international soil and water conservation …

ISSN 2095-6339

CN 10-1107/P

Available online at www.sciencedirect.com

INTERNATIONAL SOIL AND WATER CONSERVATION RESEARCH (ISWCR)

Volume 8, No. 4, December 2020

CONTENTS

Contents continued on inside back cover

Special issue: Soil erosion assessment tools and data; creation, consolidation, and harmonization – A special issue from the Global Symposium on Soil Erosion 2019 (Rome, FAO HQ).

Guest Editors: Richard Cruse, Costanza Calzolari, Lucia Anjos, Nigussie Haregeweyn and Clara Lefèvre

International Soil and Water C

onservation Research (ISW

CR

) Vol. 8 (2020) 333 – 452

84


C. Lef è vre , R.M. Cruse , L.H. Cunha dos Anjos , C. Calzolari , N. Haregeweyn Guest editorial – soil erosion assessment, tools and data: A special issue from the Global Symposium on soil Erosion 2019 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333

M. Biddoccu , G. Guzm á n , G. Capello , T. Thielke , P. Strauss , S. Winter , J.G. Zaller , A. Nicolai , D. Cluzeau , D. Popescu , C. Bunea , A. Hoble , E. Cavallo , J.A. G ó mez

Evaluation of soil erosion risk and identif cation of soil cover and management factor (C) for RUSLE in European vineyards with different soil management. . . . . . . . . . . . . . . . . . . . . . . . 337

C.M. Rosskopf , E. Di Iorio , L. Circelli , C. Colombo , P.P.C. Aucelli Assessing spatial variability and erosion susceptibility of soils in hilly agricultural areas in Southern Italy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354

C. Wang , L. Shan , X. Liu , Q. Yang , R.M. Cruse , B. Liu , R. Li , H. Zhang , G. Pang Impacts of horizontal resolution and downscaling on the USLE LS factor for different terrains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363

T. Yue , Y. Xie , S. Yin , B. Yu , C. Miao , W. Wang Effect of time resolution of rainfall measurements on the erosivity factor in the USLE in China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373

M.J. Marques , A. Alvarez , P. Carral , B. Sastre , R. Bienes The use of remote sensing to detect the consequences of erosion in gypsiferous soils . . . . . 383

M. Gharibreza , M. Zaman , P. Porto , E. Fulajtar , L. Parsaei , H. Eisaei Assessment of deforestation impact on soil erosion in loess formation using 137 Cs method (case study: Golestan Province, Iran) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393

J.L. Peralta Vital , R.H. Gil Castillo , Y. Llerena Padr ó n , Y.M. Cordovi Miranda , N. Labrada Arevalo , L. Alonso Pino

Integrated nuclear techniques for sedimentation assessment in Latin American region . . . . . 406

R. Torres Astorga , Y. Garcias , G. Borgatello , H. Velasco , R. Padilla , G. Dercon , L. Mabit Use of geochemical f ngerprints to trace sediment sources in an agricultural catchment of Argentina . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410

ISWCR_v8_i4_COVER.indd 1ISWCR_v8_i4_COVER.indd 1 12/5/2020 2:56:09 AM12/5/2020 2:56:09 AM

AIMS AND SCOPEThe International Soil and Water Conservation Research (ISWCR), the off cial journal of World Association of Soil and Water Conservation (WASWAC) www.waswac.org, is a multidisciplinary journal of soil and water conservation research, practice, policy, and perspectives. It aims to disseminate new knowledge and promote the practice of soil and water conservation. The scope of International Soil and Water Conservation Research includes research, strategies, and technologies for prediction, prevention, and protection of soil and water resources. It deals with identif cation, characterization, and modeling; dynamic monitoring and evaluation; assessment and management of conservation practice and creation and implementation of quality standards.

Examples of appropriate topical areas include (but are not limited to):∑ Soil erosion and its control∑ Watershed management ∑ Water resources assessment and management∑ Nonpoint-source pollution∑ Conservation models, tools, and technologies∑ Conservation agricultural∑ Soil health resources, indicators, assessment, and management∑ Land degradation∑ Sedimentation ∑ Sustainable development ∑ Literature review on topics related soil and water conservation research

© 2020 International Research and Training Center on Erosion and Sedimentation and China Water and Power Press. Production and hosting by Elsevier B.V.Peer review under responsibility of IRTCES and CWPP.

NoticeNo responsibility is assumed by the International Soil and Water Conservation Research (ISWCR) nor Elsevier for any injury and/or damage to persons, property as a matter of product liability, negligence, or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein.Although all advertising material is expected to conform to ethical standards, inclusion in this publication does not constitute a guarantee or endorsement of the quality or value of such product or of the claims made of it by its manufacturer.

Full text available on ScienceDirect

Volume 8, Number 4, December 2020

International Soil and Water Conservation Research (ISWCR)

International Research and Training Center on Erosion and Sedimentation and China Water and Power Press

Special issue: Soil erosion assessment tools and data; creation, consolidation, and harmonization – A special issue from the Global Symposium on Soil Erosion 2019 (Rome, FAO HQ).

Guest Editors: Richard Cruse, Costanza Calzolari, Lucia Anjos, Nigussie Haregeweyn and Clara Lefèvre

INTERNATIONAL SOIL AND WATER CONSERVATION RESEARCH (ISWCR)


CONTENTS

C. Lef è vre , R.M. Cruse , L.H. Cunha dos Anjos , C. Calzolari , N. Haregeweyn Guest editorial – soil erosion assessment, tools and data: A special issue from the Global Symposium on soil Erosion 2019 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333

M. Biddoccu , G. Guzm á n , G. Capello , T. Thielke , P. Strauss , S. Winter , J.G. Zaller , A. Nicolai , D. Cluzeau , D. Popescu , C. Bunea , A. Hoble , E. Cavallo , J.A. G ó mez

Evaluation of soil erosion risk and identif cation of soil cover and management factor (C) for RUSLE in European vineyards with different soil management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337

C.M. Rosskopf , E. Di Iorio , L. Circelli , C. Colombo , P.P.C. Aucelli Assessing spatial variability and erosion susceptibility of soils in hilly agricultural areas in Southern Italy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354

C. Wang , L. Shan , X. Liu , Q. Yang , R.M. Cruse , B. Liu , R. Li , H. Zhang , G. Pang Impacts of horizontal resolution and downscaling on the USLE LS factor for different terrains. . . . . . . . 363

T. Yue , Y. Xie , S. Yin , B. Yu , C. Miao , W. Wang Effect of time resolution of rainfall measurements on the erosivity factor in the USLE in China . . . . . . . 373

M.J. Marques , A. Alvarez , P. Carral , B. Sastre , R. Bienes The use of remote sensing to detect the consequences of erosion in gypsiferous soils . . . . . . . . . . . . . 383

M. Gharibreza , M. Zaman , P. Porto , E. Fulajtar , L. Parsaei , H. Eisaei Assessment of deforestation impact on soil erosion in loess formation using 137 Cs method (case study: Golestan Province, Iran) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393

J.L. Peralta Vital , R.H. Gil Castillo , Y. Llerena Padr ó n , Y.M. Cordovi Miranda , N. Labrada Arevalo , L. Alonso Pino

Integrated nuclear techniques for sedimentation assessment in Latin American region . . . . . . . . . . . . . 406

R. Torres Astorga , Y. Garcias , G. Borgatello , H. Velasco , R. Padilla , G. Dercon , L. Mabit Use of geochemical f ngerprints to trace sediment sources in an agricultural catchment of Argentina. . 410

P. Tsymbarovich , G. Kust , M. Kumani , V. Golosov , O. Andreeva Soil erosion: An important indicator for the assessment of land degradation neutrality in Russia. . . . . . 418

B. Liu , Y. Xie , Z. Li , Y. Liang , W. Zhang , S. Fu , S. Yin , X. Wei , K. Zhang , Z. Wang , Y. Liu , Y. Zhao , Q. Guo The assessment of soil loss by water erosion in China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430

Q. Yang , M. Zhu , C. Wang , X. Zhang , B. Liu , X. Wei , G. Pang , C. Du , L. Yang Study on a soil erosion sampling survey in the Pan-Third Pole region based on higher-resolution images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440

Editorial

Guest editorial e soil erosion assessment, tools and data: A specialissue from the Global Symposium on soil Erosion 2019

a b s t r a c t

This special issue on soil erosion assessment, tools and data creation, consolidation and harmonizationpresents advances in soil erosion research with a focus on new tools that are being used to assess soilerosion rates. This publication includes eleven selected contributions presented at the Global Symposiumon Soil Erosion (GSER, 15e17 May 2019, Rome, Italy) dealing with erosion indicators’ improvement, theuse of remote sensing, nuclear techniques and geochemical fingerprinting as promising methods toassess soil losses, management practices that reduce soil erosion in vineyards and olive groves planta-tions and their modelling, and national and regional erosion assessments.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-

ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Background

Five years after the release of the Status of the World’s Soil Re-sources report identifying soil erosion as one of the ten major soilthreats (FAO & ITPS, 2015), soil erosion is still widely recognizedas a major driver of soil degradation, soil biodiversity loss and soilorganic carbon losses. Accelerated soil erosion rates are expectedto continue in the coming decades as a result of anthropogenic ac-tivities, among them agriculture and land cover changes, andclimate change (Borrelli et al., 2020; FAO, 2019b; Guerra et al.,2020; Naipal et al., 2018; Olson et al., 2016). However, it is alsoknown that in most erosion-prone regions, rates of human-induced soil erosion can be decreased by adopting soil conservationpractices and sustainable soil management (SSM) measures, resto-ration programmes or natural recovery after land abandonment(e.g. Gharibreza et al., this issue; Tsymbarovich et al., this issue).The role of policies in scaling up SSM measures is also highlightedby national (e.g. in Spain, van Leeuwen et al., (2019), in China, Xieet al. (2019),), regional (e.g. European Union, Turpin et al. (2017),),and international agreements and programmes, such as the Volun-tary Guidelines for Sustainable Soil Management (FAO, 2017) or theLand Degradation Neutrality concept (UNCCD, 2016). The FAO alsopredicts that reducing and restoring eroded soils would foster theachievement of at least half of the Sustainable Development Goals(SDGs) (FAO, 2019a). Our greatest opportunity to meet these soilconservation needs revolves around improved science allowing ac-curate estimates of soil erosion losses at increased spatial and tem-poral resolutions, policy informed by better soil erosion science,and data availability/sharing between partners striving to achievethis common goal (FAO, 2019a).

2. Global Symposium on Soil Erosion (GSER), 15e17 May 2019

This special issue was proposed after the Global Symposium onSoil Erosion (GSER), held at the Food and Agriculture Organizationof the United Nations (FAO of the UN) headquarters, Rome, Italy on15e17 May 2019. More than 500 participants from 104 countriesattended the symposium, where about 100 oral and 30 posterswere presented (FAO, 2019a). The GSER was organized by theFAO’s Global Soil Partnership (GSP) and its Intergovernmental Tech-nical Panel on Soils (ITPS), together with the Science-Policy Inter-face (SPI) of the UN Convention to Combat Desertification(UNCCD), and the Joint FAO/IAEA Programme of Nuclear Tech-niques in Food and Agriculture. This science-policy meeting wasseparated in three different themes targeting the different aspectsof soil erosion as follows (1) Use of data and assessment tools in soilerosion control, (2) Policy in action to address soil erosion and (3)The economics of soil erosion control and restoration of erodedland. The manuscripts proposed in this special issue are focusedon the 1st theme and aim to address the objective of identifying op-tions to consolidate, generate and harmonize soil erosion data andassessment tools for promoting their use in decision making at alllevels.

3. This issue

This special issue compiles eleven research articles coming fromAsia, Europe and Latin America. The eleven papers deal withdifferent aspects of erosion assessment and modelling and provideexamples of the ongoing research and research needs. The studiespublished in this special issue were initially selected by the GSERScientific Committee (composition available in FAO (2019a)), and

Contents lists available at ScienceDirect

International Soil and Water Conservation Research

journal homepage: www.elsevier .com/locate/ iswcr

International Soil and Water Conservation Research 8 (2020) 333e336

https://doi.org/10.1016/j.iswcr.2020.11.0042095-6339/© 2020 International Research and Training Center on Erosion and Sedimentation and China Water and Power Press. Production and Hosting by Elsevier B.V. Thisis an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

invited after a call for interest launched by the GSP Secretariat inNovember 2019 (“Evaluation of soil erosion risk and identificationof soil cover and management factor (C) for RUSLE in Europeanvineyards with different soil management”, Bidoccu et al.; “Assess-ing spatial variability and erosion susceptibility of soils in hilly agri-cultural areas in Southern Italy”, Rosskopf et al.; “Impacts ofhorizontal resolution and downscaling on the USLE LS factor fordifferent terrains”, Wang et al.; “Effect of time resolution of rainfallmeasurements on the erosivity factor in the USLE in China”, Yueet al.; “The use of remote sensing to detect the consequences oferosion in gypsiferous soils”, Marques et al.; “Assessment of defor-estation impact on soil erosion in loess formation using 137Csmethod (case study: Golestan Province, Iran)”, Gharibreza et al.;“Integrated nuclear techniques for sedimentation assessment inLatin American region”, Peralta Vital et al.; “Use of geochemical fin-gerprints to trace sediment sources in an agricultural catchment ofArgentina”, Astorga et al.; “Soil erosion: An important indicator forthe assessment of land degradation neutrality in Russia”, Tsymbar-ovich et al.; “The assessment of soil loss by water erosion in China”,Liu et al.; “Study on a soil erosion sampling survey in the Pan-ThirdPole region based on higher-resolution images”, Yang et al.)

This issue presents research addressing approaches that havebeen developed for soil erosion assessment (e.g., soil surveys, fieldsoil sediment mapping, remote sensing analyses, and sedimentfingerprinting), measuring (e.g., laboratory and plot experiments,laboratory soils and sediment analysis, plot to catchment moni-toring, and tracer dating techniques), and modelling (e.g., concep-tual, statistical, and physically based models).

This issue provides insights on the improvement of the inputsused by the USLE family models. Indeed, at present, USLE and theRevised USLE (RUSLE) are by far the most widely applied soilerosion estimation models globally, and many well-known modelsare based on the USLE/RUSLE technology, such as the Chinese SoilLoss Equation (CLSE) have been developed and tested (Alewellet al., 2019). However, despite its relative simplicity, proper (R)USLE calibration for a given location outside the existing areaswithin which calibration has occurred can be challenging, espe-cially in situations outside of those widely covered in the USA. Inthis issue, four new studies propose improvements in the measureof the RUSLE/USLE factors: soil cover and management (C), erod-ibility factor (K), slope length and slope steepness factor (LS) andrainfall erosivity (R) (Biddoccu et al., this issue; Rosskopf et al.,this issue; Wang et al., this issue; Yue et al., this issue, respectively).

The management factor (C) is important to consider especiallyin locations prone to erosion, where soil management has majorimpacts on erosion rates, such as in vineyards. Indeed, vineyardsare the second most extended perennial crops in the EU28, andare often located on sloping areas and managed with bare soils,which make them experience very high erosion rates of 30 t yr�1

ha�1 or more in certain areas (Casalí et al., 2009; Tropeano &Ploey, 1983); in such cases, management can play an importantrole in reducing runoff. Biddoccu et al. (2016) showed for examplethat the use of cover crops in vineyards could reduce soil loss by upto 58%. Biddoccu et al. (this issue) propose an analysis to predicterosion rates at the hillslope scale across different wine growing re-gions in Europe, using the RUSLE approach with long-term experi-mental data and farm-based information. The erodibility factor (K)parameterises the susceptibility of a soil to erode and is related tosoil properties such as soil organic matter content, texture, struc-ture and permeability. At the European level, Panagos et al.(2014) estimated the K factor at high resolution, but at smallerspatial scales, the low density of topsoil sampling points used inthe map of Panagos et al. (2014) may not correctly or exhaustivelyrepresent the soil features and related soil erodibility. To increasethe accuracy of the K-factor at more detailed scales, Rosskopf

et al. (this issue) propose an assessment of the spatial variabilityof soil properties and soil erodibility in hilly agricultural areas ofSouth Italy and investigate the linkages between soil features andlandscape morphodynamics. In hilly terrains, the LS factor is a crit-ical soil loss parameter. It is usually derived from a digital elevationmodel (DEM), and consequently, its precision depends on the pre-cision of the DEM.Wang et al. (this issue) propose amethodology toimprove the accuracy of the LS factor. Wang et al. (this issue) stressthat most of the research addressing LS factor differences associ-ated with DEM resolution variations focused on local areas and/oron watersheds with similar topographic characteristics, whichmay explain conflicting results on the optimal resolution to choose.Wang et al. (this issue) aim to identify the influence of 5-m to 30-mresolution DEMs on LS factor estimations in varied terrains, andevaluate their data in order to build a downscale model that willallow improving the accuracy of the LS factor estimate. Rainfallerosivity is the soil erosion factor that has gained most attentionduring the last decade and much research has been done toimprove erosivity indexes (Ballabio et al., 2017; Nearing et al.,2017; Panagos et al., 2017). As part of this issue, Yue et al. providean initial study addressing the effects of time interval resolution(from 1 to 60-min) of rainfall measurements on the erosivity factorin the USLE in China. Indeed in the USLE, RUSLE and derivedmodelssuch as CSLE, with the development of automatic observations,high-temporal resolution fixed-interval rainfall data from 1 to 60-min are increasingly available and are believed to improve the esti-mation of rainfall erosivity, especially the 1-in-10-year storm event(EI30) index (Yue et al., this issue).

Remote sensing is also known as one the most important toolsto monitor soil changes and has been extensively used for decadesfor erosion detection, assessment of erosion controlling factors anddata integration for erosion mapping (Vrieling, 2006). In this issue,Marques et al. make use of remote sensing to monitor soil changesand the consequences of soil erosion through its impact on soilorganic carbon, available water and gypsum content, plot slope,and olive tree development on a gypsiferous plot use for olivetree cropping in the centre of Spain.

Innovative techniques such as the use of tracers, through nu-clear techniques or geochemical fingerprints are proven quickand efficient to monitor soil losses and increase our knowledgeabout the status and resilience of agro-ecosystems. For this specialissue, two studies used Caesium-137 (137Cs) isotopes and the inte-grated application of 3 nuclear techniques (Fallout Radionuclides(137Cs, Beryllium-7 (7Be), Lead-210 (210Pb)), Compound Specific Sta-ble Isotopes (fatty acids of the v13C) and Isotopic Hydrology (18O,2H, 3H) (Gharibreza et al., this issue; Peralta Vital et al., this issue),while one study used the geochemical fingerprint approach usingEnergy Dispersive X-Ray Fluorescence (EDXRF) (Astorga et al., thisissue). In North-East Iran, Gharibreza et al., (this issue) used Cs iso-topes to determine the on-site impacts of deforestation and the ef-fects of afforestation. The East of Hircanian forest indeed sufferedfrom severe deforestation starting in 1963 (200 000 ha) for agricul-ture and wood exploitation, while afforestation actions started in1993. Peralta Vital et al., (this issue) implemented three nucleartechniques in 21 countries of Latin America and the Caribbean re-gion to measure and monitor sedimentation in water reservoirs.The integrated use of fallout radionuclides allows to evaluate andquantify (1) the redistribution of soil in a study region to determineareas of soil lost (erosion) and deposition (sedimentation), (2) theorigin of the deposited soil, and (3) the water dynamics (surfaceand underground). Astorga et al. (this issue) used geochemical fin-gerprints to identify sediment sources in a small mountain catch-ment in a semiarid region of central Argentina, that has beensubject to a rapid land-use change for the 30 last years, with theintense conversion from native vegetation to agriculture and

C. Lef�evre, R.M. Cruse, L.H. Cunha dos Anjos et al. International Soil and Water Conservation Research 8 (2020) 333e336

334

livestock production. The study showed that feedlots are the mainsources of sediments in most of the analysed channel sedimentmixtures, followed by riverbanks and dirt roads. Likewise, range-lands and pastures were shown as a relevant source of sediments,while the native vegetation presents, as expected, one of the lowestcontributions to soil erosion in the studied area. Linkages are alsomade between soil erosion and contaminations due to livestock ef-fluents that appear as a rising issue in the area.

Eventually two national and one regional soil erosion assess-ments are presented. In Russia, Tsymbarovich et al. (this issue) pro-vide a review of different erosion data measured and modelled atdifferent scales and the authors estimate that soil erosion by waterfrom the cultivated land is about 0.56 billion tons leading to afertility loss of 30e50% from the 1950s to the 1980s. Erosion rateshave been decreasing, mainly as a result of land abandonmentand climate changes in the last decades. The authors highlightthe relevance of adding a soil erosion dynamics indicator to com-plement the three existing land degradation neutrality (LDN) indi-cators. At the global level the state of LDN is indeed evaluatedthrough using the SDG indicator 15.3 “Proportion of land that isdegraded over total land area” and its three sub-indicators (changesin land cover, land productivity and soil organic carbon). Accordingto the authors, erosion assessment appears to be a good option toreport to the LDN, under the Sustainable Development Goal(SDG) 15.3, when detailed and accurate SOCmonitoring is not avail-able in a country, such as in Russia. The position of Tsymbarovichet al. (this issue) supports the numerous papers also highlightingthe importance of soil erosion as an additional subindicator forassessing LDN, as well as an alternative to the SOC indicator(Cowie et al., 2019; Gilbey et al., 2019; Ifejika Speranza et al.,2019; Kapovi�c Solomun et al., 2018). Liu et al. (this issue) proposean assessment of water erosion in China using CSLE in 10 � 10 mgrids and the use of spatial interpolation at national, river basin,and provincial scales. The results are based on a national erosionsurvey conducted in 2011. The water erosion rate was estimatedat 5 t ha�1 yr�1 in China, and the rate was more than tripled onsloping and cultivated croplands. It is estimated that 13.5% of theChinese territory is prone to intense erosion and needs to imple-ment soil conservation practices. The survey results and the erosionmaps provide basic information for national conservation planningand policymaking. Finally, Yang et al., (this issue) propose a newmethodology of erosion survey, based on high-resolution remotesensing images in the Third Pan-Pole (the region that encompassesthe Eurasian highlands and areas hydrologically affected by them).This approach addresses the issue of the complexity, time and costsof leading soil erosion surveys based on direct measurements.Through the visual interpretation of the free high-resolutionremote sensing images, detailed information on land use and soilconservation measures was obtained. The information was thencoupled with soil erosion factor data products to calculate soilloss rates with the CSLE model.

4. Take home messages

These studies exemplify the complexity associated with soilerosion estimates and soil erosion policy intended to move us to-wards sustainable development goals. The literature is rich withsoil erosion research, yet soil erosion caused land degradation re-mains critically high (Borrelli et al., 2017; Doetterl et al., 2012;Montgomery, 2007), which was the fundamental reason for hold-ing the GSER. Moreover, the recent study of Borrelli et al. (2020)showed that soil erosion could increase greatly around the worldover the next 50 years due to climate change and land-use dy-namics. Factors affecting soil erosion are spatially and temporallydynamic across the globe; the spatial and temporally dynamic

nature of the earth’s surface make accurate erosion estimation intime and space extremely difficult. Policy targeting soil conserva-tion practices must accommodate human dynamics that are alsoculturally variable and economically challenging. Addressing thespatial and temporal variations associated with the physical envi-ronment coupled with challenges predicting human response tothe policy will require the development of dynamic models toaccommodate these complexities. Spatial and temporal datarequired for more sophisticated modelling of the physical environ-ment allowing better soil erosion estimates necessary for informedpolicy are only marginally available; to make major improvementsin soil loss estimates and resulting policy, higher resolution spatialdatabases must be developed and made freely available to soil sci-entists. Building bridges to improve our soil erosion knowledge bysharing science between international partners is a very positivestep. However, investing in data development will bring the nextmajor advance in our science of soil erosion and the improvementof global soil resources.

Declaration of competing interest

The authors declare no conflict of interest.

Acknowledgements

The Guest Editors thank the contributors of the Special Issue fortheir contribution and efforts towards this publication. The Guesteditors are thankful to the Editor-In-Chief Mark Nearing and the ex-ecutive editor Paige Chyu, for their support and the opportunity tohost this special issue, and aknowledge the fruitful collaboration.

References

Alewell, C., Borrelli, P., Meusburger, K., & Panagos, P. (2019). Using the USLE: Chan-ces, challenges and limitations of soil erosion modelling. International Soil andWater Conservation Research, 7(3), 203e225. https://doi.org/10.1016/j.iswcr.2019.05.004

Ballabio, C., Borrelli, P., Spinoni, J., Meusburger, K., Michaelides, S., Beguería, S.,Klik, A., Petan, S., Jane�cek, M., Olsen, P., Aalto, J., Lakatos, M., Rymszewicz, A.,Dumitrescu, A., Tadi�c, M. P., Diodato, N., Kostalova, J., Rousseva, S., Banasik, K.,& Panagos, P. (2017). Mapping monthly rainfall erosivity in Europe. The Scienceof the Total Environment, 579, 1298e1315. https://doi.org/10.1016/J.SCITOTENV.2016.11.123

Biddoccu, M., Ferraris, S., Opsi, F., & Cavallo, E. (2016). Long-term monitoring of soilmanagement effects on runoff and soil erosion in sloping vineyards in AltoMonferrato (NortheWest Italy). Soil and Tillage Research, 155, 176e189.https://doi.org/10.1016/j.still.2015.07.005

Borrelli, P., Robinson, D. A., Fleischer, L. R., Lugato, E., Ballabio, C., Alewell, C.,Meusburger, K., Modugno, S., Schütt, B., Ferro, V., Bagarello, V., Oost, K. V.,Montanarella, L., & Panagos, P. (2017). An assessment of the global impact of21st century land use change on soil erosion. Nature Communications, 8(1),2013. https://doi.org/10.1038/s41467-017-02142-7

Borrelli, P., Robinson, D. A., Panagos, P., Lugato, E., Yang, J. E., Alewell, C.,Wuepper, D., Montanarella, L., & Ballabio, C. (2020). Land use and climatechange impacts on global soil erosion by water (2015-2070). Proceedings ofthe National Academy of Sciences, 117(36), 21994e22001.

Casalí, J., Gim�enez, R., De Santisteban, L., �Alvarez-Mozos, J., Mena, J., & Del Valle deLersundi, J. (2009). Determination of long-term erosion rates in vineyards ofNavarre (Spain) using botanical benchmarks. Catena, 78(1), 12e19. https://doi.org/10.1016/j.catena.2009.02.015

Cowie, A. L., Waters, C. M., Garland, F., Orgill, S. E., Baumber, A., Cross, R.,O’Connell, D., & Metternicht, G. (2019). Assessing resilience to underpin imple-mentation of land degradation neutrality: A case study in the rangelands ofwestern new South Wales, Australia. Environmental Science & Policy, 100,37e46. https://doi.org/10.1016/j.envsci.2019.06.002

Doetterl, S., Van Oost, K., & Six, J. (2012). Towards constraining the magnitude ofglobal agricultural sediment and soil organic carbon fluxes. Earth Surface Pro-cesses and Landforms, 37(6), 642e655. https://doi.org/10.1002/esp.3198

Fao. (2017). Voluntary Guidelines for sustainable soil management (p. 15). Rome, Italy:155th Session of the FAO Council, 12.

Fao. (2019a). Outcome document of the global symposium on soil erosion (p. 28). Foodand Agriculture Organization of the United Nations. http://www.fao.org/3/ca5697en/ca5697en.pdf.

Fao. (2019b). Soil erosion: The greatest challenge for sustainable soil management.


335

Food and Agriculture Organization of the United Nations.Fao, & Itps. (2015). Status of the World’s soil resources (main report) (p. 608). Food

and Agriculture Organization of the United Nations, Intergovernmental Tech-nical Panel on Soils. http://www.fao.org/3/a-i5199e.pdf.

Gilbey, B., Davies, J., Metternicht, G., & Magero, C. (2019). Taking Land DegradationNeutrality from concept to practice: Early reflections on LDN target setting andplanning. Environmental Science & Policy, 100, 230e237. https://doi.org/10.1016/j.envsci.2019.04.007

Guerra, C. A., Rosa, I. M. D., Valentini, E., Wolf, F., Filipponi, F., Karger, D. N., NguyenXuan, A., Mathieu, J., Lavelle, P., & Eisenhauer, N. (2020). Global vulnerability ofsoil ecosystems to erosion. Landscape Ecology, 35(4), 823e842. https://doi.org/10.1007/s10980-020-00984-z

Ifejika Speranza, C., Adenle, A., & Boillat, S. (2019). Land degradation neutrali-tydpotentials for its operationalisation at multi-levels in Nigeria. Environ-mental Science & Policy, 94, 63e71. https://doi.org/10.1016/j.envsci.2018.12.018

Kapovi�c Solomun, M., Barger, N., Cerda, A., Keesstra, S., & Markovi�c, M. (2018).Assessing land condition as a first step to achieving land degradation neutrality:A case study of the Republic of Srpska. Environmental Science & Policy, 90, 19e27.https://doi.org/10.1016/j.envsci.2018.09.014

van Leeuwen, C. C., Cammeraat, E. L., de Vente, J., & Boix-Fayos, C. (2019). The evo-lution of soil conservation policies targeting land abandonment and soil erosionin Spain: A review. Land Use Policy, 83, 174e186.

Montgomery, D. R. (2007). Soil erosion and agricultural sustainability. Proceedings ofthe National Academy of Sciences, 104(33), 13268e13272. https://doi.org/10.1073/pnas.0611508104

Naipal, V., Ciais, P., Wang, Y., Lauerwald, R., Guenet, B., & Oost, K. V. (2018). Globalsoil organic carbon removal by water erosion under climate change and landuse change during AD 1850 e 2005. Biogeosciences, 15(July), 4459e4480.https://doi.org/10.5194/bg-15-4459-2018

Nearing, M. A., Yin, S., Borrelli, P., & Polyakov, V. O. (2017). Rainfall erosivity: An his-torical review. Catena, 157, 357e362. https://doi.org/10.1016/j.catena.2017.06.004

Olson, K. R., Al-Kaisi, M., Lal, R., & Cihacek, L. (2016). Impact of soil erosion on soilorganic carbon stocks. Journal of Soil and Water Conservation, 71(3), 61Ae67A.https://doi.org/10.2489/jswc.71.3.61A

Panagos, P., Ballabio, C., Meusburger, K., Spinoni, J., Alewell, C., & Borrelli, P. (2017).Towards estimates of future rainfall erosivity in Europe based on REDES andWorldClim datasets. Journal of Hydrology, 548. https://doi.org/10.1016/j.jhydrol.2017.03.006

Panagos, P., Meusburger, K., Ballabio, C., Borrelli, P., & Alewell, C. (2014). Soil erod-ibility in Europe: A high-resolution dataset based on LUCAS. The Science ofthe Total Environment, 189e200. https://doi.org/10.1016/j.scitotenv.2014.02.010.s 479e480.

Tropeano, D., & Ploey, D. (1983). Soil erosion on vineyards in the Tertiary piemon-tese basin (northwestern Italy). Studies on experimental areas in Rainfall simu-lation, runoff and soil erosion. Catena. An Inderdisciplinary Journal ofGeomorphology, Hydrology, Pedology Giessen, 10(4), 115e127.

Turpin, N., ten Berge, H., Grignani, C., Guzm�an, G., Vanderlinden, K., Steinmann, H.-H., Siebielec, G., Spiegel, A., Perret, E., Ruysschaert, G., Laguna, A., Gir�aldez, J. V.,Werner, M., Raschke, I., Zavattaro, L., Costamagna, C., Schlatter, N., Berthold, H.,

Sand�en, T., & Baumgarten, A. (2017). An assessment of policies affecting Sustain-able Soil Management in Europe and selected member states. Land Use Policy,66, 241e249. https://doi.org/10.1016/j.landusepol.2017.04.001

Unccd. (2016). The LDN target setting programme. United Nations Convention toCombat Desertification. https://www.unccd.int/actions/ldn-target-setting-programme.

Vrieling, A. (2006). Satellite remote sensing for water erosion assessment: A review.Catena, 65(1), 2e18. https://doi.org/10.1016/j.catena.2005.10.005

Xie, Y., Lin, H., Ye, Y., & Ren, X. (2019). Changes in soil erosion in cropland in north-eastern China over the past 300 years. Catena, 176, 410e418. https://doi.org/10.1016/j.catena.2019.01.026

Clara Lef�evre*

Global Soil Partnership Secretariat, Food and Agriculture Organizationof the United Nations, Viale delle Terme di Caracalla, Rome, Italy

Richard M. CruseDepartment of Agronomy, Iowa State University, Ames, IA, 500011,

USA

Lucia Helena Cunha dos AnjosDepartment of Soils, Federal University of Rio de Janeiro, Serop�edica,

RJ, 23890-000, Brazil

Intergovernmental Technical Panel on Soils, Food and AgricultureOrganization of the United Nations, Rome, Italy

Costanza CalzolariCNR Institute for Bioeconomy, Sesto Fiorentino, Italy

Intergovernmental Technical Panel on Soils, Food and AgricultureOrganization of the United Nations, Rome, Italy

Nigussie HaregeweynInternational Platform for Dryland Research and Education, Tottori

University, 1390 Hamasaka, Tottori, 680-0001, Japan

* Corresponding author.E-mail address: [email protected] (C. Lef�evre).

10 November 2020Available online 19 November 2020


336

Original Research Article

Evaluation of soil erosion risk and identification of soil cover andmanagement factor (C) for RUSLE in European vineyards withdifferent soil management

M. Biddoccu a, *, G. Guzm�an b, G. Capello a, T. Thielke c, P. Strauss c, S. Winter d, J.G. Zaller e,A. Nicolai f, D. Cluzeau f, D. Popescu g, C. Bunea h, A. Hoble h, E. Cavallo a, J.A. G�omez b

a Institute for Agricultural and Earthmoving Machines (IMAMOTER), National Research Council of Italy (CNR), Torino, Italyb Institute for Sustainable Agriculture, CSIC, Cordoba, Spainc Institute for Land and Water Management Research, Federal Agency for Water Management, Petzenkirchen, Austriad Institute of Plant Protection and Institute of Integrative Nature Conservation Research, University of Natural Resources and Life Sciences, Vienna, Austriae Institute of Zoology, University of Natural Resources and Life Sciences Vienna, Vienna, Austriaf Universit�e Rennes 1, Station Biologique de Paimpont, UMR 6553 EcoBio, 35380, Paimpont, Franceg SC JIDVEI SRL, Research Department, Jidvei, Romaniah University of Agriculture Science and Veterinary Medicine, Cluj Napoca, Romania

a r t i c l e i n f o

Article history:Received 16 March 2020Received in revised form3 July 2020Accepted 7 July 2020Available online 17 July 2020

Keywords:VineyardErosionSoil managementRUSLEEurope

a b s t r a c t

Vineyards show some of the largest erosion rates reported in agricultural areas in Europe. Reported ratesvary considerably under the same land use, since erosion processes are highly affected by climate, soil,topography and by the adopted soil management practices. Literature also shows differences in the effectof same conservation practices on reducing soil erosion from conventional, bare soil based, management.The Revised Universal Soil Loss Equation (RUSLE) is commonly adopted to estimate rates of water erosionon cropland under different forms of land use and management, but it requires proper value of soil coverand management (C) factors in order to obtain a reliable evaluation of local soil erosion rates. In thisstudy the ORUSCAL (Orchard RUSle CALibration) is used to identify the best calibration strategy againstlong-term experimental data. Afterwards, ORUSCAL is used in order to apply the RUSLE technology fromfarm based information across different European wine-growing regions. The results suggest that thebest strategy for calibration should incorporate the soil moisture sub-factor (Sm) to provide better soilloss predictions. The C factor, whose average values ranged from 0.012 to 0.597, presented a large spatialvariability due to coupling with local climate and specific local management. The comparison across thefive wine-growing regions indicates that for the soil protection management, permanent cover crop isthe best measure for accomplishing sustainable erosion rates across the studied areas. Alternate andtemporary cover crops, that are used in areas of limited water resources to prevent competition withvines, failed to achieve sustainable erosion rates, that still need to be addressed. This raises the need for acareful use of C values developed under different environmental conditions.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction

Vine is the second most extended perennial crop across Europewith 3.2 million ha in the EU28, only surpassed by olives (G�omez,

2017), but, as opposed to other perennial crops that tend to belocated in specific regions, vines are grown in southern and centralEurope, covering a broad range of soils and climate conditions.Vineyards are usually managed with agronomic practices, such astillage or weed control with herbicides, that leave the bare soilexposed to rainfall and runoff during all or part of the year. Thesepractices result in a degradation of soil structure and in acceleratedwater erosion, particularly for those located on slopes (Ruíz-Colmenero et al., 2013; Salom�e et al., 2016). Although the

* Corresponding author. Institute for Agricultural and Earthmoving Machines(IMAMOTER), National Research Council of Italy, Strada delle Cacce, 73, 10135,Torino, Italy.

E-mail address: [email protected] (M. Biddoccu).






reported erosion rates in vineyards across Europe present a largevariability (Prosdocimi, Cerd�a, & Tarolli, 2016) they tend to behigher than tolerable soil loss levels when measured in vineyardson sloping areas managed with bare soil. For instance, Tropeano(1983) measured soil losses of 33 t ha�1 year�1 in a deeplyplowed vineyard in Piemonte, and (Casalí et al. (2009)) measuredan average long-term erosion rate of 30 t ha�1 year�1 in N Spain onvineyards managed with tillage.

The use of cover crops in vineyards is widely considered as aneffective agricultural conservation measure, providing variousecosystem services such as reduction of runoff and erosion pro-cesses, increasing soil organic matter, weed control, pest and dis-ease regulation, water supply, water purification, improvement offield trafficability, and conservation of biodiversity (Garcia et al.,2018; Winter et al., 2018; Hall et al., 2020). Several studies indi-cated a large reduction of erosion rates in vineyards when a covercrop is used. In Italy, after a 14-years monitoring experiment,Biddoccu et al. (2016) demonstrated how the use of a permanentgrass cover reduced the average soil losses on large runoff plots oneorder of magnitude, from 13.85 t ha�1 year�1 (average of the twotillage treatments considered) to 1.8 t ha�1 year �1. Measurementscarried out during 3-year long experiment from large runoff plotson vineyards located in Portugal and France reported 58% reductionin soil losses, from an average of 48.1 t ha�1 y�1 on tilled plots to27.9 t ha�1 y�1 in the cover crop plots (G�omez et al., 2011). Therelatively large losses reported for the cover crop treatment byGomez et al. (2011) was explained by the occurrence of an extremestorm in one of the experiments. Furthermore, it was difficult toimplement a good temporary ground cover during all the rainyseason, due to the need tominimize competition for soil water withthe wines under Mediterranean dry conditions. Due to divergentsoil and climate conditions in wine-growing areas, there are dif-ferences in the implementation of agronomical practices acrossEurope that can be overlooked in the literature. In the Mediterra-nean region wine growers are reluctant to use permanent covercrops due to concerns over soil water competition with the vine,which is supported by experimental data (Celette et al., 2008; Ruiz-Colmenero et al., 2011). As a result, most wine growers that usecover crops in Mediterranean conditions tend to establish bare soilvineyards or use temporary cover crops, defined by Salom�e et al.(2016) as those providing ground cover between 4 to 7 monthsper year. In contrast, in more humid conditions in central Europe,permanent cover crops are more commonly implemented.

Evaluation of soil losses due to the adoption of different soilmanagement techniques in vineyards has been carried out withdifferent methods (Rodrigo Comino, 2018), and results can help inaddressing local soil management policies in the vine-growingsector. Among the adopted methodologies, direct soil loss mea-surements require someyears of observations, adequate equipment,they are limited to some selected vineyards, and the results are notalways comparable because obtained at different temporal andspatial scale. Furthermore, most of the available studies and resultsare related to some specific areas in some European countries (andfew in the rest of the world), thus they don’t cover the high vari-ability of soil, topographical, management, and climate conditionsthat characterize the European wine-growing regions. An alterna-tive to experimental measurements to evaluate the impact ofchanges in soil management on soil erosion risk under contrastingsituations is the analysis with simulation models. The Revised Uni-versal Soil Loss Equation RUSLE (Renard, Foster, Weesies, McCool,&Yoder, 1997) remains as the most widely used tools for this.

The model estimates average annual erosion using rainfall, soil,topographic, and management, using the following equation:

A ¼ R * K *C * LS *P

where A: annual average soil loss (t ha�1 yr�1), R: rainfall erosivityfactor (MJ mm ha�1h�1yr�1), K: soil erodibility factor (t ha hha�1MJ�1mm�1), C: cover-management factor (dimensionless), LS:slope length and slope steepness factor (dimensionless), and P:support practices factor (dimensionless).

Despite its relative simplicity, proper RUSLE calibration for agiven situation outside the modeling community can be chal-lenging, especially in situations outside of those widely covered inthe USA. To overcome this problem some authors have proposedsimplified procedures, such as the summary model proposed byG�omez et al. (2003) for the use of RUSLE in olives, the determina-tion of USLE C factors from cumulative field measurement soilerosion in vineyards (e.g. Novara et al., 2011) or calibration of theparameters required to determine the USLE C (soil cover andmanagement) factor (Auerswald & Schwab, 1999). However, thevalues derived following these approaches are sometimes difficultto extrapolate beyond the conditions for which they were devel-oped. For instance, Novara et al. (2011) developed their C values asthe ratio of soil losses between the cover crops and the bare soilplots, providing no C values for bare soil management. Auerswaldand Schbwab, 1999 provide a comprehensive evaluation of annualcrop and management factors (C) values from bare soil by tillage topermanent cover crop. Nevertheless, their values are valid forrainfall conditions similar to those or which they were developed,given the interaction with seasonal rainfall erosivity in C-factordetermination (Auerswald et al., 2015). Some of the assumptionsmade in the determination of these C factors can be far from theassumptions and equations used by the most recent RUSLE version2 (Dabney et al., 2012), that has been incorporated in the recentwater erosion map of Europe by Panagos et al. (2015c). In that studythe authors obtained C factors for tree crops, such as vineyards,based on the calibration from the fraction of ground cover derivedfromMERIS satellite images, as explained in Panagos et al. (2015b).In this approximation they update the seasonal C values based on alinear relationship between a range of maximum Cmax (for zeroground cover) andminimum Cmin (for full ground cover) with theseextreme values taken as 0.45 and 0.15 respectively from a review ofpublished values. The analysis carried out by Panagos et al. (2015b)resulted in an average C values for vineyards across Europe of 0.35.

To our knowledge, no attempt has been made to evaluate RUSLEbased on long-term soil loss measurements in vineyards, neither acalibration strategy based on a comprehensive analysis of vineyardconditions across Europe. Our hypothesis is that this evaluation canprovide relevant information to orientate its future applications forlocal and regional studies, as well as to provide insight for adaptingcurrent soil conservation strategies in vineyards, best suited for thedifferent environmental conditions in Europe. This manuscriptpresents a study with these objectives:

1 To present an Excel tool to facilitate calibration of RUSLE2 invineyards in hillslope applications.

2 To evaluate the performance of several calibration strategies ofRUSLE for permanent cover crop and bare soil by tillage using along-term experimental record previously published.

3 To identify C values and to compare soil erosion risk for themostcommon soil management operations in four different winegrowing areas in Europe (S. Spain, NW France, NW Italy, EAustria and central Romania)

2. Materials and methods

2.1. Model description

An EXCEL tool, ORUSCAL (Orchard RUsle CALibration), was built

M. Biddoccu et al. / International Soil and Water Conservation Research 8 (2020) 337e353338

to allow the calibration of RUSLE2 by assuming the simplified sit-uation of a hillslope with a regular slope profile for tree crops suchas vines, olives, almonds or citrus (G�omez Calero, Biddoccu, &Guzm�an, 2020). The structure and main input data of the tool isshown in Fig. 1. Each of the factors of RUSLE is calculated in one ofthe Excel sheets, e.g soil erodibility (K), or in several ones from eachof the corresponding subfactors, e.g. cover and management factor(C). Each factor is accompanied by a set of explanations andaccompanying information to allow their determination, indi-cating, when necessary, sources of this information (for examplerainfall erosivity) or data taken from the RUSLE2 database (e.g. rootbiomass for different orchard management system). This allowstheir use in situations with limited data in a straightforward wayfor users with limited programming or modelling training, facili-tating the use of a combination of field measurements (for instanceevaluation of ground cover) complemented with data from othersources. It makes it a convenient use for education or training onevaluation of water erosion through RUSLE, facilitating the under-standing the erosion risk of different management by stakeholders,or other applications such as comparison of erosion risk amongdifferent management, soil, temperature and rainfall regimes.ORUSCAL can be downloaded with its instructions from:https://digital.csic.es/handle/10261/216656. For more complex situations,e.g. complex slope profiles, any interested user should use the fullRUSLE2 software.

Within this approach, for the determination of the rainfallerosivity factor (R factor), there is the option of using the quartererosivity for the location of interest which are usually availablefrom several sources (e.g., ICONA, 1988; Panagos et al., 2015a),calculating this from the correlation of daily rainfall and dailyrainfall erosivity, which is also a common approach where thesecorrelations have been developed (e.g. Domínguez Romero, AyusoMu~noz, & García Marín, 2007) since daily rainfall is a commonlyavailable data or upload the quarter rainfall erosivity calculatedexternally from high resolution rainfall data using software such asRIST (Rainfall Intensity Summarization Tool) (ARS-USDA, 2015).

For the determination of soil erodibility (K factor), ORUSCALuses the standard RUSLE equation based on the textural composi-tion and organic matter of the soil, corrected by soil infiltrationcapacity and soil structural class. It allows the use of a K valuesexternally calculated, for instance from the global pedotransferfunction of Borselli et al. (2012) or the soil erodibility map of Europeby Panagos et al. (2015a). From this baseline value, it gives theopportunity to use this as constant annual value or use theempirical correction proposed by RUSLE2 to calculate a quarter Kvalue, based on daily Ta and rainfall (USDA-ARS, 2013).

For the determination of the slope steepness (S) and length (L)factors, the simplified model uses the standard equations used byRUSLE2, differentiated for steepness lower or greater than 9%(standard USLE unit plot steepness). The steepness of most of theinvestigate plots is lower than 18%, the range for which the standardRUSLE2 equations have been developed, but for those that aresteeper, for the shake of simplicity, we have kept the original RUSLEequations, which implies a slight overestimation (Liu et al., 1994).There is also an option to attribute a value to the m factor, used fordetermination of L to incorporate the relative importance of rill andinterrill erosion, or let the model calculate it based on the ratio of rillto interrill erodibilty and the evolution of ground cover using theequations described in RUSLE2 documentation (USDA-ARS, 2013).

For the determination of the cover and management factor (C),ORUSCAL calculates each of the seven subfactors from which thisfactor depends, giving the option (as done by RUSLE2) of notconsidering some of them in the determination. The first subfactoris the soil consolidation subfactor, Sc which takes into account thedecrease in soil erodibility with time after a mechanical

disturbance (like tillage). It can be used in the transition periodfrom tillage to a management with no mechanical disturbance ofthe soil surface, such as a permanent cover crop, but in most situ-ations it is not used. The second considered subfactor is the canopycover subfactor, Cc which takes into account the reduction of soilerosion due to the provision of ground cover by living vegetation. Itis calculated from the height and shape of this vegetation (which inthe case of the vineyards can be the vines plants and the herba-ceous vegetation in the lanes) and the ground cover provided bythis vegetation in each quarter. The third subfactor is the surfaceroughness subfactor, Sr, which includes the decrease of erosionwith increasing roughness. This section provides values of therecommended roughness and disturbance values for the mostcommon tillage operations and surface conditions from the RUSLE2database and it calculates quarter Sr values considering reductiondue to rainfall after tillage. The soil biomass subfactor, Sb, considersthe reduction of erosion due to the presence of residues and rootbiomass within the top 25 cm of the soil. Since this is a value that itis usually difficult to obtain, the information sheet includes infor-mation on the amount of root (live and dead) and residues for themost common situations in vineyards and other tree crops. The fifthsubfactor is the ground cover subfactor, Gc, which considers thereduction of erosion due to the presence of mulch or residues overthe soil surface directly in contact with it. The ridge subfactor, hr,considers the increase in erosion due to flow concentrationoccurring when there are well formed ridges oriented in up-downslope. Although it is not commonly used, it is included to provide allthe possibilities allowed by RUSLE2. Furthermore, in vineyards therow orientation assumes a primary role in forcing the direction oftractor traffic and tillage implementation and thus ridge formation.In any case, it must not to be confused with the contour tillagewhich is included in the support (P) factor. The final subfactor, Sm, isa subfactor which is optional in RUSLE to introduce the modifica-tion of the erosion risk due to changes in the soil moisture content,been higher when close to saturation and close to zero when it isclose to totally dry. It cannot be used in combination with theseasonal adjustment of soil erodibility mentioned above, and themodel give the opportunity to use one of these two corrections ornone. The ORUSCAL formulation gives the opportunity to introducemeasured soil moisture values or to estimate these values usingdaily rainfall and ETo (or estimating these using Hargreaves equa-tion) and the ET from the vineyard determined using the FAOmethod (Allen et al., 1998). The use of this factor has improved thedetermination of C values for olive trees in Mediterranean condi-tions (Gomez et al., 2003) which has resulted in erosion predictioncapable to reproduce the experimental measurements (Marin et al.,2014). For this reason, the model incorporates explicitly this option.Once all these factors are calculated ORUSCAL presents their annualand quarter values for each RUSLE factor, as well as the calculatedsoil loss.

2.2. Experimental data for evaluation

The Italian soil erosion dataset was derived from the CannonaData Base (Biddoccu et al., 2016) collected in the Tenuta CannonaErosion Plots. The monitored plots are part of a larger vineyard,within the Experimental Vine and Wine Centre of Agrion Founda-tion, which is located in the Alto Monferrato hilly area of Piemonte,North-West Italy. The climate is Hot-summerMediterranean (Csa inthe K€oppen climate classification, Kottek et al., 2006). At the studysite the average annual precipitation in the period 2000e2016 was852 mm, mainly concentrated in October, November and March,with the driest month been July. The Cannona vineyards lies onPleistocenic fluvial terraces in the Tertiary Piedmont Basin,including highly altered gravel, sand and silty-clay deposits, with

M. Biddoccu et al. / International Soil and Water Conservation Research 8 (2020) 337e353 339

red alteration products. Soil is classified as Typic Ustorthents, fine-loamy, mixed, calcareous, mesic (Soil Survey Staff, 2010).

The measurements concern two vineyard plots on a hillslopewith SE aspect, that are managed in according conventionalfarming for wine production. The two plots have the same length,width and slope, but the inter-rows have been differently managedfor the 14 monitored years, with conventional tillage (CT) andpermanent grass cover (GC), as summarized in Table 1. Each plotwas hydraulically bounded: a channel at the top of the plotscollected upstream water, in order to measure separately runoffand sediment yield generated from each rainfall event. The mea-surements have been conducted since 2000 and they are currentlyongoing, with methods indicated in Table 1 and fully explained inBiddoccu et al. (2016). For this analysis a careful evaluation of theoperation management and experimental records was made usingonly the years inwhich therewas complete information to calibrateRUSLE without large uncertainty in any of the required information(such as for instance number of tillage operations, etc …).

2.3. Evaluation of different calibration strategies

One of the objectives of the study was to evaluate the perfor-mance of different calibration strategies of RUSLE among those thatmight be used. For this purpose, a subset of the Cannona dataset,which passed the requirements of having all the available infor-mation, was used in the calibration step of the model (years 2004,2005, 2007, 2008, 2010, 2011, 2012, 2013). The calibration wasperformed following four different calibration strategies describedin Table 2, which in short result from using a constant or variable K-factor and considering, or not, the soil moisture subfactor, Sm.

The best calibration of themodel was determined on the basis ofboth quarter and annual values, through the following statistics:the efficiency coefficient of Nash and Sutcliffe (NSE), root meansquare error of residuals (RMSE) and the coefficient of

determination of the linear regression (R2), see Nash and Sutcliffe(1970) and Moriasi et al. (2007).

2.4. Calibration of RUSLE for five wine growing regions acrossEurope

Afterwards, ORUSCAL was fully calibrated for different kind oftypical, or possible, soil management in five contrasting winegrowing regions across Europe. This calibration was madecombining field surveys to characterize vineyard conditions, pub-licly available data and information provided by wine growers.These regions were: S Spain, NW France, NW Italy, E Austria andCentral Romania. In all cases each farm and management wassimulated for 15 or 16 consecutive years to capture temporal vari-ability in rainfall.

2.4.1. Brief description of the five study areaThe five study areas include vineyards from wine-growing re-

gions located from the southern and western to the eastern coun-tries of Europe (Fig. 2), and ranging from a variety of soils,landscapes and climatic conditions, that represent the variety ofgrowing conditions for grapevine across Europe (Table 3). Theseareas are: a) the Montilla-Moriles wine-growing region in southernSpain (Andalusia), b) the Coteaux-du-Layon in Anjou, in north-western France (Loire Valley), the Carnuntum and Leithaberg re-gion in eastern Austria (Lower Austria and Burgenland), theTarnave wine-growing region in Central Romania (Transylvania),and the Alto Monferrato and Gavi wine-growing regions in northItaly (Piedmont). The climate varies from the summer-dry (Spain)and hot-summer Mediterranean (Italy) climate, to temperateoceanic (Austria) andwarm-summer humid continental (Romania).Viticultural landscapes also differ: from viticulture and olivedominated landscape in Andalusia (Spain) and the sloping vine-yards dominating the hilly landscape of the Italian area to the

Fig. 1. Structure of the simplified calculation procedure in ORUSCAL (left side) and essential input data with indication of factors/subfactors where they are used (right side).


vineyards stretching from flat areas up to the forests at the hill-topswith interspersed arable fields, semi-natural elements in E Austria,the valley of the river Layon with very steep slopes (up to 60%) andthe plateaux area with vine and crop fields in NW France, andfinally the vineyards mixed with other agricultural crops, pasturesand woods in the Romanian area. The investigated vineyards lie ona variety of soils, from the finest clay to clay-loam soils of theMonferrato, from sandy silt to sandy loam soils in eastern Austria,from clay to sandy-loam in Transylvania and in the DGO Coteaux duLayon.

2.4.2. Description of the vineyards and management operations invineyards of the study areas

The investigated vineyards were chosen to represent the mostcommon vineyard management practices in each region. Table 4summarizes the main vineyard characteristics and managementoptions, including the specifications of the soil management tech-niques employed in the area. In all study areas tillagewas simulatedas following the maximum slope as this was the dominant imple-mentation when plowing was used in the region. In Spain, France,Austria and Romania, detailed determination of soil management,soil properties and ground cover was carried out experimentally fora whole season for two group of farms. One including farmsapplying the usual low soil protection management (with 7 or 8farms in this group depending on the country) and the secondgroup implementing common the soil protection management(with 7 or 8 farms depending on the country) in the respective area(Figs. 2 and 3). In Italy, the dataset includes 7 vineyards with lowprotective management and 5 vineyards with high protectivemanagement (Figs. 2 and 3). This resulted, as expected, in a mul-tiplicity of soil management strategies that can be grouped in 5classes: by decreasing intensity: 1- bare soil obtained by herbicideapplication and no tillage (NT); 2- bare soil using conventionaltillage (CT); 3- partial soil cover, using temporary cover crops (TCC),defined as those grown during the rainy (fall and winter) seasonkilled at the onset of the dry season period (early spring); 4- or

partial cover using alternate cover crops (ACC) defined as perma-nent cover crop every two lanes with the intercalated lane havingthe CT typical of the area; 5- permanent cover crops (PCC), resultingin complete cover of the inter-rows. Management information wascollected in the form of personal semi-structured interviews withwine growers in all countries but in Italy, where this wasmadewithdirect consultation to farm managers. These interviews allow toclassify the existing management as high or low soil protectingstrategies. Furthermore, interviews provided information about theperception of farmers and stakeholders of these practices, and theidentification of two additional soil management techniques thathad been implemented recently in the region or that it might beintroduced (or reintroduced) in the future, see Table 4. For eachmanagement scenario and identified vineyard soil losses werepredicted using the ORUSCAL to apply the RUSLE2 methodologythrough 16 consecutive years (2000e2015) for Spain, Austria andRomania and France, and for 15 years (2004e2018) in Italy,obtaining a total of 2246 years of simulated scenarios, 1097 forcurrently used scenarios and 2149 years for hypothetical scenarios.

2.4.3. Sources of information for calibration of the different RUSLEparameters

Daily temperature and rainfall depth was obtained from aweather station located within each of the study area for the wholestudy period. To minimize bias due to different calibration of therainfall erosivity, we standardize the rainfall erosivity among areasto the long-term average R values (annual and monthly) providedby Ballabio et al. (2017). To do this, we calculated daily rainfallerosivity from daily rainfall using an exponential function R ¼ aPb,adjusted to match long-term monthly and annual average withthose of Ballabio et al. (2017). We did not used this approach for theevaluation of calibration strategies, where daily rainfall erosivitywas calculated using the RUSLE2 methodology (Dabney et al.,2012).

The topographic, soil, cover and management informationrequired for calibration of ORUSCAL for the study areas other than

Table 1Summary of the experiment used to evaluate the calibration strategies for ORUSCAL.

Location Carpeneto (AL), 296 m asl, 44�40057.4500N, 8�37035.2400E

Soil texture Clay to Clay-loamPlot slope & size slope: 15%

area:1221 m2

length: 74 m, width: 16.5 mManagement Conventional Tillage (CT), depth 0.25 m,

cultivated in spring and autumnControlled Grass (GC), spontaneous vegetation mulched in spring and autumn.In autumn 2011, the inter-rows of the GC plot were tilled and a grass mixture was sown

Soil losses 7 (±12.5) 1.8 (±1.6)Average (± st.dev)(t ha¡1 year¡1)Rainfall measurements Automatic rainfall gauge (resolution 0.2 mm), hourly dataRunoff measurements Runoff is collected at the bottom of each plot by a channel,

connected to a sedimentation trap and then to a tipping bucket device (0.1 mm resolution)to measure the hourly volumes of runoff from each plot.

Soil loss measurements A portion of the runoff-sediment mixture is sampled for each tip and, after each erosive event, a 1.5 Lsample of runoff-sediment mixture was collected. Sediments deposited along drains and inthe sedimentation traps are also collected and dry-weighed

Table 2Summary of the four calibration strategies evaluated.

K factor sm subfactor

Option 1 constant (standard RUSLE nomograph) ConsideredOption 2 constant (standard RUSLE nomograph) not consideredOption 3 variable, using the empirical function based on Ta and rainfall included in

RUSLE and baseline K calculated from soil properties from the nomographnot considered

Option 4 constant, calculated from soil properties based on the model proposed by Borselli et al. (2012) Considered


Fig. 2. Locations of study areas in Spain, France, Italy, Austria, and Romania. Symbols in each box indicate the investigated vineyards, with the currently implemented soilmanagement (non protective or protective for soil). The background shows the land use according to CORINE land cover.

Table 3Basic information of the five wine growing study regions. For each study region average values, standard deviation (in brackets) and minimum and maximum values areindicated for soil organic matter content (OM), vineyards length and slope.

Country Coordinates(extremes of thearea)

Wine-growingregion(administrativeregion/district)

Climate (K€oppenclimateclassification)

Landscape and landuse

Prevalent soil andOM (%,average ± SD)

Vineyards length(m)

Vineyards slope (%)

Average (±SD) Average (±SD)

Elevation (range) Min-Max Min-Max

Spain 37� 380-290N, 4�

450-310WMontilla-Moriles(Andalusia)

Hot-summerMediterraneanclimate (Csa)

Agricultural area.Rugged relief, 220e682m a.s.l.

Clay to sandy-loamOM: 1.6 (±0.3)

43.7 (±64.8) Min:53.4, Max: 279

5.8 (±3.3)Min: 0.5,Max: 11.0

Italy 44�39’ e 44�520 N,8�37’ e 8�540 E

Alto Monferrato,Gavi (Piemonte)

Hot-summerMediterraneanclimate (Csa)

Mix of vineyardswith arable fieldsand woods, hillyregion with flatareas, 270e300ma.s.l.

Clay to clay-loamOM: 1.2 (±0.3)

87.9 (±32.9)Min:72.4, Max: 169

16.7 (±6.9)Min: 5.0,Max: 26.3

France 47.41e47.04N, 0.85e0.24W

Coteaux-du-Layonin Anjou (LoireValley )

Temperate oceanicclimate (Cfb)

16% Vineyards(1435 ha for theDGO label, total10900ha), 44%arable fields withmainly cereals andsun flowers orpastures, 30% semi-natural elements,8% urban areas, 2%water

Sandy-LoamOM:2.0 (±0.7)

106.3 (±43.5)Min:50.0, Max: 175.0

7.1 (±8.8)Min: 0.4,Max: 32.3

Austria 47�54-48�70N,16�380-16�730E

Carnuntum andLeithaberg (LowerAustria andBurgenland)

Temperate oceanicclimate (Cfb)

Mix of vineyardsand arable fields,hilly region withflat areas, 120e235m a.s.l.

sandy silt to sandyloamOM: 3.2 (±1.0)

301.3 (±100.5),Min: 160.0, Max:440.0

8.6 (±5.3), Min: 2.0,Max: 15.0

Romania 46�020-46�190N,23�430-24�010E

Arud and Tarnave(Transylvania)

Temperatecontinental climate(Dfb)

Vineyardssurrounded byother agricultural,forest andseminatural areas,280e440m a.s.l.

Clay to sandy-loamOM: 1.4 (±0.9)

168.8 (±92.9) Min:54.0, Max: 296.0

8.9 (±4.8)Min: 0.7,Max: 17.9


Italy was taken from field survey carried out during the VineDiversproject (https://short.boku.ac.at/vinedivers). Information for Italianvineyards derived from previous studies (Biddoccu et al., 2016;Bagagiolo, Biddoccu, Rabino, & Cavallo, 2018; Capello et al., 2019),field surveys, interviews to farmers and vineyard’s technicians. Soilmeasurements were used to calculate the erodibility (K-factor) andincludes, for each investigated vineyard, soil particle fraction andorganic matter content, class of soil structure and soil permeability.

Topographic data, namely field length and steepness, necessary todefine the LS-factor, were obtained by GIS analysis or Google Earth,in the gaps where it was not possible to be measured directly in thefield. The different farm-derived management information wasused to calibrate the different C subfactors. So Cc and Gc subfactorswere determined from measurements made during a wholegrowing season (see for instance Guzm�an et al., 2019) or estimatedfrom available measurements in hypothetical scenarios projecting

Table 4Characteristics of vineyards evaluated in each of the five study areas, soil management, type and number of real and simulated scenarios. Number in brackets are number offarms simulated for each management scenario.

Country Vine-growingmanagement (training,mechanization)

V arieties Row orientation (andwidth)

Mechanical soilmanagementtechniques,operation/tools,depth, times

Managementtechnique family(number ofvineyards)

Number of realscenarios (singlemanagement x yearx vineyard)

Managementtechnique familyand number ofsimulated scenarios

Spain Globet and trellis,mechanized

Pedro Ximenez Up-and-down (2.85 m) Conventionaltillage, cultivator,0.15 m, 1-4

CT (8) 128 NT (16), 256

Temporary covercrops, cultivator,0.15 m, 1e2;mowing, 1e2.

TCC (8) 128 PCC (16), 256

Italy Single Guyot,mechanized (tired andtracked tractors)

BarberaDolcettoGavi Up-and-down (2.5e2.75 m)

Conventionaltillage, ripper,0.25 m, 2 - 3

CT (7) 105 ACC (6), 90

Reduced tillage,rotocultivator,0.15 m, 2Spontaneouscontrolled grass,mulching, na, 2- 3

PCC (5) 75

France Trellis, mechanized Large variety from redto white wine varieties

Up-and-down Bare soil withherbicide, 2�4

NT (7) 112 CT (15), 240

Spontaneouscontrolled grass,mulching, na, 3

PCC (8) 128 ACC (15), 240

Austria Trellis, mechanized Large variety from redto white wine varieties

Up-and- down Tillage in everysecond inter-rowwith different typesof machinery(chisel, disc harrow,rotary, plough),0.05e0.15 m, 1e3;mulching in theother inter-row, 2-5

ACC (8) 128 NT (16), 256; CT(16), 256

Permanent covercrop, mulching inevery inter-row, 2-5

PCC (8) 128

Romania Trellis, mechanized Large variety from redto white wine varieties

Up-and-down (2.0e3.0 m)

Bare soilmanagementthrough frequentsoil tillage in allinter-row. Tillagedepths rangedbetween 5 and30 cm and thenumber ofapplication per year(in inter-row)ranged between 2e5 time/year acrossthe vineyards.

CT (8) 128 NT (16), 240

Alternate covercrop. The inter-rowvegetation cover(alternative) wasmowing/mulchedapproximately 2e4times per year

ACC (7) 112 PCC (16), 240

Abbreviations in bold are management scenarios considered by stakeholders at each area less soil protecting scenarios, and those in italics more soil-protecting scenarios.Abbreviation used for soil managements are: NT¼ no tillage, CT¼ conventional tillage, TCC¼ temporary cover crop, ACC¼ alternate cover crop, PCC¼ permanent cover crop.


expected vegetation and residue cover. Soil management opera-tions were used to determine the soil consolidation and soilroughness subfactors (Sc and Sr, subfactors of the C-factor). Dataabout residues and biomass, also needed to calculate the Sb sub-factor, were taken fromRUSLE database. The soil moisture (Sm) sub-factor was calculated internally by ORUSCAL using the simplifiedwater balance model incorporated in the model using daily rainfalland ET, based on Allen et al. (1998).

2.5. Data statistical analysis

Differences among regions in values of SL and RUSLE factors forthe currently adopted non-protecting and protecting soil manage-ments were determined using ANOVA (with a ¼ 0.05). The same

test was used to compare SL and C-factor among regions for thesame management class (currently adopted or hypothetical), andamongmanagement for each studied region. For identifying factorsthat play the major role in differentiating the currently adopted soilmanagement in the considered study areas an exploratory analysisusing a Principal Component Analysis (PCA) was performed. Thiswas complemented with stepwise multiple lineal regression (MLR)to evaluate the relative contribution of each variable to differencesin determining soil losses. The cumulative probability distributionfunctions of soil loss and C factors were determined. In all casesSTATA SE 14.0 was used.

3. Results and discussion

3.1. Evaluation of calibration strategies

The comparison of the statistics obtained with the four differentcalibration strategies (Table 5) showed that the performance of themodel was unsatisfactory (Moriasi et al., 2007) for the two option 2and 3, both for yearly and for quarter results (NSE<0). The intro-duction of variable K in option 3 (using the empirical functionbased on daily temperature and rainfall, included in RUSLE, andbaseline K calculated from soil properties) improved the perfor-mance of the model with respect to the model with constant K(option 2, Table 3) but remains performing poorly in predicting soillosses. Using the two options that include the determination of asoil-moisture subfactor, option 1 and 4, the performance of themodel improved both for yearly and for quarter values. The simu-lation resulted in predictions that give acceptable values for NSE(between 0.35 and 0.82) and satisfactory values for RMSE (>1/2 sm),with exception of yearly values for CT (NSE ¼ �0.48). The bestperformance of the model was obtained for CT, quarter values,using option 4, that resulted in NSE ¼ 0.82, RMSE ¼ 0.34 t ha�1,R2 ¼ 0.82. Considering the annual values the same option gaveNSE ¼ 0.67, RMSE ¼ 2.44 t ha�1, R2 ¼ 0.79. The performance wasgenerally poorer for GC, with satisfactory results that were ob-tained only with option 1, for yearly values (NSE ¼ 0.58,RMSE¼ 0.44 t ha�1, R2 ¼ 0.65) (Table 5). The corresponding quarterresults resulted in NSE ¼ 0.43, RMSE ¼ 0.10 t ha�1, R2 ¼ 0.46.

It is apparent, in our evaluation analysis, that the introduction ofthe temporal evolution of soil moisture improved the predictions,and that this temporal impact cannot be introduced using theempirical correction offered as a possibility in the K factor (option3). This contrasts with the results of Khalegpanah et al. (2016) whoobserved a considerable improvement of RUSLE2 model efficiencyusing a calibrated and variable K-factor. This difference might berelated to the different climate conditions of our study areas, morecontrasting to those of the empirical K value adjustment of RUSLE2than those of the study of Khalegpanah et al. (2016). It is also worthnoting that the better predicting capabilities of RUSLE2 incorpo-rating the Sm subfactor for Mediterranean conditions was alsonoted by Marin et al. (2014) for a similar analysis in olive orchards.

Comparing the model performance between options 1 and 4considering the Sm subfactor, the model predicted the low values oferosion in quarters better than for options 2 and 3, and regressionlines showed intercept (n) and slope (m) closer to 0 and 1,respectively, than those obtained without considering the Sm sub-factor. However, for option 1, the improvement in prediction of lowamounts of soil losses was not enough to give good annual valuesfor CT, that weremore overpredicted than in option 4 and led to lowperformance of the model. The performance of the model consid-ering quarter values for GC was similar to those obtained forquarters with WEPP (Zhang et al., 1998), and with models of theUSLE family for event values (Di Stefano et al., 2016; Khaleghpanahet al., 2016; Kinnell, 2017; Spaeth, 2003). Quarter results in CT were

Fig. 3. Some of the investigated vineyards, in Spain, France, Italy, Austria, and Romania,representing some of the most currently implemented soil management (non/lessprotective or protective). (Photo credits: Jean-Paul Gislard for France, Davide Ferrarese(VignaVeritas) for Italy).


satisfactory for option 1, but yearly results gave negative values forthe Nash-Sutcliffe coefficient and were not satisfactory in terms ofRMSE. For yearly values simulated in the CT plot with option 4, theNSE, slope (m) and R2 are in the range of those obtainedwith RUSLE(or RUSLE2) by other authors (Risse et al., 1993; Tiwari et al., 2000;Zhang et al., 1996). The low performance of the model in predictionfor the GC plot, could be due to difficulty of the model in predictinglow amounts of soil losses, as was observed by Spaeth et al. (2003)on relatively undisturbed rangeland sites (with soil losses<0.5 t ha1). Finally, the model version using constant K obtainedfrom RUSLE2 equations, and including the Sm subfactor, was chosenas the best option to apply to the selected study areas and differentmanagement options. Overall, by introducing the Sm subfactor, themodel was able to capture the differences observed experimentallybetween the bare (CT) and ground covered (GC) treatment with anaccuracy similar in analogous exercises by other erosion models,although it shows difficulties when predicting accurately situationswith low erosion rates (quarter or GC annual values), a situationalso previously noted in erosion models by previous studies (e.g.Risse et al., 1993).

3.2. Comparison of the predicted water erosion under current soilmanagement across the five different wine growing areas

The differences in water erosion predictions using ORUSCALamong the fine wine growing areas across Europe and currentlyimplemented managements are summarized in Fig. 4 and Table 6.In this comparison there is always a soil protective management,using total or partial cover crop (PCC in Austria, France and Italy;ACC in Romania, TCC in Spain) and one that considers a non or lesssoil protective management in the study area, which were bare soil,namely NT (in France) or CT (in Romania, Spain and Italy), with theonly exception of Austria, where the less conservative managementwas ACC (partially bare soil).

The comparison of the less conservativemanagements currentlyadopted in each region provides an overview of the maximumerosion rates across these different areas, based on a modellinganalysis using the best possible farm information. In this respect, SLpredicted in different countries showed significant differences,according to 1-way ANOVA (a ¼ 0.05). The traditional vineyard’smanagement with bare soil (CT) resulted in the highest predicted

Table 5Summary of the evaluation of the calibration results.

Yearly values Quarter values

Measured Option 1 Option 2 Option 3 Option 4 Measured Option 1 Option 2 Option 3 Option 4

CTMean SL (t ha�1) 4.22 8.19 24.42 18.14 5.51 0.18 0.37 1.04 0.77 0.25St.Dev. (t ha�1) 4.20 5.89 15.52 10.6 3.95 0.79 1.03 2.75 1.95 0.7NSE e �0.48 �33.69 �14.7 0.67 e 0.64 �7.29 �2.55 0.82RMSE (t ha�1) e 4.78 23.12 15.55 2.24 e 0.47 2.27 1.48 0.34R2 e 0.80 0.77 0.71 0.79 e 0.83 0.72 0.67 0.82M e 1.26 3.24 2.13 0.83 e 1.16 2.95 2.01 0.78N e 2.89 10.77 9.16 1.99 e 0.14 0.5 0.4 0.09

GCMean (t ha�1) 0.63 0.74 2.4 2.22 0.59 0.03 0.03 0.12 0.1 0.03St.Dev. (t ha�1) 0.74 0.47 1.54 1.26 0.33 0.13 0.07 0.24 0.2 0.06NSE e 0.58 �13.41 �7.02 0.46 e 0.43 �1.99 �1.04 0.35RMSE (t ha�1) e 0.44 2.62 1.95 0.51 e 0.10 0.23 0.19 0.11R2 e 0.65 0.19 0.14 0.56 e 0.46 0.2 0.2 0.43M e 0.51 0.94 0.63 0.33 e 0.33 0.78 0.65 0.24N e 0.42 2.29 1.83 0.38 e 0.02 0.1 0.08 0.02

SL ¼ soil losses, NSE ¼ Nash-Sutcliffe efficiency, RMSE ¼ root mean square error, R2 ¼ coefficient of determination, m and n ¼ Regression coefficients: PredictedSL ¼ m*Measured SL þ n.

Fig. 4. Average annual soil losses predicted by area and by current management implemented in the area. Brown bars are non-protecting soil management and green bars are soilprotected management. Error bars indicate standard deviation. Letters indicates significant differences at a ¼ 0.05, according to ANOVA within the same class of management.Capital letters for non-soil protective management and lower case letters for soil protective management.


soil losses in Romania (22.24 ± 11.13 t ha�1 year�1), and lowestvalues in Spain (7.1 ± 6.05 t ha�1 year�1). In Austria, annual soillosses estimated with implementation of the ACC, reached 11.98(±8.83) t ha�1 year�1 even if this management considers partialcover crop. The predicted average soil losses are comparable withvalues measured at field or hillslope scale in some vineyardsmanaged with bare soil (CT or NT). Indeed, Brenot et al. (2006)estimated soil losses from 7.61 to 14.94 t ha�1 year�1 on 32-54years-old vineyards in Burgundy (France), and Gomez et al. (2011)measured soil losses from 4.47 to 90 t ha�1 year�1 over 4-yearsexperiment in different locations in southern France. These differ-ences across areas can be better understood having a closer look tothe different components of the erosion predictions, namely theRUSLE factors, which present a high variability among regions(Fig. 5) with significant differences among countries, according toANOVA (at a ¼ 0.05 significance level). Despite having a higherrainfall erosivity, Italy and Spain present farms with slight lowersoil erodibility, shorter or gentle slopes than other countries. Inaddition, CT soil management can present relatively higher groundcover due to vegetative growth between tillage passes, as noted forSpain by Guzman et al. (2019), which explain differences in thepredicted erosion values across countries and raised the need for aproper calibration based on the actual situation at the vineyards.

The exploratory analysis using the PCA for the RUSLE calibratedfactors (R, K, L, S and C) showed that 60.6% of the variance amongsoil losses can be explained by two components (Table 7). The firstcomponent PC1 represented 42.3% of the variance and the variableswith highest loadings are L (0.559) and K (0.475), followed by S andR, that showed negative correlation (�0.440 and �0.423, respec-tively). The second component PC2, represents 18.3% of the vari-ance and has a good correlation (0.868) with the C-factor. PC1 ismore related to local topographical and climate conditions, and soilcharacteristics, whereas PC2 indicates primarily the differencesrelated to soil management. Fig. 6 depicts the individual scores onPC1 and PC2 of the average simulation by farm and managementtype for the five areas. As expected there was a clear shift amongdifferent areas for similar management class (protecting and non-protecting soil) and between management classes for the samearea. Among the non-protecting soil managements, the valuesrelated to NW Italy predictions are placed on the left side of theplan, since they are associated to the lowest values for L and K andhighest R; they are followed towards the right side by the cloudsrepresenting simulations made for S Spain, Austria and centralRomania, and finally then the ones from France to the most rightside, likely following the decreasing trend of R-factor, with thelatter showing very high values also for PC1. The stepwise multiplelinear regression analysis (MLR) for non-conservative soil man-agements, Table 8A, shows that the first parameter included (theonewith the highest explanatory value) was the slope factor, linkedto local topographical conditions, followed by variables dependingon management (C-factor) and soil properties (K-factor). Rainfallerosivity and slope length were less relevant in determining soillosses in this comparative analyses across 5 areas in Europe. Theseresults suggest that soil losses predictions with adoption of lessconservative managements are not only dependent on the localconditions (topographical, soil and climate characteristics), but alsoon the specificities of local management and soil properties.Indeed, the less protective managements in most regions contem-plate bare soil during some months, have a significant role indetermining the soil erosion risk. Bare soil is obtained by herbicidesor by tillage, with a variety of passages and tools, so the resultingground cover, and associated C-factor values, showed very highvariability among but alsowithin regions. Although R-factor is a keyparameter for estimating soil erosion risk and larger soil loss pre-dictions are usually expected with higher rainfall erosivity, e.g.Ta

ble

6Ave

rage

annual

soillosses

andfactors(andstan

darddev

iation

s)fordifferents

oilm

anag

emen

tsin

each

region

.xindicates

curren

tmosta

dop

tedsoilman

agem

ents.D

ifferentletters

indicatestatisticald

ifferentv

alues

acco

rdingto

ANOVA(a

¼0.05

):capital

lettersindicatedifferencesam

ongman

agem

ents

(curren

tlyad

optedor

hypothetical)forthesamearea

,low

ercase

lettersindicatedifferencesam

ongarea

sforthesameman

agem

ent(cu

rren

tlyad

opted

orhy

pothetical).

Austria

Rom

ania

Fran

ceItaly

Spain

NT

CT

ACCx

PCCx

NT

CTx

ACCx

PCC

NTx

CT

ACC

PCCx

NT

CTx

ACC

PCCx

NT

CTx

TCCx

PCC

Annual

soilloss

67.99

21.2

11.98

3.31

32.43

22.24

11.92

1.75

18.87

17.58

9.41

1.48

34.61

17.13

8.57

1.22

26.35

7.14

9.53

1.51

t/ha

Aa

Ba

BCa

CaAb

ABa

CaDb

Ac

Aa

Ba

CbAb

Ba

CaDb

Abc

Bb

Ba

Cbst

dev

44.31

15.29

8.83

2.11

23.4

11.13

5.29

1.36

20.57

19.95

10.78

1.98

16.21

9.51

5.26

0.96

19.92

6.05

4.5

1.08

Ave

rage

rainfallerosivity,R

.37

1.9

371.9

371.9

371.9

635.4

635.4

635.4

635.4

467.8

467.8

467.8

467.8

1527

.615

27.6

1527

.615

27.6

1049

.710

49.7

1049

.710

49.7

stdev

191.9

191.9

191.9

191.9

228.7

228.7

228.7

228.7

305.7

305.7

305.7

305.7

816

816

816

816

7.8

7.8

7.8

7.8

Ave

rage

soil

0.06

70.06

40.05

10.04

0.05

80.06

0.04

50.03

30.04

10.04

60.03

90.03

60.02

90.02

50.02

40.02

40.03

70.02

70.03

10.02

2erodibility,

K.

stdev

0.02

0.02

0.01

10.00

40.02

40.03

0.01

70.01

10.00

90.01

80.01

30.00

90.00

40.00

40.00

40.00

40.00

80.00

90.00

70.00

8

Ave

rage

slope

4.73

3.59

3.22

2.81

3.26

2.66

2.42

2.17

2.57

2.15

1.99

1.82

1.99

1.28

1.22

1.18

3.01

2.42

2.56

2.07

lengthco

eff,L.

stdev

0.98

0.66

0.53

0.41

1.14

0.92

0.41

0.52

0.68

0.46

0.38

0.31

0.35

0.13

0.07

0.06

0.81

0.47

0.66

0.37

Ave

rage

slope

1.08

1.09

1.08

1.08

1.1

1.27

1.09

1.1

0.69

0.93

0.93

1.13

2.32

2.32

2.08

2.09

0.67

0.48

0.86

0.67

stee

pco

eff,S.

stdev

0.7

0.71

0.73

0.73

0.67

0.77

0.46

0.67

0.8

1.23

1.24

1.56

1.12

1.12

1.01

1.13

0.37

0.36

0.29

0.37

Ave

rage

cove

r0.59

70.24

0.15

70.07

90.26

80.22

20.12

70.03

40.58

40.54

0.29

40.04

90.16

30.13

30.07

30.01

20.32

10.21

10.13

40.04

1an

dman

ag.,C.

Aa

Bb

CbDa

Ad

Bc

CcDd

Ab

Ba

CaDb

Ae

Bd

CdDe

Ac

Bc

CcDc

stdev

0.01

30.05

60.03

0.00

60.00

50.01

20.00

50.00

50.01

40.00

40.00

70.01

0.04

70.04

0.02

30.00

30

0.04

80.06

60

Abb

reviationusedforsoilman

agem

ents

are:

NT¼

notilla

ge,C

T¼

conve

ntion

altilla

ge,T

CC¼

temporaryco

vercrop

,ACC¼

altern

ateco

vercrop

,PCC¼

perman

entco

vercrop

.


Panagos et al. (2015b) in their large scale analysis, our resultssuggest that when the characteristics of the specific farms (in thiscase vineyards) are introduced our overall vision of the relativedifferences among different areas change dramatically. This is nottotally surprising, since it is well known that hillslope scale topo-graphical factors likely have greater importance, because they canbe drivers for the relative dominance of rill erosion rather thaninterrill, especially on bare soil (Bagarello& Ferro, 2010; Bagio et al.,2017; Prosdocimi et al., 2016). Also the ground cover, a major driverin water erosion, can vary largely within apparently similar man-agement due to local specificities like (climate, seed bank, tillageintensity) that need to be properly appraised, see for instanceGuzman et al. (2019). Our results are a reminder that overlooking

these local factors can bias large scale analysis and the policy de-cisions relying on such analysis. The RUSLE technology, in our casewith a relative simple tool like ORUSCAL, can be a powerful in-strument to capture local conditions, when properly calibratedcombining several sources of information capturing the local re-ality. ORUSCAL, as a simple tool based on the RUSLE technology,allow its application to awide range of situations, namely at field orfarm level, and can be a powerful instrument to capture localconditions. Indeed, it does not necessarily need expensive ordetailed datasets, on the contrary it can be properly calibratedcombining publicly available information, field surveys andfarmers’ interviews, which in our opinion remain a fundamentalelement in this calibration phase.

Fig. 5. Average and standard deviation (error bars) of estimation of RUSLE factors value for non-soil protective (brown) and soil protective (green) soil managements in each region.Letters indicate significant differences among regions for non-soil protecting (lowercase) and soil protecting (capital) managements, according to ANOVA (a ¼ 0.05).


Table 7A) Results of the Principal Component Analysis using the 5 factors of RUSLE as variables. B) Loadings for the first two components for these five factors.

A

PC Eigenvalues %Variance Cum % Variance

PC1 2.117 0.423 0.423PC2 0.914 0.183 0.606PC3 0.812 0.162 0.769PC4 0.686 0.137 0.906PC5 0.472 0.094 1.000

B

Variable Load in PC1 Load in PC2

R-factor �0.4284 0.3414K-factor 0.4746 �0.3021L-factor 0.5589 �0.0783S-factor �0.4397 �0.1808C-factor 0.2925 0.8679

Fig. 6. Representation on components 1 and 2 of the Principal Component Analysis of predicted annual soil losses as individuals on the principal component plan, classified non-soilprotective (brown) and soil protective (green) in each study area.

Table 8Summary of the stepwise multiple linear regression models for soil losses predicted for the currently most adopted managements, for A) non-protecting and B) soil protectingmanagement.

(A)

Model for less conservative managements R2 RMSE

SL ¼ Constant þ a1 Sfactor 0.228 14.02SL ¼ Constant þ a1 Sfactor þ a2 Cfactor 0.314 13.21SL ¼ Constant þ a1 Sfactor þ a2 Cfactor þ a3 Kfactor 0.399 12.37SL ¼ Constant þ a1 Sfactor þ a2 Cfactor þ a3 Kfactor þ a4 Rfactor 0.526 11.02SL ¼ Constant þ a1 Sfactor þ a2 Cfactor þ a3 Kfactor þ a4 Rfactorþ a5 Lfactor 0.590 10.22

(B)Model for more conservative managements R2 RMSE

SL ¼ Constant þ a1 Cfactor 0.361 5.58SL ¼ Constant þ a1 Cfactor þ a2 Rfactor 0.507 4.91SL ¼ Constant þ a1 Cfactor þ a2 Rfactor þ a3 Lfactor 0.537 4.76SL ¼ Constant þ a1 Cfactor þ a2 Rfactor þ a3 Lfactor þ a4 Sfactor 0.577 4.54SL ¼ Constant þ a1 Cfactor þ a2 Rfactor þ a3 Lfactor þ a4 Sfactor þ a5 Kfactor 0.591 4.47

SL ¼ soil losses, RMSE ¼ root mean square error, R2 ¼ coefficient of determination.


A comparison of the currently used more soil-protectingmanagements can provide an assessment of the best possiblesituation in terms of soil losses reduction under current condi-tions (Table 6 and Fig. 4). As expected, results showed a reductionof the predicted annual soil losses for four of the study areas,with exception of Spain where predicted soil losses were slightlyhigher for TCC (9.53 ± 4.50 t ha�1 year�1) than for CT. This laterresult can be explained by the fact that ground cover in TCC wasextremely low in the study area, since most of the farmers optedfor natural vegetation as the alternative for cover crop and this, incombination with higher soil compaction a low nutrient content,resulted in a poor ground cover as noted by Guzman et al. (2019).In other study areas, predictions show that the implementationof PCC reduced soil losses by 87% to 93% with respect to CTmanagement, reaching values close to sustainable erosion rates,from 3.31± 2.11 in Austria, to 1.21 ± 0.96 t ha�1 year�1 in Italy.The predicted values and reduction are similar to those observedin the already cited monitoring experiment in north Italy(Biddoccu et al., 2016), where the average soil losses in a vineyardwith permanent grass cover resulted in 1.8 t ha�1 year�1,showing 74%e91% reduction with respect to tilled vineyards. Thelowest values for SL estimated with PCC management are mostlyrelated to lower C-values, lower than 0.1 only for this treatmentand up to 91% smaller than in the higher intensity managementin the same region (Fig. 5). It is also worth noting, and in the lineof the discussion above on the need to proper local calibration,that different management classes differed in some of thetopography related model parameters, because farms withdifferent management tend to be located in areas with slightlydifferent topography. In Italy PCC vineyards showed also thelowest values of K and L, in Austria and France the same factorspresented at least 12% reduction compared to the high intensitymanagement.

The exploratory analysis using PCA shows (Table 7, Fig. 6), asexpected, differences between the conservative and the non-conservative soil managements. In Fig. 6 there is a cleardistinction between management classes with the non-protecting soil management, showing both higher PC1 and PC2values. For the same area, the soil protective managements wereassociated to lower values of C-factor, resulting in lower PC2values, and also higher S-factor and lower K and L-factors,resulting in smaller value for PC1. Different distribution of pointsacross the graph is also evident among regions: on the most leftside of the graph (with negative PC1 values), the plot showsagain the Italian values (lowest values for K and L and highestvalues for R and S) together with some values from Frenchvineyards. Increasing values for PC1 are associated to pointscorresponding to SL estimations that were made for other vine-yards in France, and then Spain, Romania and Austria, which tendto overlap. Such distribution reflects primarily the increasingvalues of the L-factor, which is lower than 2 in Italy and France,and greater than 2.4 in other areas, and also the variabilityamong regions of K and R factors. Stepwise multiple linearregression (Table 8) showed that in predicting the soil losses forthe soil protecting managements, the highest determination co-efficient (and so the explanatory power) was found for the C-factor (R2 ¼ 0.383), and secondarily for average annual erosivity(increase of R2 ¼ 0.136), which is characteristics of each region,followed by local topographical conditions (L and S factors) andby soil erodibility. This reflects the fact that once achieved a highground cover, as is the case of PCC treatments, the differencesdue to management tend to be minimized across different areas(as the C values tend to converge) and differences due totopography and rainfall erosivity tend to be relevant forexplaining variability among areas.

3.2.1. Soil loss predictions for hypothetical alternative soilmanagements

This exercise allows the evaluations of the potential impact ofalternative soil management that might reduce, or increase soillosses as compared to current management, and it is summarizedin Table 6. In Romania and in Spain the simulations that consideredPCC as alternative soil management resulted in a reduction ofannual SL by 85%, with respect to those predicted with partial covercrop, reaching values close to the upper limit of the tolerable soilerosion rates (1.4 t ha�1 year�1) proposed for Europe by Verheijenet al. (2009):1. 8 ± 1.4 t ha�1 in Romania and 1.5 t ha�1 ± 1.1 inSpain. Such reductions in predicted soil losses are consistent withthose estimated for real scenarios, and above mentioned in-fieldmeasurements. In combination with the unsustainable erosionrates predicted for both real managements resulting in partial andtotal bare soil in Spain (TCC and CT) and Romania (ACC and CT),these results highlight the need to keep working in the regions forsoil management that maximize ground cover, as close as possibleto PCC. For this, considering the seeding of cover crops andimprovement of soil conditions should be a priority. Also, theyhighlight the need in continuing research in minimizing andproperly appraise the risk of competition for soil water with thevines under these improved ground cover based management (e.g.G�omez & Soriano, 2020). In all cases, as expected, the hypotheticalshift to a bare soil using herbicide management, NT, resulted in asubstantial increase in predicted soil losses as compared to the non-protective soil management currently used in the area, with in-creases of 3.7, 2.0, 1.5 times as compared to CT in Spain, Italy andRomania, respectively. This is in the range of the observations ofRaclot et al. (2009) in southern France: they compared soil lossesmeasured during 18 rainfall events in two vineyards, findingaverage soil losses 4.5 times higher in NT than in CT. Similarly, thealternative soil managements with bare soil that were simulated inAustria resulted in worsening of predicted soil erosion risk by 5.7and 1.8 times with respect to ACC, for NT and CT, respectively. Theresults suggest the relevant effect of adoption of agro-environmental schemes related to the CAP promoting soil conser-vation measures to reduce soil erosion, as those implemented inthe last 15e20 years in Austria.

3.2.2. C factor estimation for different managements and countriesThe cover and management C factor of RUSLE is the main factor

where changes due to soil management in erosion risk are intro-duced and then proper predictions can only be made if thisparameter actually conforms to the situation to be predicted. To theauthors, one of the most interesting results of our analysis in thissection is to provide a first attempt to appraise variability indetermination of this C factor at local level for a give managementdefinition, and among different areas for similar managementdefinitions. They are summarized in Fig. 7, where the cumulativeprobability distribution appears by different soil management andstudy area. In all the graphs it is apparent a lower variability forhypothetical scenarios (e.g. NT for areas other than France) whichreflects the bias towards a lower farm to farm variability thanwhenthis variability is appraised with actual farm surveys (e.g. CT inSpain). A more relevant observation from Fig. 7 is that same man-agement under different rainfall erosivity distribution presentsdifferent C values, which is not a novel result but again tend to beoverlooked in large scale studies. This is quite apparent in Fig. 7,showing the C cumulative distribution for C values for NT soilmanagement, showing clearly two different, non-overlapping,distributions for central Romania and S Spain from one side, withlower C values probably due to lower values of the Sm subfactor,and E Austria and NW France with higher C values. Italian C-factorvalues for NT show are the lowest, with higher variability than


other countries. An analogous shift in C distribution appears for PCCfor the five countries, with NW Italy showing the lowest values,then S Spain and central Romania with slightly lower variability,whereas larger variability was observed for values obtained in NWFrance and E Austria, with the latter presenting the highest values.The cumulative probability distribution of C values for ACC-TCC andCT presents a wider distribution reflecting also the higher farm tofarm variability. But despite this also presents a shift associated tothe area. This shift reflects the interaction of annual distribution ofrainfall erosivity and rainfall directly in the model in differentsubfactors, Sm, Sr, R, weighted averaged of annual C; and also theeffect of rainfall depth and temperature on differential plant andvegetation growth across areas which subsequently is incorporatedinto the model predictions. The analysis of statistically significantdifferences among C values in Table 6 shows that C values differedamong managements for the same area, but also that they tend todiffer significantly for the samemanagement among the areas. Thisresult suggests again that the variability in C values among differentareas needs to be properly addressed and interpreted, particularlybecause in most RUSLE based erosion analysis, it tends to beextrapolated from tables offering average values for various man-agement classes (e.g. Panagos et al., 2015b). Another interestingexercise is that of comparing our array of C values with thoseproposed in the literature for different soil management. Forinstance, the C-factor values calculated for PCC managements,actually happening at the vineyards, resulted lower, up to morethan 10 times smaller, than the minimum value estimated at Eu-ropean scale (Panagos et al., 2015b). The range of variability of Cvalues between protecting, and non-protecting soil managementfound in our study tend to be slightly wider than the one used byPanagos et al. (2015b) for vineyards for RUSLE application at Eu-ropean scale, 0.15e0.45, although interestingly it is not far off of

those proposed by Auerswald and Schwab (1999) for Germany, whodetermined C-factor for vineyards with different soil management,obtaining 0.59 for bare soil and 0.03 for permanent grass. Thesedifferences with some of the published values, particularly thelower C value for PCC might be the effect of differences in rainfalldistribution in comparison to the studies where previous C valueshave been proposed. Furthermore, the incorporation of specificvalues of vegetation cover, ground cover and surface roughness aswell as the incorporation of the Sm subfactor in the calibration of C,contributed to extend the range of C-factor values, something thatseems to be supported by the analysis of the calibration strategiesagainst experimental data. Although the experimental dataset usedin our analysis is small, this result suggests that this is point ofconcern in our current use of C values in vineyards across Europe,and should be further explored.

Overall, theseresults showclearly the importanceofusingC-factorobtained taking into account local management and climate, espe-cially rainfall, which allows better soil losses prediction than usingliterature values, referred to unknown or very different local andmanagement conditions, namely for a land use that is characterizedfor such high variability. Also the need to introduce uncertainty inproposedCvalues for simulationanalysis basedon farmmanagementclasses and/or ground cover determined by remote sensing should beconsidered. For instance, Baiamonte et al. (2019) estimated the inter-and intra-annual variability of C-factor in vineyards from remotesensingdata,withmethodsusingNDVI.Almagroetal. (2019)used thesame method to estimate C-factor, and then calibrated it withmeasured data: they obtained significant improvement in soil lossesestimation with RUSLE. ORUSCAL allows local estimation of C-factorfor a single vineyard, bymeans of datasets that can be easily obtained,based on farmers’ knowledge, eventually integrated with some fieldobservations, and from local weather services.

Fig. 7. Cumulative probability distribution curves for C-factor for the different management and countries.


3.3. Implications for further research, management options for winegrowers and related policies

Soil erosion risk predicted by ORUSCAL was exceeding thetolerable erosion rate for most of current and hypothetical (butpossible) scenarios, up to more than 48 times higher in the worstcase scenario. While different regions, vineyards and managementstrategies resulted in high variability of RUSLE factors and soil lossesestimation, predictions showed a good agreement with mid to long-term monitoring data at similar scale. Only with the use of perma-nent cover crop predicted soil losses were lower or close to soilformation rate in Europe. The use of partial (alternate or temporarycrops) resulted in a substantial lowering of soil erosion risk (at least43%), in agreement with the objectives of agri-environmental mea-sures that have been implemented in the last decade in some Eu-ropean countries, and also in some of the studied regions. In Austria,for example, participation rates in erosion protection measures ac-count for about 50% of the total vineyard area in Eastern Austria(BMLFUW, 2016; €OsterreichWein, 2019); in Italy,15.4% of Piedmont’sagricultural area utilized for vineyards (and orchards) adopted grasscovering for soil erosion prevention (Regione Piemonte, 2016).Indeed, a recent analysis included in the Impact Assessment of thepost 2020 Common Agricultural Policy (European Commission, CAP2021e2027, 2018) estimated the impact of cover crops on reducingsoil erosion in permanent crops by up to 37%. In addition, otherrecent studies demonstrated that lower intensity managements invineyards generally leads to improved soil properties, biodiversityand other eco-system services (Guzman et al., 2019; Winter et al.,2018). Results of this study highlight the need for maintaining and/or increasing the adoption of cover crops in order to achieve toler-able soil erosion rates in European vineyards.

To reach such objective, local soil erosion risk estimation, also atfarm level is necessary, to increase awareness of stakeholders.Models of the USLE family, especially RUSLE in more recent times,have been widely used to assess the soil erosion risk in Europeanwine growing regions, at different spatial scales (Martínez-Casasnovas & Bosch, 2000; Napoli et al., 2016; Pappalardo et al.,2019; Pijl et al., 2019; Rodrigo Comino et al., 2016). Although manyauthors have tried to use C values reflecting in the local managementconditions most adequately, in most occasion they used referencevalues that were not locally defined and calibrated. The GIS maprepresenting the C-factor assessment for EU28, at 100 m resolution(Panagos et al., 2015b), represents a widely used source for thisfactor. The estimation of C-factors obtained in the present studyshows a great variability (also vineyard to vineyard) for this factorand led to the identification of values that are not included in therange proposed in the European map for vineyards. The imple-mentation of ORUSCAL in different study areas across Europe showsthat it could be an effective strategy for the scientific community, forcalibration and identification of site-specific values for C-factor. Itcould also be used for the evaluation of soil erosion risk underdifferent intensities of soil management with limited datasets, withdemonstrative and technical purposes. Indeed, ORUSCAL is proposedas a simple and effective tool to simulate at field scale the effect ofimplementation (or not) of soil conservationmeasures on soil losses,based on datasets that can be obtained by a collaborative approachinvolving farmers and technicians. The output can assist in designingthe most appropriate soil conservation measures for their commer-cial vineyards and into disseminating best management practices.Such approach would be crucial for increasing the participation andawareness of farmers towards soil conservation measures toimplement effective soil conservation policies, as pointed byMarques et al. (2015). Indeed, the advantage of using ORUSCAL is itsability to estimate soil losses, but also the RUSLE factors taking intoaccount of the variability in soil, topographical, climate,

management conditions by means of datasets that are low-cost,easily accessible (climate data, topographical) or that can beretrieved from field observations and farmers’/technicians’experience.

4. Conclusions

To best of our knowledge, this is the first comprehensive analysisof predicting erosion rates at hillslope scale across different winegrowing regions in Europe, using the RUSLE approach with long-term experimental data and farm based information. Our findingssuggest that the best strategy for calibration should incorporate thesoil moisture subfactor (Sm), since it provided the best soil losspredictions. The comparison across the five wine-growing regionsindicates that PCC is the only soil management practice achievingsustainable erosion rates across the areas studied, whereas ACC andTCC failed to achieve this goal. Improvements in the implementationof TCC, providing higher ground cover and/or during a longer timeperiod, or expansion of PCC in areas where it is currently rarelyimplemented needs to consider the competition for soil water,particularly in more dry areas. Differences in predicted erosion ratesacross areas highlight the need to properly consider differences inclimate, topography, soil variability, and impact of management onground cover. This can only be made with a careful calibrationincorporating local specific features in order to avoid introducingmajor bias in large scale studies when extrapolating RUSLE param-eters, particularly the C factor. The C factor presented a large vari-ability due to coupling with local climate and specific localmanagement. This raises the need for a careful use of C valuesdeveloped from different conditions, during RUSLE implementation.Furthermore, it is essential to consider the farm to farmvariability inC values within the same soil management type, even within thesame area. A probabilistic approach to the C distributionmight resultin more reliable data which could address the uncertainty of theerosion predictions and the statistical significance of differences inmean values among different areas and vineyard management.

Acknowledgments

This study was funded by the European BiodivERsA projectVineDivers (https://short.boku.ac.at/vinedivers) through the Bio-divERsA/FACCE JPI (2013e2014 joint call) for research proposals,with the national funders: Austrian Science Fund (grant numbers I2044-/I 2043-/I 2042-B25 FWF), French National Research Agency(ANR), Spanish Ministry of Economy and Competitiveness (PCIN-2014-098), Romanian Executive Agency for Higher Education,Research, Development and Innovation Funding (UEFISCDI) andFederal Ministry of Education and Research (BMBF/Germany). Alsoto the CNR Short Term Mobility Program 2016 for funding a stay atIAS-CSIC during which M.Biddoccu contributed to this study andthe SHui project funded by the European Commission (GA 773903),which supported the final steps of the analysis presented in thismanuscript. We would like to thank especially all the winegrowers,the staff of the Experimental Vine and Wine Centre of AgrionFoundation and VignaVeritas, who provided access to their vine-yards as study sites and/or vineyard management information.

Appendix A. Supplementary data

Supplementary data to this article can be found online athttps://doi.org/10.1016/j.iswcr.2020.07.003.

References

Allen, R. G., Pereira, L. S., Raes, D., & Smith, M. (1998). Crop evapotranspiration


dguidelines for computing crop water requirements. FAO Irrigation and drainagepaper 56. Rome: Food and Agriculture Organization.

Almagro, A., Thom�e, T. C., Colman, C. B., Pereira, R. B., Marcato Junior, J.,Rodrigues, D. B. B., & Oliveira, P. T. S. (2019). Improving cover and managementfactor (C-factor) estimation using remote sensing approaches for tropical re-gions. Int. Soil Water Conserv. Res., 7, 325e334. https://doi.org/10.1016/j.iswcr.2019.08.005

ARS-USDA. (2015). RIST. Rainfall intensity summarization tool. Last accessed on10.03.2020 http://www.ars.usda.gov/Research/docs.htm?docid¼3251.

Auerswald, K., Fiener, P., Gomez, J. A., Govers, G., Quinton, J. N., & Strauss, P. (2015).Comment on “Rainfall erosivity in Europe” by Panagos et al. (Sci. Total Environ.,511, 801-814, 2015). The Science of the Total Environment, 532, 849e852. https://doi.org/10.1016/j.scitotenv.2015.05.019

Auerswald, K., & Schwab, A. (1999). Erosion risk (C factor) of different viticulturalpractices. Viticultural and Enological Science, 54e60.

Bagagiolo, G., Biddoccu, M., Rabino, D., & Cavallo, E. (2018). Effects of rowsarrangement, soil management, and rainfall characteristics on water and soillosses in Italian sloping vineyards. Environmental Research, 166, 690e704.https://doi.org/10.1016/j.envres.2018.06.048

Bagarello, V., & Ferro, V. (2010). Analysis of soil loss data from plots of differinglength for the Sparacia experimental area, Sicily. Italy. Biosyst. Eng, 105,411e422. https://doi.org/10.1016/j.biosystemseng.2009.12.015

Bagio, B., Bertol, I., Wolschick, N. H., Schneiders, D., dos Santos, M. A., & do, N.(2017). Water erosion in different slope lengths on bare soil. Rev. Bras. Cienc. doSolo, 41, 1e15. https://doi.org/10.1590/18069657rbcs20160132

Baiamonte, G., Minacapilli, M., Novara, A., & Gristina, L. (2019). Time scale effectsand interactions of rainfall erosivity and cover management factors on vineyardsoil loss erosion in the semi-arid area of southern Sicily. Water (Switzerland) 11.https://doi.org/10.3390/w11050978.

Ballabio, C., Borrelli, P., Spinoni, J., Meusburger, K., Michaelides, S., Beguería, S.,Klik, A., Petan, S., Jane�cek, M., Olsen, P., Aalto, J., Lakatos, M., Rymszewicz, A.,Dumitrescu, A., Tadi�c, M. P., Diodato, N., Kostalova, J., Rousseva, S., Banasik, K.,Alewell, C., & Panagos, P. (2017). Mapping monthly rainfall erosivity in Europe.The Science of the Total Environment, 579, 1298e1315. https://doi.org/10.1016/j.scitotenv.2016.11.123

Biddoccu, M., Ferraris, S., Opsi, F., & Cavallo, E. (2016). Long-term monitoring of soilmanagement effects on runoff and soil erosion in sloping vineyards in AltoMonferrato (North-West Italy). Soil and Tillage Research, 155, 176e189. https://doi.org/10.1016/j.still.2015.07.005

BMLFUW. (2016). Grüner Bericht 2016 e bericht über die Situation der€osterreichischen Land- und Forstwirtschaft (Vol. 57). Auflage, Wien).

Borselli, L., Torri, D., Poesen, J., & Iaquinta, P. (2012). A robust algorithm for esti-mating soil erodibility in different climates. Catena, 97, 85e94. https://doi.org/10.1016/j.catena.2012.05.012

Brenot, J., Quiquerez, A., Petit, C., Garcia, J. P., & Davy, P. (2006). Soil erosion rates inBurgundian vineyards. Boll. della Soc. Geol. Ital. Suppl, 6, 169e173.

Capello, G., Biddoccu, M., Ferraris, S., & Cavallo, E. (2019). Effects of tractor passes onhydrological and soil erosion processes in tilled and grassed vineyards. Water(Switzerland) 11. https://doi.org/10.3390/w11102118.

Casalí, J., Gim�enez, R., De Santisteban, L., �Alvarez-Mozos, J., Mena, J., & Del Valle deLersundi, J. (2009). Determination of long-term erosion rates in vineyards ofNavarre (Spain) using botanical benchmarks. Catena, 78, 12e19. https://doi.org/10.1016/j.catena.2009.02.015

Celette, F., Gaudin, R., & Gary, C. (2008). Spatial and temporal changes to the waterregime of a Mediterranean vineyard due to the adoption of cover cropping. Euro-pean Journal of Agronomy, 29, 153e162. https://doi.org/10.1016/j.eja.2008.04.007

Dabney, S. M., Yoder, D. C., Yoder, & Vieira, D. A. N. (2012). The application of theRevised Universal Soil Loss Equation, Version 2, to evaluate the impacts ofalternative climate change scenarios on runoff and sediment yield. Journal ofSoil and Water Conservation, 67, 343e353.

Di Stefano, C., Ferro, V., & Pampalone, V. (2017). Applying the USLE family of modelsat the sparacia (south Italy) experimental site. Land Degradation & Development,28, 994e1004. https://doi.org/10.1002/ldr.2651

Domínguez Romero, L., Ayuso Mu~noz, J. L., & García Marín, A. P. (2007). Annualdistribution of rainfall erosivity in western Andalusia, southern Spain. Journal ofSoil and Water Conservation, 62(6), 390e403.

Garcia, L., et al. (2018). Management of service crops for the provision of ecosystemservices in vineyards: A review. Agriculture, Ecosystems and Environment, 251,158e170. https://doi.org/10.1016/j.agee.2017.09.030

G�omez, J. A. (2017). Sustainability using cover crops in Mediterranean tree crops,olives and vines e challenges and current knowledge. Hungarian Geogr. Bull., 66,13e28. https://doi.org/10.15201/hungeobull.66.1.2

G�omez, J. A., Battany, M., Renschler, C. S., & Fereres, E. (2003). Evaluating the impactof soil management on soil loss in olive orchards. Soil Use & Management, 19(2),127e134. https://doi.org/10.1079/SUM2002179

G�omez, J. A., Llewellyn, C., Basch, G., Sutton, P. B., Dyson, J. S., & Jones, C. A. (2011).The effects of cover crops and conventional tillage on soil and runoff loss invineyards and olive groves in several Mediterranean countries. Soil Use &Management, 27, 502e514. https://doi.org/10.1111/j.1475-2743.2011.00367.x

G�omez, J. A., & Soriano, M. A. (2020). Evaluation of the suitability of threeautochthonous herbaceous species as cover crops under Mediterranean con-ditions through the calibration and validation of a temperature-basedphenology model. Agriculture, Ecosystems & Environment, 291. https://doi.org/10.1016/j.agee.2019.106788, 106788.

G�omez Calero, J. A., Biddoccu, M., & Guzm�an, G. (2020). ORUSCAL: RUSLE calculator

for orchards, Version 4. DIGITAL.CSIC. https://doi.org/10.20350/digitalCSIC/12552.

Guzm�an, G., Cabezas, J. M., S�anchez-Cuesta, R., Lora, Bauer, T., Strauss, P., Winter, S.,Zaller, J. G., & G�omez, J. A. (2019). A field evaluation of the impact of temporarycover crops on soil properties and vegetation communities in southern Spainvineyards. Agriculture, Ecosystems & Environment, 272, 135e145. https://doi.org/10.1016/j.agee.2018.11.010

Hall, R. M., Penke, N., Kriechbaum, M., Kratschmer, S., Jung, V., Chollet, S.,Guernion, M., Nicolai, A., Burel, F., Fertil, A., Lora, A., S�anchez-Cuesta, R.,Guzm�an, G., G�omez, J., Popescu, D., Hoble, A., Bunea, C.-I., Zaller, J. G., &Winter, S. (2020). Vegetation management intensity and landscape diversityalter plant species richness, functional traits and community compositionacross European vineyards. Agricultural Systems, 177, 102706. https://doi.org/10.1016/j.agsy.2019.102706

ICONA. (1988). Agresividad de la lluvia en Espa~na. Valores del factor R de la ecuaci�onuniversal de p�erdidas de suelo.[Rainfall erosivity in Spain. R-factor values for theUniversal Soil Loss Equation]. Ministerio de Agricultura, Pesca y Alimentaci�on,Madrid. In Spanish).

Khaleghpanah, N., Shorafa, M., Asadi, H., Gorji, M., & Davari, M. (2016). Modelingsoil loss at plot scale with EUROSEM and RUSLE2 at stony soils of Khamesanwatershed. Iran. Catena, 147, 773e788. https://doi.org/10.1016/j.catena.2016.08.039

Kinnell, P. I. A. (2017). A comparison of the abilities of the USLE-M, RUSLE2 andWEPP to model event erosion from bare fallow areas. The Science of the TotalEnvironment, 596e597, 32e42. https://doi.org/10.1016/j.scitotenv.2017.04.046

Kottek, M., Grieser, J., Beck, C., Rudolf, B., & Rubel, F. (2006). World Map of theK€oppen-Geiger climate classification updated. Meteorologische Zeitschrift, 15,259e263.

Liu, B. Y., Nearing, M. A., & Risse, L. M. (1994). Slope gradient effects on soil loss forsteep slopes. Transactions of the ASAE, 37, 1835e1840.

Marín, V., Taguas, E. V., Redel, M. D., & G�omez, J. A. (2014). Validation of RUSLEmodel in olive orchard land use by the virtual tool SECO. In Proceeding of the ELS2014 conference (Bari, Italy).

Marques, M. J., Bienes, R., Cuadrado, J., Ruiz-Colmenero, M., Barbero-Sierra, C., &Velasco, A. (2015). Analysing perceptions attitudes and responses of wine-growers about sustainable land management in Central Spain. Land Degradation& Development, 26, 458e467. https://doi.org/10.1002/ldr.2355

Martı

nez-Casasnovas, J. a, & S�anchez-Bosch, I. (2000). Impact assessment ofchanges in land use/conservation practices on soil erosion in thePened�eseAnoia vineyard region (NE Spain). Soil and Tillage Research, 57,101e106. https://doi.org/10.1016/S0167-1987(00)00142-2

Moriasi, D. N., Arnold, J. G., Liew, Van, M, W., Bingner, R. L., Harmel, R. D., &Veith, T. L. (2007). Model evaluation guidelines for systematic quantification ofaccuracy in watershed simulations. Transactions of the American Society ofAgricultural and Biological Engineers, 50, 885e900. https://doi.org/10.13031/2013.23153

Napoli, M., Cecchi, S., Orlandini, S., Mugnai, G., & Zanchi, C. A. (2016). Simulation offield-measured soil loss in Mediterranean hilly areas (Chianti, Italy) with RUSLE.Catena, 145, 246e256. https://doi.org/10.1016/j.catena.2016.06.018

Nash, J. E., & Sutcliffe, J. V. (1970). River flow forecasting through conceptual modelspart I d a discussion of principles. J. Hydrol., 10, 282e290.

Novara, A., Gristina, L., Saladino, S. S., Santoro, A., & Cerd�a, A. (2011). Soil erosionassessment on tillage and alternative soil managements in a Sicilian vineyard.Soil and Tillage Research, 117, 140e147. https://doi.org/10.1016/j.still.2011.09.007

€Osterreich Wein. (2019). Dokumentation €Osterreich Wein 2018. €Osterreich Wein.Panagos, P., Ballabio, C., Borrelli, P., Meusburger, K., Klik, A., Rousseva, S., Tadi�c, M. P.,

Michaelides, S., Hrabalíkov�a, M., Olsen, P., Aalto, J., Lakatos, M., Rymszewicz, A.,Dumitrescu, A., Beguería, S., & Alewell, C. (2015a). Rainfall erosivity in Europe.The Science of the Total Environment, 511, 801e814. https://doi.org/10.1016/j.scitotenv.2015.01.008

Panagos, P., Borrelli, P., Meusburger, K., Alewell, C., Lugato, E., & Montanarella, L.(2015b). Estimating the soil erosion cover-management factor at the Europeanscale. Land Use Policy, 48, 38e50. https://doi.org/10.1016/j.landusepol.2015.05.021

Panagos, P., Borrelli, P., Poesen, J., Ballabio, C., Lugato, E., Meusburger, K.,Montanarella, L., & Alewell, C. (2015c). The new assessment of soil loss by watererosion in Europe. Environ. Sci. Policy, 54, 438e447. https://doi.org/10.1016/j.envsci.2015.08.012

Pappalardo, S. E., Gislimberti, L., Ferrarese, F., De Marchi, M., & Mozzi, P. (2019).Estimation of potential soil erosion in the Prosecco DOCG area (NE Italy), to-ward a soil footprint of bottled sparkling wine production in different land-management scenarios. PloS One, 14, 1e20. https://doi.org/10.1371/journal.pone.0210922

Pijl, A., Barneveld, P., Mauri, L., Borsato, E., Grigolato, S., & Tarolli, P. (2019). Impact ofmechanisation on soil loss in terraced vineyard landscapes. Geogr. Res. Lett., 45,287e308. https://doi.org/10.18172/cig.3774

Prosdocimi, M., Cerd�a, A., & Tarolli, P. (2016). Soil water erosion on Mediterraneanvineyards: a review. Catena, 1e21. https://doi.org/10.1016/j.catena.2016.02.010

Raclot, D., Le Bissonais, Y., Louchart, Y., Andrieux, P., Moussa, R., & Voltz, M. (2009).Soil tillage and scale effects on erosion from fields to catchment in a Mediter-ranean vineyard area. Agriculture, Ecosystems & Environment Catena, 66,198e210.

Regione Piemonte. (2016). Rapporto di valutazione ex post. Programma di SviluppoRurale della Regione Piemonte 2007-2013. Available on line https://www.regione.piemonte.it/web/sites/default/files/media/documenti/2019-03/


valutazione_ex-post_integrale.pdf.Renard, K. G., Foster, G. R., Weesies, G. A., McCool, D. K., & Yoder, D. C. (1997).

Predicting Soil Erosion by Water: A Guide to Conservation Planning with theRevised Universal Soil Loss Equation (RUSLE). USDA Washington DC: USDepartment of Agriculture Agricultural Handbook No. 703.

Risse, L. M., Nearing, M. A., Laflen, J. M., & Nicks, A. D. (1993). Error assessment inthe universal soil loss equation. Soil Science Society of America Journal, 57, 825.https://doi.org/10.2136/sssaj1993.03615995005700030032x

Rodrigo Comino, J. (2018). Five decades of soil erosion research in “terroir”. TheState-of-the-Art. Earth-Science Reviews, 179, 436e447. https://doi.org/10.1016/j.earscirev.2018.02.014

Rodrigo Comino, J., Quiquerez, A., Follain, S., Raclot, D., Le Bissonnais, Y., Casalí, J.,Gim�enez, R., Cerd�a, A., Keesstra, S. D., Brevik, E. C., Pereira, P., Senciales, J. M.,Seeger, M., Ruiz Sinoga, J. D., & Ries, J. B. (2016). Soil erosion in sloping vineyardsassessed by using botanical indicators and sediment collectors in the Ruwer-Mosel valley. Agriculture, Ecosystems & Environment, 233, 158e170. https://doi.org/10.1016/j.agee.2016.09.009

Ruiz-Colmenero, M., Bienes, R., Eldridge, D. J., & Marques, M. J. (2013). Vegetationcover reduces erosion and enhances soil organic carbon in a vineyard in thecentral Spain. Catena, 104, 153e160. https://doi.org/10.1016/j.catena.2012.11.007

Ruiz-Colmenero, M., Bienes, R., & Marques, M. J. (2011). Soil and water conservationdilemmas associated with the use of green cover in steep vineyards. Soil andTillage Research, 117, 211e223. https://doi.org/10.1016/j.still.2011.10.004

Salom�e, C., Coll, P., Lardo, E., Metay, A., Villenave, C., Marsden, C., Blanchart, E.,Hinsinger, P., & Le Cadre, E. (2016). The soil quality concept as a framework toassess management practices in vulnerable agroecosystems: A case study in

mediterranean vineyards. Ecological Indicators, 61, 456e465. https://doi.org/10.1016/j.ecolind.2015.09.047

Soil Survey Staff. (2010). Keys to soil taxonomy (11th ed.). Washington, DC: USDA-Natural Resources Conservation Service.

Spaeth, K. E., Pierson, F. B., Weltz, M. A., & Blackburn, W. H. (2003). Evaluation ofUSLE and RUSLE estimated soil loss on rangeland. Journal of Range Management,56, 234e246. https://doi.org/10.2307/4003812

Tiwari, a K., Risse, L. M., & Nearing, M. a (2000). E valuation of wepp and I Ts Comparison (Vol. 43, pp. 1129e1135).

Tropeano, D. (1983). Soil erosion on vineyards in the tertiary Piedmontese basin(northwestern Italy): Studies on experimental areas. Catena Supplement, 4,115e127.

USDA-ARS. (2013). Revised Universal Soil Loss Equation Version 2 (RUSLE2), ScienceDocumentation. USDA-ARS. https://www.ars.usda.gov/ARSUserFiles/60600505/RUSLE/RUSLE2_Science_Doc.pdf.

Verheijen, F. G. A., Jones, R. J. A., Rickson, R. J., & Smith, C. J. (2009). Tolerable versusactual soil erosion rates in. Europe Earth Sci. Rev., 94, 23e38, 10.1016/j.earscirev.2009.02.003.

Winter, S., Bauer, T., Strauss, P., Kratschmer, S., Paredes, D., Popescu, D., Landa, B.,Guzm�an, G., G�omez, J. A., Guernion, M., Zaller, J. G., & Bat�ary, P. (2018). Effects ofvegetation management intensity on biodiversity and ecosystem services invineyards: A meta-analysis. Journal of Applied Ecology, 1e12. https://doi.org/10.1111/1365-2664.13124

Zhang, X., Nearing, M., Risse, L., & McGregor, K. (1996). Evaluation of WEPP runoffand soil loss predictions using natural runoff plot data. Transactions of the ASAE,39, 855.



Assessing spatial variability and erosion susceptibility of soils in hillyagricultural areas in Southern Italy

Carmen Maria Rosskopf a, Erika Di Iorio b, *, Luana Circelli b, Claudio Colombo b,Pietro P.C. Aucelli c

a University of Molise, Department of Biosciences and Territory, c.da Fonte Lappone, 86090, Pesche, IS, Italyb University of Molise, Department of Agricultural, Environmental and Food Sciences, Via F. De Sanctis, 1, 86100, Campobasso, CB, Italyc University of Napoli “Parthenope”, Department of Environmental Sciences, Centro Direzionale Isola C/4, 80143, Napoli, Italy


Article history:Received 1 March 2020Received in revised form21 September 2020Accepted 24 September 2020Available online 1 October 2020

Keywords:Hillslope morphodynamicsRUSLE equationK factorMediterranean regionSoil catenaLand units

a b s t r a c t

Soil erosion is one of the main environmental problems in the Mediterranean area. This problem isbecoming even more important especially in Italy, in the Apennines, where severe erosive processesoccur due to the action of concentrated running water. The erodibility (K-Factor) of a soil, estimated usingthe Revised Universal Soil Loss Equation (RUSLE), is a measure of its susceptibility to erosion and dependson several soil properties such as organic matter, texture and permeability and structure.

To assess the spatial variability of soil properties and soil erodibility in hilly agricultural areas and toinvestigate the relationships between soil features and landscape morphodynamics, a detailed study inMolise region (southern Italy), in a small drainange basin located along its hilly Adriatic flank, was carriedout. In this catchment, 63 topsoil samples (A horizons) were collected and 10 soil profiles, forming acatena crossing 3 land units, were sampled. The calculated K-Factors ranges between 0.012 and 0.048 t hah ha�1 MJ�1 mm�1 indicating a complex spatial distribution, due to the several local pedological andgeomorphological factors affecting soil erodibility. The results give clear evidence about the relationshipsamong soil characteristics, soil erodibility and landscape morpho-dynamics (land units).

Comparing the soil loss rates estimated for the study area with those reported in literature, a goodcorrespondence can be observed only for the more stable land unit, not characterized by intense erosiveprocesses. The proposed methodology is suitable to highlight areas characterized by similar morpho-dynamics features, and comparable soil erodibility, for a more effective spatialization of K factor.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction

Soil erosion by water is one of the most significant environ-mental problems worldwide (e.g. Borrelli et al., 2017) andthroughout Europe and the Mediterranean region (Boardman &Poesen, 2006; García-Ruiz et al., 2013; Panagos et al., 2015; Torriet al., 2006 and references therein).

In Europe, the highest average annual soil loss rates (at countrylevel) are found in Italy (8.46 t ha�1 yr�1), Slovenia(7.43 t ha�1 yr�1) and Austria (7.19 t ha�1 yr�1), due to a combi-nation of high rainfall erosivity and steep topography (Panagos

et al., 2015).In Italy, highest soil erosion rates are found along the Apennines

and the surrounding hilly areas (Borrelli et al., 2018). Soil erosion islargely concentrated in arable crop areas where it is frequentlytriggered by the adopted agricultural and management practises(Costantini & Lorenzetti, 2013; Panagos et al., 2015). Especially theincreasing exploitation of agricultural areas and the use of intensetillage and farming are largely favouring soil loss due to watererosion with a sensible reduction of the topsoil (Borrelli et al.,2014). Furthermore, in many areas, the worsening of soil physicalproperties due to soil compaction and related soil porosity reduc-tion, caused by the passage of machinery, can be observed (Faticaet al., 2019).

In the hilly-mountainous sectors of Molise region (SouthernItaly), due to their orography and complex lithological-structuralsetting (Aucelli et al., 2001; 2010), water erosion (Aucelli et al.,

* Corresponding author. University of Molise, Department of Agricultural, Envi-ronmental and Food Sciences, Via F. De Sanctis, 1, 86100, Campobasso, CB, Italy.

E-mail address: [email protected] (E. Di Iorio).






2006) and mass wasting (e.g. Aucelli et al., 2013; Borgomeo et al.,2014; Rosskopf & Aucelli, 2007) play a fundamental role in landdegradation and related loss of agricultural areas (Zumpano et al.,2018). Especially on steep slopes, the prevalence of impermeableand highly erodible clayey terrains, common land use practises(Pisano et al., 2017; Zumpano et al., 2018), and the abandonment ofconservation measures such as terraces and waterways, increasesignificantly the phenomena of water erosion and flooding, whichare easily triggered by intense and/or prolonged rainfalls (Aucelliet al., 2004).

To predict soil losses and plan soil conservation, the most usedempirical models are the USLE equation (Wischmeier & Smith,1978) and its revised form, the RUSLE version, which allow theestimation of average annual soil loss based on the product of fiveerosion risk factors. Two of these five factors are the result ofspecific local geologic-geomorphological and climate characteris-tics, namely the rainfall erosivity factor (R-factor) (Borrelli et al.,2014) and the soil erodibility factor (K-factor) (Panagos et al.,2014) that reflects the geopedological conditions and relatedspatial variations.

The concept of soil erodibility is generally accepted to be relatedto the susceptibility of the soil to erosion and is influenced by alarge number of soil properties related to the soil pedon that in-fluence its depth and the vegetative growth (Parysow et al., 2003).The main soil properties that control soil erodibility and, therefore,potential soil losses, include particle size distribution, shape, size ofaggregates, bulk density, porosity and permeability, organic matterand clay content, and chemical composition (Faulkner et al., 2004;Pulice et al., 2009).

For areas lacking field experimental data, the K values can bepredicted by using the soil erodibility (K-factor) nomograph andthe related algebraic approximation proposed by Wischmeier andSmith (1978) and reported in Panagos et al. (2014) that includesfive soil parameters, i.e. texture, organic matter, coarse fragments,structure and permeability.

As demonstrated by Favis-Mortlock (1998), the results of RUSLEmodels are often poor in areas other than those in which they havebeen tested. Such models may also require input data that are notalways available for large areas with the necessary accuracy (e.g.,Benavidez et al., 2018; Jones et al., 2003).

Regarding particularly soil erodibility, according to Panagoset al. (2014) direct measurements of the K-factor are not finan-cially sustainable at the regional or national levels. To calculate theK-factor at the European level, these authors used the LUCAS topsoildatabase, which is based on a density of topsoil sampling of about 1per 199 km2, corresponding to a grid cell size of around14 km � 14 km. The overall range of K values obtained by Panagoset al. (2014) at the European level is 0.004e0.076 t ha hha�1 MJ�1 mm�1. The mean K-factor for the 25 Member States ofEurope is 0.032 t ha h ha�1 MJ�1 mm�1 with a standard deviation of0.009 t ha h ha�1 MJ�1 mm�1. Their study provided a 500 m res-olution K-factor map of Europe that was then employed in thecalculation of mean annual soil loss rates at the European scale bymeans of a modified version of the RUSLE model (Panagos et al.,2015).

For Italy, the mean K-factor calculated by Panagos et al. (2014) is0.0322 t ha h ha�1 MJ�1 mm�1 with a standard deviation of0.0077 t ha h ha�1 MJ�1 mm�1. Considering the protective effect ofsurface stone cover, this value decreases to 0.0276 t ha hha�1 MJ�1 mm�1, corresponding to a reduction of 14.5% due tostoniness.

The constructed soil erodibility dataset (Joint Research Centre ofthe European Commission, 2014) surely represents a high qualityresource for soil erosion estimations at regional to national levelsand allows overcoming the limited data availability for K-factor

assessment especially for large-scale studies. Nevertheless giventhe low density of topsoil sampling points, it may not correctly orexhaustively represent the soil features and related soil erodibilityat smaller spatial scales.

Therefore, to avoid the risk of using too rough K-factor values, inthe present study, a detail scale analysis was carried out in a smalltest area located in Southern Italy with the aim to assess the spatialvariability of soil properties and soil erodibility in hilly agriculturalareas and to investigate the relationships between soil features andlandscape morphodynamics.

The test area (about 2.67 km2) is located on the hilly Adriaticflank of the Molise region, and falls in the headwater portion of theRivo River catchment, a small-sized catchment of approximately80 km2 that well represents the geological and geomorphologicalconditions of this sector of the Apennine chain.

To characterize the local geopedological context, i.e. the rela-tionship between soil properties and landscape features, 10 soilprofiles forming a soil Catena (Birkeland, 1999) were analysed. Toassess the spatial variability of the K-factor, we analysed the top-soils sampled along the nodes of a 300 m square grid covering theentire test area (on total 63 samples) and those coming from the 10soil profiles of the catena. The resulting spatial resolution ofcalculated K-factors (measured by the RUSLE equation) is 13 pointsfor km2.


2.1. Site description

The Rivo catchment is located on the Adriatic flank of the MoliseApennine. This sector of the Apennine chain is characterized by adominant hilly morphology and elevations ranging between 150 m(minimum height of main valley floors) and 1000 m a.s.l.(maximum height of some isolated peaks) (Aucelli et al., 2001,2010). The landforms are mainly developed on clayey-marly lime-stone successions and siliciclastic deposits belonging to the UpperCretaceous-Miocene Sannio Units and the Upper Oligocene-Miocene Molise units (Vezzani et al., 2004; ISPRA, 2015). Theprevalence of mostly pelitic rocks with scarce to nil permeabilityhas allowed the development of a dense fluvial network. Majorwater courses have incised deep V-shaped valleys suspendingancient mature landforms, that are represented by polycyclic,gently dipping erosion surfaces in summit position and hangingvalley side glacis (Amato et al., 2017; Aucelli et al., 2011). Valleyflanks are affected by intense and widespread erosion phenomena.Especially steeper portions are evolving under the action of runningsurface water and mass wasting processes (Aucelli et al., 2010). Thelatter affect widespread especially clayey steep slopes (Borgomeoet al., 2014; Pisano et al., 2017; Rosskopf & Aucelli, 2007) andrepresent one of the main issues for landscape management(Zumpano et al., 2018).

The Rivo catchment has altitudes ranging from 225 to 887 ma.s.l. (Fig. 1) and is underlain by prevailingly argillitic, marly andarenaceous rocks. It is characterized by a temperate climate, withmean annual rainfall and temperatures ranging respectively be-tween 650 and 800mmand 5 and 30 �C. The regime of temperatureand precipitation leads to the identification of an arid summerperiod that lasts about 4 months (Aucelli et al., 2006). Most of theannual precipitation occurs during the late autumn and springmonths. Prolonged and extreme rainfall events occur periodicallyand are responsible for huge erosion related both to landslides andwater erosion (Aucelli et al., 2004).

From a morpho-dynamic point of view, the Rivo catchment and,with it, the test area, is characterized by five major land units. Thelatter are distinguished from each other not only by different slope

C.M. Rosskopf, E. Di Iorio, L. Circelli et al. International Soil and Water Conservation Research 8 (2020) 354e362

355

gradients, morphology and elevation, but most of all by differentdominant erosive processes responsible for their current geomor-phological evolution.

- Unit A: flat or slightly rounded summit erosion surfaces, subject tosplash and sheet erosion. This unit is made of flat or slightlyrounded remnants of morphologically mature surfaces, withslope gradients ranging generally from 3% to 10%, which arelocated at altitudes between 750 and 800 m above sea level.These surfaces are generally of erosional origin, i.e. are relicts ofancient valley floor morphologies that preceded the progressiveincision and valley downcutting by the current hydrographicnetwork. In some cases, however, these surfaces are the result ofmorphoselective erosion and correspond to the presence of rockbanks in a sub-horizontal position that are more resistant toerosion. The lithological substratum is represented byarenaceous-marly and clayey rocks. Only mild erosion occurs onthese surfaces, especially splash and sheet erosion. Due to thecharacteristic low slope gradients, the prevailing land uses arerainfed crops and marginally vineyards.

- Unit B: low to medium gradient connecting slopes subject to rillerosion. The connection between the summit land unit A and theslope unit C or the valley floor unit A1 is represented by straightor concave-convex slopes characterized by slopes generallybetween 7% and 25%. These slopes are located at altitudes be-tween approximately 600 and 800 m above sea level. Amongerosive processes, rill erosion prevails, to which solifluctionphenomena and sporadic episodes of incipient gully erosion areadded. The prevailing land use are rainfed crops, vineyards andonly in a marginal area a forest is present.

- Unit C:moderately steep to very steep valley slopes subject towatererosion and gravitational processes. This unit, according to pre-vailing erosion processes, has been divided into the followingthree sub-units. Despite unsuitable morphology, this unit ispartly occupied by crops, and only in the steepest areas grass-land and wood in the early stages of development are found.

- Sub unit C1: high gradient connecting slopes mainly affected bylandslides. These generally concave-convex slopes are prevail-ingly located at 500 and 600 m of height. They have high slopegradients, generally exceeding 25%, and are widespread affected

by landslides. The dominant substratum is clayey, and soils aregenerally strongly reworked.

- Sub unit C2: High gradient connecting slopes affected prevailinglyby badlands. This landscape unit is mainly located near to themainwaterways and has a reduced areal extension. It is affectedby badland erosion, i.e. by particularly severe concentratedsurface and sub-surface water erosion, and devoid of bothvegetation cover and fertile soil cover.

- Sub unit C3: High gradient connecting slopes prevailingly affectedby gully erosion. This landscape unit is very similar to sub-unitC1, but affected prevailingly by gully erosion. This erosion phe-nomenon occurs especially in areas that were affected bylandslides in the past, that had stabilized or had becomeintermittent.

- Unit Al: Aggraded valley floor. This landscape unit corresponds tothe flat, low gradient valley floor reaches generated by sedi-mentation and riverbed aggradation.

- Unit E: bare rock outcrops. A characteristic element of the Rivocatchment and the study area is represented by some outcropsof bare rock, represented by huge calcareous rock spurs broughtout by selective erosion operated by surface water (Filocamoet al., 2019).

2.2. Pedological sampling and laboratory analyses

Two campaigns of soil sampling and characterization werecarried out in the test area.

Ten soil profiles, forming a soil catena, were sampled along a1.7 km long transect that crosses from waterdivide to waterdividetransversally the headwater area of the Rivo catchment, inter-cepting land units A, B and C (Fig. 2). For each profile, fielddescription of soils wasmade in accordancewith the FAO (2006). Toclassify the soils according to the criteria proposed by IUSSWorkingGroup WRB, 2015, main diagnostic horizons were subjected tophysical and chemical laboratory analyses.

Furthermore, a systematic sampling of topsoils in 63 georefer-enced points corresponding to the nodes of a 300 m square grid(Fig. 3a) was carried out. The uppermost, 20-cm thick soil portionswere picked up. The sampled topsoils were very poor in skeleton

Fig. 1. Location and altimetric map of the Rivo catchment. The location of the test area is indicated with a frame.


356

(rock fragments > 2mm); only those sampled on steep slopes had avery limited skeleton. Furthermore, being the study area mostlyunderlain by argillites, sampled topsoils were devoid of surfacestone cover.

All soil samples were subjected to chemical and physical labo-ratory analyses using the air dried and sieved (<2 mm) parts of thesoil samples, following the Italian Methods Manual on Soil

Chemical Analyses (Colombo & Miano, 2015): the soil pH waspotentiometrically measured on suspensions soil-H2O (1:2.5 ratio);the soil organic matter (SOM) content was determined by means ofthe dichromate oxidation procedure and based on results; thecation exchange capacity (CEC) was determined by BaCl2; textureand the related particle size distribution were determined aftersieving by the pipette method, the very fine sand (0.05e0.1 mm)

Fig. 2. Photos illustrating the 10 soil profiles (psn1-psn10) forming the catena (psn is the acronym of “Salcito profile number”).


357

was determined as 20% of the sand fraction; furthermore, the bulkdensity and calcium carbonate content were analyzed, as theseparameters can be considered good indicators of acceleratedtopsoil erosion.

The most common statistical parameters (Table 2) were deter-mined by using the software Statistica 10.

2.3. Soil erodibility estimation

The soil erodibility or K-Factor (t ha h ha�1 MJ�1 mm�1) wasestimated on the basis of soil properties of the 63 sampled topsoils,such as particle size composition, content of organic matter, soilstructure and permeability. This factor represents the soil loss rateper erosion index unit for a specified soil as measured on a standardplot. The values of the K-factor of the topsoil samples were

Fig. 3. Location of the 63 topsoil samples, according to a square grid of 300 m (a). Location of the 10 soil profiles (psn1-psn10) forming the catena (b) and related cross-profileshowing their relationship with the distinguished land units (c). U.A. ¼ Unit A: flat or slightly rounded summit erosion surfaces; U.B. ¼ Unit B: low to medium gradient con-necting slopes subject to rill erosion; U.C. ¼ Unit C: moderately steep to very steep valley slopes subject to water erosion and gravitational processes. U.Al ¼ Unit Al: aggraded valleyfloor. A detailed description of landscape units is given in the text.

Table 1FAO classification and main characteristics of the catena soil profiles.

Profile code FAO Classification Depth (cm) Elevation (m a.s.l.) Slope gradient (�) Lithology a Cracks (mm) SOM (%) Clay (%)

psn4 Haplic Vertisols 210 739 0.4 M >10 2.5 65psn1 Haplic Vertisols 160 723 4.6 ARG <10 1.8 61psn2 Haplic Vertisols 150 690 4.9 ARG >10 2.1 60psn3 Haplic Hypercalcic Vertisols 190 650 15.4 ARG <10 2.1 58psn8 Luvic Calcisols 45 720 16.9 ARG 3e5 5.7 41psn7 Cambisols 260 714 6.5 ARG No 2.3 21psn6 Calcaric Cambisols 150 725 16.3 A <10 3.1 45psn10 Haplic Calcisols 130 700 11.5 ARG No 1.5 38psn9 Luvic Calcisols 150 700 11.5 ARG No 1.0 27psn5 Brunic Skeletic Leptosols 45 770 3.9 A No 2.2 13

a Parent materials are principally sedimentary rocks: Argillite (ARG), Arenite (A), Marls (M).


358

calculated using the following formula of Wischmeier and Smith(1978):

K ¼ 2:1� 10�4�12� OM

�M1:14 þ 3:25

�S� 2

�þ 2:5

�P � 3

�

�.7:59� 100

(1)

where K is expressed in t ha h ha�1 MJ�1 mm�1, OM represents theorganic matter content (%), M defines the relations between per-centages of silt, very fine sand and clay content (% silt þ % very finesand)/(100 - % clay), S represents the soil structure code (S¼ 1: veryfine granular, S¼ 2: fine granular, S¼ 3:medium or coarse granular,S ¼ 4: blocky, platy or massive), and P the permeability class.

To estimate the permeability, the field-saturated hydraulicconductivity was measured in the field using a double-ring infil-trometer (six measurements per land use, each one with five rep-etitions). The soil structure class was defined in accordance withRenard et al. (1997). Since the sampled soils were found to haveno surface stone cover, no adjustment of the K-factor (Panagoset al., 2014) was necessary.

2.4. Model and parameter estimation of average annual soil loss

Average annual soil loss was estimated by using the modifiedversion of the RUSLE model and the following equation:

A ¼ R � K� LS� C� P (2)

Where A is the estimate average annual soil loss (t ha�1 yr�1)caused by sheet and rill erosion.

R is the rainfall erosivity factor (MJ mm ha�1 h�1 yr�1), K is thesoil erodibility factor (t ha h ha�1 MJ�1 mm�1), LS is the slopelength-gradient factor (dimensionless), C is the cover-managementfactor (dimensionless), and P is the support practice factor(dimensionless).

To calculate the R-factor, the time-series used in this studyconsisted in a continuous 80-year precipitation data record(1920e2000) coming from the Trivento station. The investigatedtime period sufficiently represents the inter-annual and decennialvariability of rainfall.

The R-factor was derived as a mean of 6 equations (Arnoldus,1977 and, 1980; Lo et al., 1985; Renard & Freimund, 1994; Yu &Rosewell, 1996) and correspond to a value of 1579 MJ mmha�1 h�1 yr�1.

For the calculation of the K-factor, equation (1) was used asalready described in the precedent section 2.3.

The LS-factor was directly calculated on the basis of a high-resolution (5 m resolution) Digital Elevation Model of the studyarea, implemented in the ESRI ArcGIS environment according to thefollowing equation:

LS ¼ 1:4� ðAr\22:13Þ � 0:4� ðsinb\0:0896Þ � 1:3 (3)

where Ar¼ catchment area calculated for each grid cell of DEM andb ¼ slope (Mitasova & Brown, 2002; Mitasova et al., 1996).

Spatial data for the C-factor, which defines the type of land coverpresent in the study area, were extracted from the IV Level of theCORINE Land Cover dataset (Corine Land Cover, 2012).

Regarding the C-factors for the land cover types present in thestudy area, we adopted, according to Borrelli et al. (2014), thefollowing values: 0.22 for conventional crops (rainfed croplands),0.66 for vineyards, 0.05 for grassland and 0.007 for forest-basedland management.

The P-factor accounts for conservation practices that reduce theerosion potential of the surface runoff thanks to their influence ondrainage pattern, runoff concentration, runoff velocity, and hy-draulic forces exerted by surface runoff on the soil. The value of theP-factor was set equal to 1 because in the study area, water erosioncontrol procedures that respond to the soil conservation practiceare lacking, while soil tillage is frequent.

3. Results

3.1. The soil catena

The performed surveys and related analyses evidence differenttypes of soils within the study area. The differences found are to belinked not only to the different parent materials from which thesesoils have originated, but also to the geomorphological context towhich they belong. The catena allows the investigation of the re-lationships between soils and slope features, especially hydrology,acclivity, morphology and morphodynamics (erosion e deposition)(Fig. 3). The soil profiles change along this sequence together withdrainage and geomorphological conditions, and are the result of theinteraction between parent material, climate and their position onthe slope. According to the FAO classification system (IUSSWorkingGroupWRB, 2015), the soils forming the catena were classified intofour categories (Table 1): Vertisols (clay soils, well structured,deeply cracked and characterized by a modest calcium horizon),Calcisols (characterized by a superficial calcium Horizon), Cambi-sols (young soils in an initial stage of differentiation, characterizedby a Bw horizon) and Leptosols (very thin soils with a massivestructure and sandy texture in direct contact with the parentmaterial).

3.2. The Vertisols along the soil catena

In soils characterized by marked vertic properties, deep cracksextending up to a considerable depth (>1 m) are found widespreadwithin the soil profile. Cracks can play a double role. During rainfallevents of limited duration, they represent an effective system forthe deep drainage of infiltration waters, helping to limit surface

Table 2Summary statistics of the content of soil organic matter (SOM), texture (percentages of coarse sand, fine sand, silt and clay), calcium carbonate content and K-factor of the 63topsoils.

SOM Texture CaCO3 K factor

(g/Kg) Coarse Sand (%) Fine Sand (%) Silt (%) Clay (%) (g/Kg) t ha h ha�1 MJ�1 mm�1

Mean 26.35 5.23 36.8 35.13 22.09 47.51 0.027Median 24.96 2.48 33.43 38.05 23.28 15.79 0.025Variance 106.26 43.21 126.3 87.5 51.13 2967.5 4,55*10-5

Dev_st 10.31 6.57 11.24 9.35 7.15 54.47 0.0067Skewness 0.79 2.24 0.37 �0.96 �0.13 1.0 0.85kurtosis 0.49 4.52 �1.17 0.68 �1.04 �0.12 1.15


359

runoff; however, when intense and/or prolonged rainfall eventsoccur, cracks can be completely saturated, favoring in this way theformation of subsurface erosion phenomena due to piping.

The performed analyses evidence that the deepest soils occurboth on the flat gently sloping top surfaces of land unit A (psn 4)and on the uppermost portion of land unit B (psn 9). In particular,the profile of soil psn 4 (Fig. 3, Table 1) is about 1.7 m deep, wellpedogenized, very clayey (65%) and well structured (large sub-angle polyhedral structure). It is furthermore characterized byabundant and very deep cracks, good decarbonation, a deep cal-cium horizon with abundant calcium carbonate concretions, and aBss diagnostic horizonwith typical highly developed “pressure andsliding faces” (slickensides). According to the FAO classification(IUSS Working Group WRB, 2015), this soil is a Haplic Vertisols.Slightly downslope, soil profiles psn 1 and psn 2 are found (Fig. 3),which show very pronounced vertic properties, but are shallowerand at a less advanced stage of pedogenesis and alteration of parentmaterial when compared with psn 4. Moving further downslope,the psn 3 profile is found that falls in the landscape unit C1. Thisprofile stands on a steep plowed slope, showing signs of a ratherrecent surface landslide, and lacks the Ap horizon. Also in this case,the texture is mainly clayey but SOM and the CEC are lower. Aburied horizon B was found within this profile that confirms thetendency of the slope to be affected by very shallow landslides.Therefore, the psn 3 is recognized as a rejuvenated soil, in whichthe upper part has been removed, and on which erosive processesstill act intensely. In this perspective, the abundance of calciumcarbonate in the upper horizon is to be understood as a good in-dicator of soil rejuvenation: in fact, the levels of calcium carbonateare compatible with those found in the B horizon of profile psn 4,indicating the removal of part of the superficial horizon.

3.2.1. The Leptosols, Cambisols and CalcisolsThe psn 5 profile corresponds to very thin soil without well

developed horizons, although located in the uppermost part oflandscape unit A. This soil is still in an initial phase of evolution, dueto lowweathering of the underlying rock and probably because it issubject to particularly intense erosion phenomena, and has beenclassified as a Haplic Leptosols (Calcaric, Eutric). The psn 5 profile ismassive and very sandy, has no B horizon and is characterized bymoderate CEC and SOM content, as well as a rather high pH due tothe high carbonate content. These findings highlight that the soilmanifests serious symptoms of surface erosion, which can onlypartially be justified by the action of surface water. Its high sandcontent and low percentage of clay are due to the parent material,since this slope is set on weakly cemented sandstones (thearenaceous-pelitic complex), and certainly influenced the soil ahigh erodibility. Themost striking feature of this soil is certainly thelack of structure and surface cracks. We can conclude that there areimportant textural differences between this and the other profiles.The stage of pedological development may have been caused byhuman activity rather than by natural phenomena. In particular,tillage erosion most likely has influenced its development andcontributed to speeding up the erosion processes.

Cambisols (psn 6 and psn 7) and Calcisols (psn 9 e psn 10) havelower clay contents but partially show a good structural stability. Ingeneral, the Cambisols are present in areas affected by high rates ofwater erosion, and in regionsmainly developed on parent materialsthat are resistant to clay movement through the profile. They showa low differentiation of sub-superficial horizons due both to slowweathering of parent material and continuous removal of the sur-face horizons with the water runoff.

3.3. The topsoils

The mean results in Table 2 show significant differences be-tween land uses and soil classification. With reference to the par-ticle size distribution (Table 3), the top-soils (A horizons) aremostlysandy loam, formed mainly of fine sand, followed by silt and lowquantities of clay.

Tables 2 and 3 show the main soil properties used to evaluatethe K-factor. The SOM content of the soil samples reaches a meanvalue of 25 g/kg, indicating that most topsoil samples have overallmoderate to high SOM contents. As for texture, the mean referencevalues are represented by the following percentages: fine sand36.8%, silt 35.13% and clay 22.09%. Therefore, from a textural pointof view, the analysed soils were defined as sandy loam soils.

The study discloses an overall mean K-factor of 0.0027 (±0.0067SD) t ha h ha�1 MJ�1 mm�1 for the study area. Analyzing the K-factors in relation to different land units, moderate differences canbe observed. Anyway, land unit C is characterized by the highestmean value, namely 0.029 (±0.006 SD), and therefore appears moreprone to erosion.

4. Discussion and conclusions

From the pedological point of view, we can draw some conclu-sions by observing in detail some chemical and physical charac-teristics of the soils moving along the catena, starting from psn 4,i.e. the profile that shows the best state of development among the10 investigated profiles. Psn 4was developed on a planar slopewitha very weak inclination, largely used for agricultural purposes, andaffected bymild phenomena of surfacewater erosion such as splashand sheet erosion, subordinately rill erosion. The profile, mainlywith vertic properties, appears fully developed. Proceeding alongthe morphodynamic gradient, it is possible to note that the erosiveprocesses and the position of the profile itself become predominantin the evolution and state of conservation of the profile. In the mostsloping portions of the slope, we note that the profile is lessdeveloped, less deep, and characterized by possible beheadings dueto surface erosion. Furthermore, it is possible to note how the land-use that overlaps and sometimes totally alters the natural processof pedogenesis, can cause the acceleration of erosive phenomenaup to the total loss of the most fertile portion of the soil, with aconsequent serious environmental damage. It is therefore clear thatthere are significant relationships between the land units, thedenudation processes that act on them and the soils that candevelop on them; it is also clear that through a pedological survey itis possible to “read” some signs that allows us to understand whichis the most important erosion process and what damage it hascaused and is causing.

This study shows that it is possible to use a modeling approach -the RUSLE based method - to develop a detailed spatial soil infor-mation to assess the distribution of erosion risk within a smallcatchment.

For Italy, the mean K-factor calculated by Panagos et al. (2014) is0.0322 t ha h ha�1 MJ�1 mm�1with a standard deviation of0.0077 t ha h ha�1 MJ�1 mm�1. In the study area, the mean K-factoris slightly lower than the national one and corresponds to 0.027(±0.0067 SD) t ha h ha�1 MJ�1 mm�1 (Table 2). Considering thedifferent land units, the K-Factors derived from the topsoils rangefrom 0.026 to 0.029 t ha h ha�1 MJ�1 mm�1 (Table 3), with thehighest average value obtained for land unit C. In addition, the Kvalues obtained for the catena soil profiles are largely in agreementwith those derived from the topsoils (Table 3). Anyway, in our


360

statistical analysis we have not considered the K values calculatedfor the catena profiles, because the soil sampling was made usingmajor depth intervals with respect to those used for topsoil sam-pling (20 cm).

Examining the soil parameters considered, the texture datahighlight that the content of fine sand is themajor force influencingthe K-factor values. In fact, the fine sand content is 14% higher inland unit C than in land unit A, and about 8% higher than the overallmean value of the study area (data not show). In land unit C, siltcontents remain stable, while clay contents decrease with respectto the overall mean.

In general, higher K values occurr in the central part of the studyarea, with three important hot-spots in correspondence of steepslopes (>30%) of land unit C. Moderate K-Factor values are found onthe gentle slopes, where the main development of soils is charac-terized by vertic properties with slopes generally between 5% and10%. The positive influence of the clayey soil texture on the soilresistance to water erosion is evident at a glance for a group of soilsof land unit A: deep soils with good infiltration capacity and rela-tively good permeability are among the most resistant (psn 4). Insome cases (particularly on steep slopes, sub-unit C1), the Vertisols(psn 3) may have been strongly influenced by severe erosion. Landunit B hosts the soils most frequently occurring in the Apenninearea, mainly Cambisols and Calcisols. The group is very variegated,both because of different sedimentary substrates (arenaceous-marly and clayey rocks) and to variable pedogenesis. Regarding theCalcaric Cambisols (psn 9 and psn 10), major contents of calciumcarbonate in horizon A evidence that they have undergone majorwater erosion and a low differentiation of soil horizons.

Based on the RUSLE approach and calculated K-factors, annualsoil losses have been estimated for the study area according to land

units and prevailing land use categories.The mean erosion rate in land unit A (slightly rounded top

surfaces) was obtained for vineyards with 17.87 (±7.55 SD) tha�1 yr�1, followed by rainfed crops with is 5.96 ± (2.5 SD) tha�1 yr�1. Shrubland (1.35 ± 0.57 t ha�1 yr�1) and natural forests(0.19 ± 0.08 t ha�1 yr�1) have significatively lower values. The landunit B, according to higher K-Factors and slope gradients withrespect to land unit A, shows also significantly higher mean soilerosion rates for all land uses (Table 4). Especially for rainfed crops,a mean value of 19.6 (±17.2 SD) t ha�1 yr�1 was obtained. This valueis twenty fold higher than that estimated by Aucelli et al. (2006)based on direct measurements in a test plot (test plot station ofMorgiapietravalle) during only three years of monitoring(2002e2004).

For land unit C (high slope connecting slopes), a value of41.01 ± (25 SD) t ha�1 yr�1 has been estimated for rainfed crops,suggesting a very high erosion risk.

Comparing the soil loss rates estimated for the study area withthose reported in literature, a good correspondence can beobserved for land unit A (8.38 t ha�1 yr�1 for arable land, Panagoset al., 2015; 5e10 t ha�1 yr�1 estimated for the province of Cam-pobasso, Van der Knijff et al., 2000). Considering land unit B, esti-mated erosion values are considerably higher with respect to thosereported by Panagos et al. (2015). Therefore, in unit B and C, someanti-erosive practises coupled with more sustainable cultures arehighly recommended.

In conclusion, the pedological characterization of the soil catenaallowed to identify four different soil categories (Vertisols, Calci-sols, Cambisols and Leptosols) that reflect the geomorphologicalcontext to which they belong. Furthermore, the results of thisresearch clearly indicate a strict relation between soil characteris-tics, land cover, soil management and soil erodibility (K-Factor)along the catena.

Despite the limited catchment extension, the RUSLE approachand the calculated K-factors, evaluated on topsoils, highlighted acomplex spatial distribution of soil erodibility, linked to severallocal pedological and geomorphological factors. Future study coulduse the obtained pedological informations to improve soil erod-ibility evaluation in areas lacking topsoil data.

This approach can help implementing soil conservation strategy(Minimum or zero tillage, crop rotation, soil seeding etc.) carriedout at National and EU level, since local methodologies may sufferfrom a large heterogeneity, uncertainty, low consistency and, inmany regions, total lack of soil information.

This study can help the decision makers to take sustainable landuse management and soil conservation measures in similar catch-ment in Italy. Hence, substantial conservation practices should beevaluated in similar landscape in terms of soil loss. In fact, high andextreme erosion occur mainly in areas that show significantchanges in land use or in abounded area, for instance: steep slopecultivation and barren land.

References

Amato, V., Aucelli, P. P. C., Bracone, V., Cesarano, M., & Rosskopf, C. M. (2017). Long-term landscape evolution of the Molise sector of the centralsouthern Apen-nines, Italy. Geologica Carpathica, 68(1), 29e42. https://doi.org/10.1515/geoca-2017-0003

Arnoldus, H. M. (1977). Methodology use to determine the maximum potentialaverage annual soil loss due to sheet and rill erosion in Morocco. FAO SoilsBulletin, 34, 39e51.

Arnoldus, H. M. (1980). An approximation of the rainfall factor in the universal soilloss equation. In M. De Boodt, & D. Gabriels (Eds.), Assessment of soil erosion (pp.127e132). Wiley & Sons.

Aucelli, P. P. C., Amato, V., Cesarano, M., Pappone, G., Rosskopf, C. M., RussoErmolli, E., & Scarciglia, F. (2011). New morphostratigraphic and chronologicalconstraints for the Quaternary palaosurfaces of the Molise Apennines (southernItaly). Geologica Carpathica, 62(1), 17e26.

Table 3FAO classification, soil permeability, LS-factor of the catena soils and comparisonbetween K-factors of catena soils and the average K value of topsoils, according todifferent land units.

Land unit FAOClassification

SoilPermeability

K-Factor LS-factor

Unit A cm/h t ha h ha�1

MJ�1 mm�1

psn4 Vertisol 0.21 0.020 0.23psn1 Vertisol 0.26 0.020 0.66psn5 Leptosol 0.80 0.011 0.97Topsoil n. 17 0.026 ± 0.005

Unit Bpsn2 Vertisol 0.55 0.016 3.23psn9 Calcisol 2.35 0.018 2.18psn10 Calcisol 0.12 0.025 3.16psn8 Calcisol 1.31 0.021 0.79psn6 Cambisol 2.02 0.020 7.84psn7 Cambisol 0.54 0.014 3.68Topsoil n. 35 0.027 ± 0.008

Unit Cpsn3 Vertisol 0.55 0.020 3.23Topsoil n. 11 0.029 ± 0.006

Table 4Annual erosion rates (ton ha�1 year�1) estimated for the topsoils falling respectivelyin land units A, B and C according to different land cover scenarios.

Land unit Annual erosion rate (ton ha�1 year�1)

Rainfed cropland Vineyard Forest Shrubland/Forest

Unit A 5.96 ± 2.5 17.87 ± 7.55 0.19 ± 0.08 1.35 ± 0.57Unit B 19.6 ± 17.2 58.8 ± 51.5 0.62 ± 0.54 4.45 ± 3.9Unit C 41.01 ± 25.6 123.2 ± 76.8 1.3 ± 0.81 9.33 ± 5.82


361

Aucelli, P. P. C., Casciello, E., Cesarano, M., Zampelli, S. P., & Rosskopf, C. M. (2013).A deep, stratigraphically and structurally controlled landslide: The case ofmount La Civita (Molise, Italy). Landslides, 10, 645e656.

Aucelli, P. P. C., Cinque, A., & Rosskopf, C. (2001). Geomorphological map of theTrigno basin (Italy): Explanatory notes. Geografia Fisica e Dinamica Quaternaria,24(1), 3e12.

Aucelli, P. P. C., De Angelis, A., Colombo, C., Palumbo, G., Scarciglia, F., &Rosskopf, C. M. (2006). La stazione sperimentale per la misura dell’erosione delsuolo di Morgiapietravalle (Molise, Italia): Primi risultati sperimentali. Pro-ceedings of the final workshop of the PRIN 2002 project, Water erosion in Medi-terranean environment: Direct and in direct assessment in test areas andcatchments (pp. 87e104). Geniva: Brigati.

Aucelli, P. P. C., Izzo, M., Mazzarella, A., Rosskopf, C. M., & Russo, M. (2004). L’eventometeorico estremo di gennaio 2003 nel Molise. Quaderni di Geofisica Applicata,11, 101e119.

Aucelli, P. P. C., Robustelli, G., Rosskopf, C. M., Scarciglia, F., Di Paola, G., & Luc�a, F.(2010). Geomorphological map of the area between Frosolone and Trivento(Molise, Italy). Journal of Maps, 423e434. https://doi.org/10.4113/jom.2010.1039

Benavidez, R., Jackson, B., Maxwell, D., & Norton, K. (2018). A review of the (revised)universal soil loss equation ((R)USLE): With a view to increasing its globalapplicability and improving soil loss estimates. Hydrology and Earth SystemSciences, 22, 6059e6086. https://doi.org/10.5194/hess-22-6059-2018

Birkeland, P. W. (1999). Soils and geomorphology. New York: Oxford University Press.Boardman, J., & Poesen, J. (2006). Soil erosion in Europe: Major processes, causes

and consequences. In J. Boardman, & J. Poesen (Eds.), Soil erosion in Europe.Chichester, UK: John Wiley & Sons, Ltd.

Borgomeo, E., Hebditch, K. V., Whittaker, A. C., & Lonergan, L. (2014). Characterisingthe spatial distribution, frequency and geomorphic controls on landslideoccurrence, Molise, Italy. Geomorphology, 226, 148e161. https://doi.org/10.1016/j.geomorph.2014.08.004

Borrelli, P., M€arker, M., Panagos, P., & Schütt, B. (2014). Modeling soil erosion andriver sediment yield for an intermountain drainage basin of the CentralApennines, Italy. Catena, 114, 45e58.

Borrelli, P., Robinson, D. A., Fleischer, L. R., Lugato, E., Ballabio, C., Alewell, C.,Meusburger, K., Modugno, S., Schütt, B., Ferro, V., Bagarello, V., Oost, K. V.,Montanarella, L., & Panagos, P. (2017). An assessment of the global impact of21st century land use change on soil erosion. Nature Communications, 8, 2013.

Borrelli, P., Van Oost, K., Meusburger, K., Alewell, C., Lugato, E., & Panagos, P. (2018).A step towards a holistic assessment of soil degradation in Europe: Couplingon-site erosion with sediment transfer and carbon fluxes. EnvironmentalResearch, 161, 291e298.

Colombo, C., & Miano, T. (2015). Metodi di analisi chimica del suolo. Societ�a Italianadi Scienza del Suolo (3� Ed.) Modugno (BA).

Corine Land Cover. (2012). http://www.sinanet.isprambiente.it/it/sia-ispra/download-mais/corine-land-cover/corine-land-cover-2012-iv-livello/view.

Costantini, E., & Lorenzetti, R. (2013). Soil degradation processes in the Italianagricultural and forest ecosystems. Italian Journal of Agronomy, 8(4), e28.https://doi.org/10.4081/ija.2013.e28

Fatica, A., Circelli, L., Di Iorio, E., Colombo, C., Crawford, T. W., & Salimei, E. (2019).Stresses in pasture areas in South-central Apennines, Italy, and evolution atlandscape level. In M. Pessarakli (Ed.), Handbook of plant and crop stress (4th ed.,pp. 271e291). https://doi.org/10.1201/9781351104609

Faulkner, H., Alexander, R., Teeuw, R., & Zukowskyj, P. (2004). Variations in soildispersivity across a gully head displaying shallow sub-surface pipes, and therole of shallow pipes in rill initiation. Earth Surface Processes and Landforms, 29,1143e1160.

Favis-Mortlock, D. (1998). Validation of field scale soil erosion models using com-mon datasets. In J. Boardman, & D. Favis-Mortlock (Eds.), NATO ASI series I: Vol.55. Modelling soil erosion by water (pp. 89e127). Berlin: Springer.

FAO (Food and Agricultural Organization of the United Nations). (2006). Guidelinesfor soil description (4th ed., p. 97). Rome IT EU: Food and Agriculture Organi-zation of the United Nations.

Filocamo, F., Rosskopf, C. M., & Amato, V. (2019). A contribution to the under-standing of the Apennine landscapes: The potential role of Molise Geosites.Geoheritage, 11, 1667e1688. https://doi.org/10.1007/s12371-019-00365-2

Garcìa-Ruiz, J. M., Nadal-Romero, E., Lana-Renault, N., & Beguería, S. (2013). Erosionin Mediterranean landscapes: Changes and future challenges. Geomorphology,198, 20e36.

ISPRA. (2015). Carta Geologica d’Italia alla scala 1:50.000. Roma: Servizio GeologicoD’Italia. System Cart.

IUSS Working Group WRB. (2015). World Reference Base for Soil Resources 2014,update 2015 International soil classification system for naming soils andcreating legends for soil maps. In World soil resources reports No. 106 (p. 192).Roma IT EU: FAO.

Joint Research Centre of the European Commission. (2014). https://esdac.jrc.ec.europa.eu/content/soil-erodibility-k-factor-high-resolution-dataset-europe.

Jones, R. J. A., Grima, M., & Montanarella, L. (2003). Use of Meteorological data sets atEuropean level for input to the PESERA Grid Model. PESERA (Contract QLKS-CT-1999-01323) final report.

Lo, A., El-Swaify, S. A., Dangler, E. W., & Shinshiro, L. (1985). Effectiveness of EI30 aserosivity index in Hawaii. In El-Swaify, & L. Moldenhauer (Eds.), Soil con. Soc.Amer. (pp. 384e392). Ankeny.

Mitasova, H., & Brown, W. M. (2002). Using soil eroision modeling for improvedconservation planning: A GIS-based tutorial geographic modelling system lab.University of Illinois at Urbana- Champaign.

Mitasova, H., Hofierka, J., Zloha, M., & Iverson, R. L. (1996). Modeling topographicpotential for erosion and deposition using GIS”. International Journal ofGeographical Information Science, 10(5), 629e641.

Panagos, P., Borrelli, P., Poesen, J., Ballabio, C., Lugato, E., Meusburger, K.,Montanarella, L., & Alewell, C. (2015). The new assessment of soil loss by watererosion in Europe. Environmental Science & Policy, 54, 438e447.

Panagos, P., Meusburger, K., Ballabio, C., Borrelli, P., & Alewell, C. (2014). Soil erod-ibility in Europe: A high-resolution dataset based on LUCAS. The Science of theTotal Environment, 479e480, 189e200. https://doi.org/10.1016/j.scitotenv.2015.01.008

Parysow, P., Wang, G., Gertner, G., & Anderson, A. B. (2003). Spatial uncertaintyanalysis for mapping soil erodibility based on joint sequential simulation. Ca-tena, 53, 65e78.

Pisano, L., Zumpano, V., Malek, �Z., Rosskopf, C. M., & Parise, M. (2017). Variations inthe susceptibility to landslides, as a consequence of land cover changes: A lookto the past, and another towards the future. The Science of the Total Environment,601e602, 1147e1159.

Pulice, I., Cappadonia, C., Conoscenti, C., Robustelli, G., Rotigliano, E., & Scarciglia, F.(2009). Calanchi landforms in South Italy: A comparison between Calabria andSicilia case studies. In Epitome, 418 pp.

Renard, K. G., Foster, G. R., Weesies, G. A., McCool, D. K., & Yoder, D. C. (1997).Predicting soil erosion by water: A guide to conservation planning with the reviseduniversal soil loss equation (RUSLE) (p. 404). U.S. Department of Agriculture.Agriculture Handbook No 703.

Renard, K. G., & Freimund, J. R. (1994). Using monthly precipitation data to estimatethe R-factor in the revised USLE. Journal of Hydrology, 157, 287e306.

Rosskopf, C. M., & Aucelli, P. P. C. (2007). Analisi del dissesto da frana in Molise.Rapporti APAT, Rapporto sulle frane in Italia. Il progetto IFFIdmetodologia, risultatie rapporti regionali (pp. 493e508).

Torri, D., Borselli, L., Guzzetti, F., Calzolari, M. C., Bazzoffi, P., Ungaro, F., Bartolini, D.,& Salvador Sanchis, M. P. (2006). Italy. In J. Boardman, & J. Poesen (Eds.), Soilerosion in Europe. https://doi.org/10.1002/0470859202.ch20

Van Der Knijff, J. M., Jones, R. J. A., & Montanarella, L. (2000). Soil erosion riskassessment in Europe. EUR 19044 EN, 34pp.

Vezzani, L., Ghisetti, F., & Festa, A. (2004). Carta geologica del Molise (scala 1:100.000). Firenze: S.EL.CA.

Wischmeier, W. H., & Smith, D. D. (1978). Predicting rainfall erosion loss: A guide toconservation planning. In Handbook No 537 (p. 58). U.S. Department of Agri-culture, Agricultural Research Service: Washington D.C..

Yu, B., & Rosewell, C. J. (1996). A robust estimator of the R-factor for the universalsoil loss equation. Transactions of the ASAE, 39(2), 559e561.

Zumpano, V., Pisano, L., Malek, Z., Micu, M., Aucelli, P. P. C., Rosskopf, C. M.,Balteanu, D., & Parise, M. (2018). Economic losses for rural land value due tolandslides. Frontiers of Earth Science. https://doi.org/10.3389/feart.2018.00097


362


Impacts of horizontal resolution and downscaling on the USLE LSfactor for different terrains

Chunmei Wang a, b, c, d, Linxin Shan a, b, e, Xin Liu a, b, Qinke Yang a, b, c, *, Richard M. Cruse d,Baoyuan Liu f, g, Rui Li g, Hongming Zhang h, Guowei Pang a, b, c

a Shaanxi Key Laboratory of Earth Surface System and Environmental Carrying Capacity, Northwest University, Xi’an, 710127, PR Chinab College of Urban and Environmental Sciences, Northwest University, Xi’an, 710127, PR Chinac Key Laboratory of National Forestry Administration on Ecological Hydrology and Disaster Prevention in Arid Regions, Xi’an University of Technology, Xi’an710048, PR Chinad Department of Agronomy, Iowa State University, Ames, IA, 50011, USAe Xi’an Institute of Prospecting and Mapping, Xi’an, 710054, PR, Chinaf Faculty of Geographical Science, Beijing Normal University, Beijing, 100875, PR Chinag State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau, Institute of Soil and Water Conservation, Northwest A&F University,Yangling, 712100, Shaanxi, PR Chinah College of Information Engineering, Northwest A & F University, Yangling, Shaanxi, 712100, China


Article history:Received 1 March 2020Received in revised form8 July 2020Accepted 3 August 2020Available online 11 August 2020

Keywords:LS factorSRTMResolutionDownscaleUSLE

a b s t r a c t

Slope length and slope steepness are critical topographic factors (L and S) in the Universal Soil LossEquation (USLE) and Chinese Soil Loss Equation (CSLE) for soil erosion modelling. Both slope length andslope gradient are potentially sensitive to spatial resolution when calculated in a GIS framework. Theresolution effect on the LS factor and approaches suitable for improving the LS factor at a coarse reso-lution have not been well identified. To address this problem, the LS factor at 5-m and 30-m resolution intwenty-four watersheds with various terrains was estimated. And a downscale model based on matchingof the lower resolution LS cumulative frequency curves to a higher resolution (“Histogram Matching”method) was tested for its potential to improve LS factor estimation accuracy. In the larger reliefmountainous area, compared to 5-m resolution, the 30-m resolution generated LS was generally over-estimated by more than 20% and in lower relief areas underestimated by more than 15%. This bias is lessthan 10% in medium relief areas. The downscale model improved LS factor estimates compared to the 30-m resolution estimate by more than 10% when comparing frequency distribution curves and more than20% in mean values in larger relief areas. The downscale model worked well in all regions except for thelow relief areas, which intuitively are the low soil erosion potential areas. The results of this research helpquantify the uncertainty in soil erosion estimates and may ultimately help to improve the assessment ofsoil erosion through its impact on LS factor estimates, especially at regional and global scales.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction

Topography is an essential factor affecting soil erosion. The slopelength and steepness factors (L and S) are the topographic param-eters in widely used soil erosion models such as USLE (Wischmeier& Smith, 1978), RUSLE (Renard, 1997), and CSLE (Liu et al., 2002).The DEM based procedure for LS factor calculation was developed

(Fu et al., 2015; Hickey et al., 1994; Van Remortel et al., 2001; Zhang,Wei, et al., 2017) subsequent to the development of GeographicalInformation Science (GIS), which has primarily expanded theapplication of USLE and related models, especially for large-scalesoil erosion modelling (Borrelli et al., 2017; Panagos et al., 2015;Yin et al., 2018). Topographic factor accuracy depends on the qualityand resolution of the input elevation dataset (Wilson & Gallant,2000). For the combined effect on the L and S factors, LS is usu-ally calculated together as a single LS factor in GIS applications ofUSLE based erosion models. Exploring the effect of resolution onthe LS factor and its downscaling approach is essential to

* Corresponding author. 1st Xuefu Street, Xi’an, Shaanxi, 710127, PR China.E-mail address: [email protected] (Q. Yang).






understanding the uncertainty in erosion modelling and findingpossible ways to improve erosion estimation. Most researchersagreed that slope gradient decreased with resolution reductionwhile slope length increased (Lu et al., 2020; Wolock & McCabe,2000). The results are not consistent with those for the LS factor.Wu et al. (2005) reported an LS factor decrease with a lower DEMresolution in the State of Virginia, USA; this result was supported byFu et al. (2015) and Lu et al. (2020) within sampled watersheds inChina. Shan et al. (2019) reported an increase of the LS factor asresolution became coarser in an Australia mountain area with highrelief. Under some specific situations, for example, for regions withterraces, an LS factor increase was observed when resolution wentfrom 0.5 m to 30 m (Zhang, Baartman, et al., 2017).

However, most of the research addressing LS factor differencesassociated with DEM resolution variations focused on a local areaor several watersheds with similar topographic characteristics.That might be the reason for conflicting results in the previouslymentioned researches. More exploration across varied terraintypes is needed to reach a more reasonable explanation of reso-lution impact on LS values and to identify possible solutions toreduce the uncertainty caused by a coarser resolution DEM usedwith large-scale erosion modelling. This is especially important atthe 30-m resolution level since several of the global publicelevation datasets, such as the Shuttle Radar Topography Mission(SRTM), are at or near a 30-m DEM. Additionally, SRTM has beenused in large scale erosion modelling (Borrelli et al., 2017) evenwith the lower 3 arc-seconds resolution version. The use of 30-mresolution public elevation datasets would offer better options tofuture global scale erosion and other land surface processmodelling.

The objectives of this study are to identify the influence of 5-mto 30-m resolution DEMs on LS factor estimations in varied terrains,evaluate the uncertainty of these estimates, and build a downscalemodel to reduce this uncertainty. This research will help quantifythe uncertainty and improve the quality of LS factor estimates andultimately help to improve the assessment of soil erosion by usingUSLE related models, especially at regional and global scales.

2. Material and methods

2.1. Study area

Six terrain relief types from different North American regionswere investigated in this study. Regions are classified based onrelative relief, or differences between maximum and minimumelevations in a 25 km2 domain. The relative relief was calculatedbased on DEM with a resolution of 1 km. The relative relief classi-fication levels used in this study are Small (<30-m [SR]), Small andMedium (30-m - 50 m, [SMR]), Medium (50 me150 m [MR]), Largeand Medium (150 me300 m [LMR]), Large (300 me800 m [LR],Extremely Large (�800 m [ELR]). In each region (see Fig. 1), fourwatersheds with area from 6.7 km2 to 16.4 km2 were chosen. Thesewatersheds were selected due to their high-resolution datasetavailability. The detailed description for the twenty-four water-sheds is shown in Table 1.

2.2. Base datasets and data processing

Elevation datasets were the primary inputs for this research. Forthe extraction of the six relief regions, the 1-km resolution GlobalMulti-resolution Terrain Elevation Data 2010 (“GMTED2010” forshort) (Danielson & Gesch, 2011) developed by United StatesGeologic Survey (USGS) in 2010 was used. At the watershed scale,two elevation datasets were used as base data for LS factor calcu-lation. The 3D Elevation Program (3DEP) 1/9 arc-second

(approximately 3 m) product (USGS, 2017) derived from Lidarclouds was used as a high-resolution data source. The NASA SRTMVersion 3.0 Global 1 arc-second (SRTMGL1v003) (approximately30 m) (NASA, 2013) was used as a low-resolution data source.

Both the 3DEP and the SRTMGL1v003 datasets were projectedto user-defined UTM projection choosing the central watershedlongitude as the central longitude in UTM and resampled to 5-mand 30-m separately. Subsequently, a Gaussian low pass filterwith a neighborhood of 3 cells by 3 cells was applied.

2.3. Procedures of the experiment

The experimental procedures are shown in Fig. 2. The 1-kmglobal relief dataset was obtained from GMTED2010 to classifythe land into six relative relief regions. Watersheds in each regionwere selected based on the availability of fine-resolution elevation.Then within each watershed, the 1/9 arc-second 3DEP dataset andthe 1 arc-second SRTMwere obtained from the USGS website. Afterdata preprocessing (section 2.2), the fine-resolution DEM (5m) andthe coarse-resolution DEM (30 m) were ready for LS calculationusing the software LS-Tools (Zhang, Wei, et al., 2017). Statistics ofthe resolution influence on LS were conducted based on the tworesolution datasets. Subsequently, the LS Downscale models in thesix regions were generated.

2.4. Relative relief calculation

The relative relief was calculated based on the 1-kmGMTED2010 dataset and used to select samples for analysis. AFocal analysis used the elevation range value of the neighboring5� 5 pixels to identify the relative relief value of the centralpixel.

2.5. LS factor algorithm

The LS factor raster dataset was calculated from the slopegradient and slope length raster according to the algorithmdeveloped by Liu et al. (2002). The calculation of S varies from thealgorithm used in the USLE (McCool et al., 1987) for locationswhere the slope gradient is larger than 10�. The regions in thisresearch included steeply sloping areas, the reason why the al-gorithm developed by Liu et al. (2002) was used. The LS factor wascalculated using the software LS-Tool (Zhang, Wei, et al., 2017).The software was first published by Zhang et al. (2013). Animproved version was developed by Zhang, Wei, et al. (2017) andcould be obtained by personal communication Zhang, Wei, et al.(2017).

The most time-consuming and challenging part of the LScalculation is the slope length derivation (Hickey et al., 1994; VanRemortel et al., 2001). In this research, the Multiple-Flow Direc-tion (MFD) algorithm was used in calculating slope length byusing the latest version of LS-Tools (Zhang, Wei, et al., 2017). Ac-cording to the definition of slope length given by (Wischmeier &Smith, 1978), two types of locations for hillslope slope lengthcut-points were used. The first is the center of the cell in whichdeposition initially occurs, which means the location where theslope gradient decreased to 70% of the upper slope gradient withupper slope gradient <5%, and to 50% with upper slope gradient>5%. The other type is where the hillslope meets a channel; thechannel cells were defined in the program by setting a thresholdof contributing drainage area. In each region, at each resolution,this contributing drainage area threshold was selected based onthe visualization of channels within each elevation dataset. Thechannels here referred to the end level of the channel that couldbe observed on the elevation dataset. Where channels began, the

C. Wang, L. Shan, X. Liu et al. International Soil and Water Conservation Research 8 (2020) 363e372

364

flow direction extracted based on the elevation dataset tended tobe more concentrated on the channelized pixels, and also thecolor on the DEM changed more than the hillslope pixels. Tenchannel initiation points from each watershed were observed, anda mean contributing drainage area value for all points for eachregion and resolution was used as the threshold value in theprogram and is shown in Table 2.

2.6. LS downscale method

The LS downscale model was developed based on the HistogramMatching theory often used in digital image processing. Thismethod was used by Yang et al., (2008) in the field of digital terrainanalysis for slope gradient downscaling. It was based on the rela-tionship between cumulative frequencies of fine and coarse-

Fig. 1. Study regions.

Table 1Basic information of each sampled watershed based on 5-m resolution.

Region Name Mean elevation (m) Elevation range (m) Mean slope gradient (�) Mean slope Length (m) Area (km2) Longitude Latitude

SR SR01 411.4 58.5 3.5 74.6 8.3 97.336� W 39.341� NSR02 421.3 48.3 2.7 59.3 6.7 97.381� W 39.477� NSR03 450.5 72.2 4.3 64.9 15.0 97.485� W 39.360� NSR04 429.1 40.9 2.6 53.3 11.3 97.439� W 39.460� N

SMR SMR01 296.0 116.0 10.1 72.2 9.7 81.566� W 39.735� NSMR02 273.9 113.1 11.7 71.5 8.0 81.638� W 39.652� NSMR03 290.8 100.9 10.5 62.8 9.1 81.633� W 39.914� NSMR04 308.4 92.9 8.2 94.0 7.2 81.701� W 39.917� N

MR MR01 371.2 256.0 22.8 126.0 9.4 80.569� W 39.609� NMR02 348.6 228.4 21.3 127.5 10.9 80.705� W 39.653� NMR03 339.3 197.2 20.4 112.6 12.1 80.682� W 39.305� NMR04 340.7 217.3 20.9 111.1 16.4 80.621� W 39.353� N

LMR LMR01 493.8 371.7 27.5 103.0 9.4 81.805� W 37.523� NLMR02 450.0 384.9 27.0 107.6 10.5 81.923� W 37.559� NLMR03 475.5 446.8 28.1 118.6 10.6 81.813� W 37.636� NLMR04 496.5 451.0 30.3 117.5 15.8 81.922� W 37.660� N

LR LR01 947.0 585.1 28.2 107.7 11.0 123.707� W 42.754� NLR02 717.6 910.5 31.7 115.5 10.3 123.636� W 42.868� NLR03 800.5 822.4 31.6 115.1 15.8 123.655� W 42.711� NLR04 526.7 722.5 29.0 96.4 7.8 123.543� W 42.632� N

ELR ELR01 1192.1 1251.7 27.4 128.0 11.7 121.601� W 48.638� NELR02 1055.1 1149.2 24.3 114.0 10.5 121.653� W 48.586� NELR03 1124.3 1138.6 26.9 150.6 6.7 121.598� W 48.541� NELR04 1100.7 1102.1 27.3 127.9 8.6 121.557� W 48.534� N

*SR: small relief; SMR: small and medium relief; MR: medium relief; LMR: large and medium relief; LR: large relief; ELR: extremely large relief; The elevation range refers tothe maximum elevation minus the minimum elevation in each watershed.


365

resolution terrain factors. Four steps for LS downscale processes areillustrated in Fig. 3. In the first step illustrated in Fig. 3 (a), thecumulative frequency curves were obtained based on both fine andcoarse-resolution LS rasters. Then pairs of LS values with the samecumulative frequencies were identified and summarized in a tablein the second step shown in Fig. 3 (b). The third step shown in Fig. 3(c) involves developing the mathematical model with coarse res-olution LS values as X and fine resolution LS values as Y based onthe table generated by step two. The last step shown in Fig. 3 (d)was to apply the mathematical model built in step three in theArcGIS raster calculator, with the coarse resolution LS raster datasetas X, and the downscaled LS raster dataset as the Y raster in theoutput..

Assuming the fine resolution LS factor is more accurate than thecoarse, this method adjusts the values of the coarse LS factor ac-cording to the fine LS factor to make the former more closelyapproximate the fine resolution dataset, without changing the pixelsize of the raster dataset. We anticipated the coarse and fine res-olution LS would be different in varied terrain types, which makesit important to classify the topographic surface into differentregions.

There are four watersheds in each relative relief region. For eachregion, a unique downscale model was generated based on theweighted average of frequencies at the four watersheds and then tocreate averaged coarse and fine resolution cumulative frequencycurves, which helped to generate an overall downscale model (Eq.(1))

PðXÞ¼Xn

i¼1

si � PðxiÞ,Xn

i¼1

si (1)

where PðXÞ refers to the weighted average LS frequency value at

Fig. 2. Procedures of the experiment.

Table 2Threshold of contributing drainage area for channel cut for the slope length calcu-lation (m2).

Region 5 m 30 m

SR 33,500 71,100SMR 30,000 64,800MR 29,500 53,100LMR 29,400 53,100LR 57,500 90,000ELR 120,900 160,200

*SR: small relief; SMR: small and medium relief; MR: medium relief; LMR: large andmedium relief; LR: large relief; ELR: extremely large relief.

Fig. 3. Illustration of the four steps for the LS Downscaling method.


366

each step of the frequency curve; PðxiÞ refers to the frequency valueat that step in the ith watershed; Si refers to the drainage area of theith watershed. Here in this research, n equals 4.

2.7. Statistics

Relative Error (RE) was used to quantify the LS factor mean valueaccuracy. It was defined as in Eq. (2).

RE¼�meanc �meanf

�.meanf � 100% (2)

Where RE refers to Relative Error; meanC refers to the mean LSfactor value at the coarse resolution; meanf refers to the LS meanfactor value at the fine resolution, which was defined as thereference value. This parameter was used to quantify how muchuncertainty of the mean LS factor value would be attributed to theresolution difference from 5-m to 30 m. Moreover, it was used forquantifying the changes attributed to the LS downscale models onmean LS values. A larger RE value indicates a poorer estimate of theLS factor.

Relative Similarity of Cumulative Frequency was used to quan-tify the similarity of two cumulative frequency curves, and it wasdefined as Eq. (3).

RSCF ¼�1�

��Sf � Sc��.Sf�� 100% (3)

Where RSCF was Relative Similarity of Cumulative Frequency, Sfreferred to the area between the cumulative frequency curve andthe X-axis of the fine-resolution LS factor; Sc referred to the areabetween the cumulative frequency curve and the X-axis of thecoarse-resolution LS factor.

The larger the RSCF value, the more similar are the cumulativefrequency curves. In this research, The RSCF value was used as thesecond parameter to show the downscale model efficiency. Themodel worked better if the RSCF value between the downscaled LSfactor and the 5-m LS factor is larger than that between the original30-m LS factor and the 5-m LS factor, which is three cumulativefrequency distribution curves we involved. RSCF estimates arebased on the area under the three cumulative distribution curvesfrom 0% to an arbitrarily defined cumulative frequency value. Thisvalue was set as the largest of the LS values representing 95% cu-mulative distribution among the three curves.

3. Results

3.1. LS scaling laws in varied terrain types

Fig. 4 (a) displays the mean LS factor value for each watershed at

5-m and 30-m resolution, and Fig. 4 (b) shows the RE (RelativeError, Eq. (2)) values of the LS factor, S factor, and L factor for eachregion. The mean LS factor values for the 30-m resolution for wa-tersheds within SR, SMR, and MR were small relative to those forthe 5 m resolution, where the RE values of the LS factor werenegative. And 30 m resolution LS factor mean values were largerwithin LMR, LR, and ELR, where the RE of LS factor was positive. Thelargest RE value for the LS factor value between 30 m and 5 mresolution, 56.0%, occurred in watersheds within the ELR region.The absolute RE value was as small as -3.8%, which occurred in theMR region. The RE value of the S factor was always negative; it wasalways positive for the L factor within each region. In summary, theS factor decreased as resolution became coarser, and the L factorconcurrently increased. The absolute values for the S factor REvalues were larger than those for L in SR and SMR regions, whichresulted in a negative RE value for the LS factor. This observationwas opposite in LMR, LR, and ELR regions, where the absolute REvalues of the L factor are greater than those for S factor. For the MRregion the absolute RE value of the L factor was slightly larger thanthat of the S factor, while the LS value was still shown to be slightlydecreasing.

Fig. 5 shows the influence of resolution on the frequency andcumulative frequency curves of the LS factor. The cumulative fre-quency curves of LS at 30 m resolution generally moved to the leftside of 5 m resolution for SR, SMR, and MR regions, while it movedto the right side for LR and ELR regions; for the LMR region, itmoved to the left side before the LS value reached 15, while itmoved to the right side after that. The LS values are underestimatedfor coarser resolution in smaller relative relief regions and over-estimated in larger relative relief regions.

Fig. 6 (a) and (b) show the land surface of the different water-sheds spatially color-coded to depict the estimated LS factordetermined from DEMs with 5-m and 30-m resolutions. Note lessred and yellow color (lower frequency of high LS values, see Fig. 6)exists in SR and SMR watersheds for the 30-m resolution than forthe 5 m resolution. The color-coded surfaces are quite similar forboth resolutions within the MR region. In LMR, LR, and ELR regions,a frequency of higher LS values increased for 30-m resolutioncompared to 5-m resolution estimates.

In general, the scaling effect on LS factor estimates varies byregion depending on region terrain characteristics. The mean value,frequency distribution, and spatial visualization give consistentevidence of the scaling effect. In some regions, for example, SR,SMR, and MR regions, LS factors derived from a coarser DEM aresmaller than those derived from a finer DEM. And in other regions,for example, the LMR, LR, and ELR regions, the opposite seems to betrue. Whether the LS factor is likely to be underestimated oroverestimated when using the coarser DEM dataset is determined

Fig. 4. LS Mean Value difference between 30-m to 5-m resolution for the six regions.


367

by the relative absolute RE values of the S or the L factor. The nextstep in addressing this LS scaling factor uncertainty is to test thedownscaling impact on LS predictability.

3.2. LS downscaling model and application

Table 3 shows the downscale models for each region based onthe histogrammatchingmethod described in Section 2.5. In Table 3,y refers to the downscaled values of the LS factor, and x refers to theLS values at 30 m resolution. N refers to the number of the LS pairs(between high and low resolutions, see Fig. 3 (b)) with similarcumulative frequency numbers; the values of N is decided by thecumulative distribution curve step of LS and the maximum, mini-mum values of LS in the specific area. R2 refers to the coefficient ofdetermination of the model, a higher R2 value indicates a bettermodel performance and less difference between modeled andcalculated y value in the cumulative distribution curve. The R2

values are larger than 0.95 in all regions. The relationship betweenthe 30-m resolution LS factor and 5 m resolution LS factor is mostlylinear, except in the low relief terrain.

Fig. 6 (c) shows the spatial distribution of estimated LS factorsafter the downscale process. The downscaled LS factor surface wasmore spatially similar to the 5-m LS factor (Fig. 6 (a)) than the 30-mresolution (Fig. 6 (b)) based LS surface. The improvement of LSspatial representation is more obvious in SR, LR, and ELR regions.

The effect of downscaling on LS predictability is illustrated inFig. 7; 30-m downscaled LS mean values are plotted against the 5-m mean LS values across all terrains (in all of the 24 watersheds).Compared to the mean LS values at 30 m resolution before down-scaling shown in Fig. 4 (a), after downscaling, the mean LS valueswere closer to 5-m LS values as the points in Fig. 7 more closelyapproximate a 1:1 line.

Fig. 8 shows the cumulative frequency curves of 5-m, 30-m, anddownscaled LS factor for each relative relief area. The downscaled

Fig. 5. Frequency and cumulative frequency curves of the LS factor with 5-m and 30-m resolutions for the six regions.


368

Fig. 6. Spatial distributions of 5-m, 30-m resolution and downscaled LS factor in typical watersheds in the six regions.


369

LS cumulative frequency curves more closely approximate the 5-mresolution LS curves. Table 4 shows the changes in RE of LS and RSCF

values before and after downscaling. In most regions, the down-scale model worked well; it decreased the RE values and increasedthe RSCF values. The downscale model worked best with thegreatest relief, ELR, in this study, for which the RE decreased from56% to -1.39%, and RSCF improved by 23.63%. The downscalingprocess improved RE for all regions except for the region with thesmallest relief (SR). The downscale model has the greatest benefitfor estimating the LS factor in regions most suspectable to soilerosion losses e the areas with the greatest relief.

4. Discussion

Based on this study, L factor increases, and S factor decreases asthe input DEM resolution becomes coarser, which agrees withprevious research (Fu et al., 2015; Lu et al., 2020; Shan et al., 2019;Wu et al., 2005). However, the DEM resolution impact on the LSfactor varied between regions due to differences in relative relief;these results are not always aligned with previous research results(Fu et al., 2015; Lu et al., 2020; Shan et al., 2019; Wu et al., 2005).The LS factor determined from 30-m resolution compared to 5-mresolution could be underestimated as much as 19% and over-estimated by 20% or more (in extremely large relief areas, could be56%). This work suggests both overestimates and underestimatesare a reality depending on the terrain of the region. Moreover, theerrors seem substantial and should not be ignored in erosionmodelling.

The reason for different conclusions between this research andprevious studies may be attributed to the different terrain regionschosen for the study. For example, Lu et al. (2020) showed that theLS factor would decrease with resolution coarseness in Chinasampled regions while Shan et al. (2019) showed it would increase.Most of the sampled watersheds in Lu’s research were located inhilly areas without extremely large relief, while Shan et al. (2019)focused on large relief Australian mountain areas. In most cases,researchers considered the S decrease associated with coarserresolutions as the driver of the LS factor changes. This seemsespecially true as much of the research focuses on hilly or agri-cultural small relief areas. In some large relief mountain areas, suchas the research area of Shan et al. (2019) and the ELR region in thisresearch, there may be some deposition areas predicted at 5 mresolution that are not predicted with 30 resolution. Additionally,

Table 3LS factor downscale model from 30-m to 5-m in each region.

Region Downscale Model N p-value R2

SR y ¼ 0:2653x2 þ 0:6019xþ 0:5742 50 <0.0001 0.959SMR y ¼ 0:0465x2 þ 0:272xþ 3:0104 39 <0.0001 0.960MR y ¼ 0:0065x2 þ 0:8067xþ 1:6046 63 <0.0001 0.974LMR y ¼ 0:0092x2 þ 0:4964xþ 3:6975 85 <0.0001 0.989LR y ¼ 0:0031x2 þ 0:5992xþ 2:9317 107 <0.0001 0.984ELR y ¼ 0:0014x2 þ 0:5774x� 0:0695 156 <0.0001 0.994

*SR: small relief; SMR: small and medium relief; MR: medium relief; LMR: large andmedium relief; LR: large relief; ELR: extremely large relief. Here y refers to thedownscaled values of the LS factor and x refers to the LS values at 30-m resolution. Nrefers to the number of LS pairs (between high and low resolution) with similarcumulative frequency numbers, the values of N are decided by the cumulativedistribution curve step of LS and the maximum, minimum values of LS factor in thespecific area. R2 refers to the coefficient of determination of model, a higher R2 in-dicates a better model performance and less difference between modeled andcalculated y value in the cumulative distribution curve.

Fig. 7. Comparison of high resolution (5-m) and the downscaled LS factor mean valuesfor each watershed.

Fig. 8. Cumulative frequency curve differences between 5-m, 30-m, and the downscaled LS factor for the six regions.


370

channel development less frequently exists in mountainous areascompared to the hilly area; this results in much longer hillslopewith coarser resolution than that which occurs at a finer resolution.However, the slope gradient would not decrease to a large extent inthis situation as few small channels or high-frequency terrainstructures exists. Thus, the relative decrease of S is smaller than theincrease of L, resulting in the increasing LS factor.

The threshold value of the contributing drainage area for achannel cut is one of the essential parameters in calculating slopelength. In this paper, this threshold value was calculated based onvisually surveyed channel heads in a specific area (Table 2). Thesurvey showed the threshold values decreased from SR to LRM,because with steeper slopes, normally a smaller amount ofconcentrated flow is required to make the channel process happen.In LR and ELR, the threshold increased because high-density forestcovered the ground and protected the surface from sheet and rillerosion and channel formation. Since in some studies, thisthreshold would be the same values for different regions, we testedthe LS sensitivity to lower resolution by using the unique thresholdcontribution area value 30000 m2. The LS factor could still be bothunderestimated (in smaller relief regions) and overestimated (inlarger relief regions) by using coarse resolution DEMs. The spatialdistribution of this threshold value needs further study since it isimportant in the application of the USLE (Wischmeier & Smith,1978), RUSLE (Renard, 1997), CSLE (Liu et al., 2002), and othermodels, especially in large scale soil erosion modelling.

The downscale model presented, worked well in most of theregions, especially in large and extremely large relief areas wherethe LS factor was seriously overestimated without downscaling.This will help to improve the accuracy of soil erosion estimateswhen using the USLE and related models. However, the down-scaling model did not work well in flat areas with relief lower than30 m. The reason for this failure might be the relatively low valuesof the LS factor and the more skewed distribution in the LS fre-quencies found in the SR region (Fig. 5 (a)). It is more difficult to finda unique relationship between LS pairs (shown in Fig. 3 (b)) fordifferent steps of LS values based on more skewed distributions.The mathematical model for downscaling (shown in Fig. 3 (c))would work well for some steps while poor for other steps. Cautionis advised for using this downscaling approach in flat areas, such asareas where the LS mean values are lower than 2. Worth empha-sizing, this approach seems to best work for areas with high soilerosion potentials, and it seems unnecessary for areas that havelimited erosion potential based on low LS values. The downscalemodel needs further improvement before it could be applied uni-versally. We also observed bimodal frequency distribution of LSestimates in some of the watersheds (see Fig. 5 (c) and Fig. 5 (d)),which may also need more exploration as we propose to use LSdownscaling for coarse resolution DEMs.

We acknowledge there were some limitations to this research.The samples are all located in North America due to the avail-ability of higher resolution elevation data. The twenty-four wa-tersheds selected in this research may not cover all of the terraintypes found globally. So, more applications and calibration workwould be useful across more diverse terrains. Since the downscalemodel did not change the raster cell size, it could not get back thedetailed terrain information lost in the coarse resolution DEMsurface. It worked for correcting the LS factor statistics in awatershed and then in a region, but not for pixel or hillslope scaleuse.

The coarser resolution elevation datasets, for example, the 1 arc-sec SRTM, might be the possible choice in large scale erosionassessment where finer resolution datasets are unavailable. How-ever, we should always be aware that erosionmodelling results willlikely be overestimated or underestimated based on LS estimatesderived from coarser DEMs. The opportunity to better estimate theLS factor used in many soil erosion models by downscaling tech-niques offers important improvement for making soil erosion es-timations. This research preliminarily provides the variation of LSfactor scaling relationships in space and a downscale model forimproving LS factor estimates. More efforts are needed to identifyspatial LS factor scaling laws and identify the appropriate applica-tion of the downscaling method to make the erosion predictionmore accurate for the large scale.

5. Conclusions

The difference in LS factor estimates based on the 30-m reso-lution SRTM data source, and 5-m resolution elevation from 3-mLidar datasets was identified. And then a downscale model wasoffered in this research within twenty-four watersheds in six reliefregions in North America. We could conclude based on the resultsof this research that:

(1) Slope length factor (L) increased, and slope steep factor (S)decreased with resolution becoming coarser, from 5 m to30 m.

(2) From fine to coarse resolution, the LS factor can bothdecrease (in small relief regions) and increase (in large reliefregions). And the influence of resolution on the LS factorcannot always be ignored as the bias could be more than 20%in some regions.

(3) A downscaling model is useful in improving regional LS ac-curacy, which improved the LS factor estimate in most of theregions. It was little to no improvement for low relief wa-tersheds, i.e., watersheds in which soil erosion tends to beless serious.

Table 4Change of LS factor after downscaling.

Regions SR SMR MR LMR LR ELR

LS mean (5-m) 1.24 5.38 15.91 19.63 21.20 20.03LS mean (30-m) 1.00 4.44 15.31 21.05 25.85 31.25LS mean (Downscale) 1.64 5.74 15.91 19.26 21.08 19.76RE (5-m, 30-m) -19.39% -17.47% -3.80% 7.25% 21.91% 56.00%RE (5-m, Downscale) 32.32% 6.64% -0.03% -1.90% -0.59% -1.39%RE change -12.93% 10.83% 3.77% 5.35% 21.32% 54.61%RSCF(5-m,30-m) 93.60% 89.94% 96.68% 92.54% 84.70% 73.17%RSCF(5-m, downscale 30-m) 81.78% 94.43% 98.71% 96.12% 95.57% 96.80%RSCF change -11.82% 4.49% 2.03% 3.58% 10.87% 23.62%

*SR: small relief; SMR: small and medium relief; MR: medium relief; LMR: large and medium relief; LR: large relief; ELR: extremely large relief. RE change means absolutevalue of RE (5-m, 30-m) minus absolute value of RE (5-m, Downscale), RSCF change means RSCF (5-m, downscale 30-m) minus RSCF (5-m,30-m). Both changes could indicatethe model works well if the change in value is positive.


371

Conflict of interest

The authors declare that they have no known competingfinancial interests or personal relationships that could haveappeared to influence the work reported in this paper.

Acknowledgements

This research was supported by the Strategic Priority ResearchProgram of the Chinese Academy of Sciences, Grant No.XDA20040202, SKL Foundation Grant No. A314021402-1718, Na-tional Natural Science Foundation of China, Grant No. 41977062,41601290, 41771315, 41930102, Program for Key Science andTechnology Innovation Team in Shaanxi Province, Grant No.2014KCT-27. Part of the work was finished during Chunmei Wang'svisit to the Department of Agronomy, Iowa State University.

References

Borrelli, P., Robinson, D. A., Fleischer, L. R., Lugato, E., Ballabio, C., Alewell, C., et al.(2017). An assessment of the global impact of 21st century land use change onsoil erosion. Nature Communications, 8(1), 1e13.

Danielson, J. J., & Gesch, D. B. (2011). Global multi-resolution terrain elevation data2010 (GMTED2010). Open-File Report.

Fu, S., Cao, L., Liu, B., Wu, Z., & Savabi, M. R. (2015). Effects of DEM grid size onpredicting soil loss from small watersheds in China. Research Journal of Envi-ronmental and Earth Sciences, 73(5), 2141e2151.

Hickey, R., Smith, A., & Jankowski, P. (1994). Slope length calculations from a DEMwithin ARC/INFO GRID. Computers, Environment and Urban Systems, 18(5),365e380.

Liu, B. Y., Zhang, K. L., & Xie, Y. (2002). An empirical soil loss equation. 12th ISCOConference.

Lu, S., Liu, B., Hu, Y., Fu, S., Cao, Q., Shi, Y., & Huang, T. (2020). Soil erosion topo-graphic factor (LS): Accuracy calculated from different data sources. CATENA,187(4), 104334.

McCool, D., Brown, L., Foster, G., Mutchler, C., & Meyer, L. (1987). Revised slopesteepness factor for the universal soil loss equation. Transactions of the ASAE

(American Society of Agricultural Engineers), 30(5), 1387e1396.NASA. (2013). NASA Shuttle radar Topography mission global 1 arc second [data

set]. In NASA EOSDIS land processes DAAC. N. JPL.Panagos, P., Borrelli, P., Poesen, J., Ballabio, C., Lugato, E., Meusburger, K.,

Montanarella, L., & Alewell, C. (2015). The new assessment of soil loss by watererosion in Europe. Environmental Science & Policy, 54, 438e447.

Renard, K. G. (1997). Predicting soil erosion by water: A guide to conservation planningwith the revised universal soil loss equation. RUSLE) United States GovernmentPrinting.

Shan, L., Yang, X., & Zhu, Q. (2019). Effects of DEM resolutions on LS and hillslopeerosion estimation in a burnt landscape. Soil Research, 57(7), 797e804.

USGS. (2017). 1/9th arc-second digital elevation models (DEMs) - USGS national map3DEP downloadable data collection. Reston, VA, USA: U.S. Geological Survey.

Van Remortel, R. D., Hamilton, M. E., & Hickey, R. J. (2001). Estimating the LS factorfor RUSLE through iterative slope length processing of digital elevation datawithin Arclnfo grid. Cartography, 30(1), 27e35.

Wilson, J. P., & Gallant, J. C. (2000). Digital terrain analysis. Terrain analysis: Principlesand applications, 6(12), 1e27.

Wischmeier, W. H., & Smith, D. D. (1978). Predicting rainfall erosion losses: A guide toconservation planning. Department of Agriculture, Science and EducationAdministration.

Wolock, D. M., & McCabe, G. J. (2000). Differences in topographic characteristicscomputed from 100-and 1000-m resolution digital elevation model data. Hy-drological Processes, 14(6), 987e1002.

Wu, S., Li, J., & Huang, G. (2005). An evaluation of grid size uncertainty in empiricalsoil loss modeling with digital elevation models. Environmental Modeling &Assessment, 10(1), 33e42.

Yang, Q., Jupp, D., Li, R., & Liang, W. (2008). Re-scaling lower resolution slope byhistogram matching (pp. 193e210). Springer.

Yin, S., Zhu, Z., Wang, L., Liu, B., Xie, Y., Wang, G., et al. (2018). Regional soil erosionassessment based on a sample survey and geostatistics. Hydrology and EarthSystem Sciences, 22, 1695.

Zhang, H., Baartman, J. E., Yang, X., Gai, L., & Geissen, V. (2017). Influence of terracedarea DEM resolution on RUSLE LS factor. EGU General Assembly Conference Ab-stracts, 11517.

Zhang, H., Wei, J., Yang, Q., Baartman, J. E., Gai, L., Yang, X., et al. (2017). An improvedmethod for calculating slope length (l) and the LS parameters of the RevisedUniversal Soil Loss Equation for large watersheds. Geoderma, 308(12), 36e45.

Zhang, H., Yang, Q., Li, R., Liu, Q., Moore, D., He, P., et al. (2013). Extension of a GISprocedure for calculating the RUSLE equation LS factor. Computers & Geo-sciences, 52(3), 177e188.


372


Effect of time resolution of rainfall measurements on the erosivityfactor in the USLE in China

Tianyu Yue a, Yun Xie a, Shuiqing Yin a, *, Bofu Yu b, Chiyuan Miao a, Wenting Wang a

a State Key Laboratory of Earth Surface Processes and Resource Ecology, Faculty of Geographical Science, Beijing Normal University, 100875, Beijing, Chinab Australian Rivers Institute, School of Engineering and Built Environment, Griffith University, Nathan, Queensland, Australia


Article history:Received 13 March 2020Received in revised form5 June 2020Accepted 5 June 2020Available online 19 June 2020

Keywords:Soil erosionRainfall erosivity1-in-10-year EI30USLECSLE

a b s t r a c t

Rainfall erosivity, one of the factors in the Universal Soil Loss Equation, quantifies the effect of rainfall andrunoff on soil erosion. High-resolution data are required to compute rainfall erosivity, but are not widelyavailable in many parts of the world. As the temporal resolution of rainfall measurement decreases,computed rainfall erosivity decreases. The objective of the paper is to derive a series of conversion factorsas a function of the time interval to compute rainfall erosivity so that the R factor computed using data atdifferent time intervals could be converted to that computed using 1-min data. Rainfall data at 1-minintervals from 62 stations over China were collected to first compute the ‘true’ R factor values. Under-estimation of the R factor was systematically evaluated using data aggregated at 5, 6, 10, 15, 20, 30, and60-min to develop conversion factors for the R factor and the 1-in-10-year storm EI30 values. Comparedwith true values, the relative error in R factor using data at fixed intervals of �10min was <10% for atleast 44 out of 62 stations. Errors increased rapidly when the time interval of the rainfall data exceeded15 min. Relative errors were >10% using 15-min data for 66.1% of stations and >20% using 30-min datafor 61.3% of stations. The conversion factors for the R factor, ranging from 1.051 to 1.871 for 5 to 60-mindata, are higher than those for the 1-in-10-years storm EI30, ranging from 1.034 to 1.489 for the 62stations.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction

Rainfall erosivity factor (R factor) is an important factor for soilerosion prediction models such as the Universal Soil Loss Equation(USLE, Wischmeier & Smith, 1965; Wischmeier & Smith, 1978) andits revised version (RUSLE, Renard, 1997) and the Chinese Soil LossEquation (CSLE, Liu et al., 2002). It reflects the potential ability ofrainfall on soil erosion, including raindrop splashing and runofferosion and determines the actual soil loss in an area together withother factors such as soil erodibility, topography, coverage, and soilconservationmeasures. Wischmeier and Smith (1958) note that theevent EI30, the product of total storm kinetic energy (E) and itsmaximum continuous 30-min intensity (I30), has the best correla-tion with single event soil erosion based on an analysis of precip-itation and soil erosion data from fallow plots at three observationstations in the United States. R factor is defined as the mean annual

summation of event EI30 values and Wischmeier (1959) confirmedits suitability based on 8000 plot-years data in 21 states in theeastern part of the United States. To cover dry and wet periods,rainfall data of long periods (usually more than 20 years) areneeded for the calculation of the R factor (Wischmeier, 1976). Inaddition, there were twomore aspects of the characteristics relatedto the rainfall erosivity factor required to be prepared before theUSLE (RUSLE, CSLE) models were run. The first is the seasonaldistribution of the rainfall erosivity, which is defined as the ratio ofthe rainfall erosivity in 24 half months to annual rainfall erosivityand required for the calculation of the cover-management factor (Cfactor) in the USLE (RUSLE) and the biological and tillage practicefactor (B and T factor) in CSLE (Liu et al., 2002). The second is the 1-in-10-year storm event EI30, which has the probability of occurringonce every 10 yr. The 1-in-10-year EI30 index is required forassessing the effectiveness of the terrace practice and the effect ofponding on rainfall erosivity. Usually, an extreme value distributionsuch as lognormal (Wischmeier & Smith, 1978) or the generalizedextreme value distribution (Hollinger et al., 2002) was first used tofit the annual maximum event EI30 series and the 1-in-10-year

* Corresponding author.E-mail address: [email protected] (S. Yin).






event EI30 was then generated based on the calibrated extremevalue distribution. To ensure the sample size is sufficiently large toobtain a reasonably good fit with the extreme value distribution, aminimum of 20 to 25 years of data would be required.

Calculating rainfall erosivity originally requires breakpoint orhyetograph rainfall data which were obtained from pluviograph(Wischmeier & Smith, 1978), which were usually in shortage notonly in the length but also in the spatial coverage. Therefore, sta-tistical models have been developed to use more commonlyavailable data, such as daily (e.g., Richardson et al., 1983; Xie et al.,2016), monthly (e.g., Renard & Freimund, 1994; Yu & Rosewell,1996) and annual rainfall (e.g., Lo et al., 1985; Renard &Freimund, 1994) to estimate R factor. Yin et al. (2015) evaluatedthe accuracy of models using different resolutions of data and re-ported that generally the precision of models increased with thetemporal resolution of precipitation and daily precipitation data areadequate to estimate the R factor whereas they often generatelarger errors for the estimation of event EI30 index. With thedevelopment of automatic observations in the tipping bucket raingauges, high-temporal resolution fixed-interval data, such as 1, 5, 6,10, 15, 20, 30 and 60-min rainfall data, are increasingly availableand they are believed to improve the estimation of rainfall erosivity,especially the event EI30 index. Fixed-interval data have beenwidely used in the estimation of rainfall erosivity and most of theprevious studies used data with intervals of 10-min or less,including 5-min (Diodato et al., 2017; Fiener et al., 2013; Martinset al., 2010; Neuhaus et al., 2010, pp. 186e189), 6-min (Mcfarlaneet al., 1986; Yu, 1998; Yu et al., 2010) and 10-min (Hanel et al.,2016; Klik et al., 2015; Meusburger et al., 2012; Santosa et al.,2010; Schmidt et al., 2016; Verstraeten et al., 2006). Some othersused 15-min or 30-min interval data (Angulo-Martínez & Beguería,

2009; Mannaerts & Gabriels, 2000; Panagos et al., 2016; Shamshadet al., 2008). Rainfall intensity data at 5 to 30-min intervals havebeen used to compute the R factor in the studies cited above.

Many researchers have reported that there were varying de-grees of discrepancy when using different resolutions of fixed-interval data. The longer the time interval, the larger the under-estimation. The underestimation could be corrected by multiplyinga conversion factor (Agnese et al., 2006; Istok et al., 1986; Panagoset al., 2015; Porto, 2016; Renard, 1997; Weiss, 1964; Williams &Sheridan, 1991; Yin et al., 2007). For example, The RUSLE (Renard,1997) updated the map of the USLE for the eastern part of theU.S. by including 790 stations with 60-min interval rainfall data.The conversion factors ranged from 1.08 to 3.16 between the EI30values calculated from the15-min and 60-min data and a constantconversion factor of 1.0667 (Weiss, 1964) between the EI30 from the

Fig. 1. Spatial distribution of the 62 stations with 1-min rainfall data. Thirty-three stations marked with dots were classified as Group A, whereas the remaining 29 stations markedwith circles as Group B.

Fig. 2. Ratio of rainfall erosivity estimated from the 5 to 60-min rainfall data to thoseestimated from the 1-min data.

T. Yue et al. / International Soil and Water Conservation Research 8 (2020) 373e382374

15-min data and the breakpoint data was further adopted. Anotherexample is the generation of rainfall erosivity maps in Europe.Panagos et al. (2015) collected 1541 precipitation stations in Europewith intervals of 5 to 60 min and normalized the R factor valuescalculated from the precipitation data of different temporal reso-lutions to the R factor values with 30-min intervals using linear

regression methods. Auerswald et al. (2015) argued that Panagoset al. (2015) used the maximum clock half-hourly intensity ratherthan I30 using high-resolution breakpoint data and the R factor intheir study would be probably underestimated since in all cases themaximum half-hourly intensities are equal or lower than I30 values.They further reported that an underestimation of 22% existed if 60-

Table 1Basic information on the 62 stations.

No. Sta. No. Sta. Name Lat. (�N) Lon. (�E) Alt. (m) Annual Rainfalla (mm) Period Effective Yearsb Available Month

1 50557 Nenjiang 49.2 125.2 243 470.7 1961e2000 30 May - Sept.2 50963 Tonghe 46 128.7 110 570.2 1961e2000 38 May - Sept.3 53663 Wuzhai 38.9 111.8 1402 466.0 1971e2000 30 May - Sept.4 53754 Suide 37.5 110.2 928.5 445.2 1964e1996 29 May - Sept.5 53845 Yan’an 36.6 109.5 958.8 534.9 1961e2000 39 May - Sept.6 53975 Yangcheng 35.5 112.4 658.8 592.1 1971e2000 30 May - Sept.7 54416 Miyun 40.4 116.9 73.1 632.8 1961e2000 37 May - Sept.8 54511 Guangxiangtai 39.9 116.3 54.7 547.7 1961e2000 40 May - Sept.9 56294 Chengdu 30.7 104 506.1 883.8 1961e2000 39 Jan.- Dec.10 56571 Xichang 27.9 102.3 1590.9 992.5 1961e2000 40 Jan.- Dec.11 56739 Tengchong 25 98.5 1648.7 1472.5 1961e1999 36 Jan.- Dec.12 56778 Kunming 25 102.7 1896.8 983.9 1961e2000 33 Jan.- Dec.13 57259 Fangxian 32 110.8 427.1 821.6 1961e2000 31 Jan.- Dec.14 57405 Suining 30.5 105.6 279.5 937.6 1962e1999 33 Jan.- Dec.15 57504 Neijiang 29.6 105.1 352.4 1032.8 1961e2000 39 Jan.- Dec.16 58407 Huangshi 30.3 115.1 20.6 1412.3 1961e2000 32 Jan.- Dec.17 58847 Fuzhou 26.1 119.3 84 1364.5 1961e2000 39 Jan.- Dec.18 58911 Changting 25.9 116.4 311.2 1701.7 1961e2000 31 Jan.- Dec.19 50639 Zhalantun 48 122.4 306.5 493.7 2005e2016 11 May - Sept.20 51431 Yining 44 81.3 662.5 276.2 2005e2016 10 May - Sept.21 51463 Urumqi 43.8 87.7 935 264.7 2005e2016 9 May - Sept.22 51705 Wuqia 39.4 75.2 2175.7 183.8 2005e2016 11 May - Sept.23 52856 Gonghe 36.3 100.6 2835 317.8 2015e2016 2 May - Sept.24 52889 Lanzhou 36 103.5 1517.2 304.8 2007e2016 10 May - Sept.25 53463 Hohhot 40.5 111.4 1153.5 398.4 2005e2016 12 May - Sept.26 53644 Wushenqi 38.6 108.8 1307.2 341.5 2006e2016 11 May - Sept.27 53646 Yulin 38.3 109.8 1157 396.2 2005e2016 11 May - Sept.28 53848 Ganquan 36.3 109.3 1005.5 535.9 2008e2016 9 May - Sept.29 53923 Xifeng 35.7 107.6 1421 546.6 2005e2016 12 May - Sept.30 54161 Changchun 43.5 125.1 236.8 569.4 2005e2016 11 May - Sept.31 54337 Jinzhou 41.1 121.1 65.9 564.1 2005e2016 12 Apr.- Oct.32 54527 Tianjin 39.1 117 3.5 531.1 2005e2016 10 Apr.- Oct.33 54602 Baoding 38.5 115.3 16.8 516.0 2005e2016 11 Apr.- Oct.34 54823 Jinan 36.4 117 170.3 690.6 2005e2016 12 Apr.- Oct.35 54843 Weifang 36.5 119.1 22.2 599.5 2005e2016 12 Apr.- Oct.36 55578 Rikaze 29.3 88.9 3836 430.4 2006e2015 10 May - Sept.37 55591 Lasa 29.7 91.1 3648.9 433.7 2006e2016 9 May - Sept.38 56146 Ganze 31.4 100 3393.5 641.6 2005e2016 12 May - Sept.39 56312 Linzhi 29.4 94.2 2991.8 669.7 2006e2016 11 Apr.- Oct.40 56331 Zuogong 29.4 97.5 3780 438.5 2008e2016 9 May - Sept.41 56434 Chayu 28.4 97.3 2327.6 777.1 2007e2016 9 Mar.- Oct.42 57006 Tianshui 34.6 105.8 1141.6 511.7 2007e2016 10 May - Sept.43 57051 Sanmenxia 34.5 111.1 409.9 544.9 2005e2016 12 Mar.- Oct.44 57083 Zhengzhou 34.4 113.4 110.4 629.8 2005e2016 12 Mar.- Oct.45 57584 Yueyang 29.2 113.1 53 1315.1 2005e2016 8 Jan.- Dec.46 57687 Changsha 28.1 112.6 68 1404.9 2005e2016 12 Jan.- Dec.47 57816 Guiyang 26.4 106.4 1223.8 1106.2 2005e2016 12 Jan.- Dec.48 57926 Libo 25.3 107.5 428.7 1236.9 2007e2016 10 Jan.- Dec.49 57957 Guilin 25.2 110.2 164.4 1878.6 2005e2016 12 Jan.- Dec.50 58038 Shuyang 34.1 118.5 8.8 911.1 2007e2016 10 Jan.- Dec.51 58238 Nanjing 31.6 118.5 35.2 1065.4 2005e2016 12 Jan.- Dec.52 58321 Hefei 31.5 117.2 27 986.0 2005e2016 12 Apr.- Nov.53 58424 Anqing 30.3 117 19.8 1394.8 2005e2016 12 Apr.- Nov.54 58457 Hangzhou 30.1 120.1 41.7 1391.7 2005e2016 11 Jan.- Dec.55 58461 Qingpu 31.1 121.1 4 1084.6 2007e2016 10 Jan.- Dec.56 58606 Nanchang 28.4 115.6 46.9 1584.1 2005e2016 8 Jan.- Dec.57 58637 Shangrao 28.3 117.6 118.2 1810.9 2006e2016 11 Jan.- Dec.58 58646 Lishui 28.3 119.6 59.7 1392.9 2005e2016 12 Jan.- Dec.59 59046 Liuzhou 24.2 109.2 96.8 1430.5 2005e2016 12 Jan.- Dec.60 59082 Shaoguan 24.4 113.4 61 1569.9 2005e2016 12 Jan.- Dec.61 59493 Shenzhen 22.3 114 63 1885.7 2005e2016 11 Jan.- Dec.62 59758 Haikou 20 110.2 63.5 1679.5 2005e2016 10 Jan.- Dec.

a Annual rainfall was calculated based on the daily rainfall data observed by simple rain gauges.b Effective years were those with the relative error of the annual or seasonal rainfall amounts from the 1-min data is less than 15% compared with those from the daily

rainfall data of the same station.

T. Yue et al. / International Soil and Water Conservation Research 8 (2020) 373e382 375

min interval data are used using the same procedures by Panagoset al. (2015) based on rainfall observation at the 5-min intervalsfor a station used in Fiener and Auerswald (2009) and they spec-ulated that the degree of underestimation may vary within Europegiven the variation in climate.

In summary, the accumulation of fixed-interval data along withthe development of automatic rainfall measurement technologyhas provided high-resolution rainfall data and increased the accu-racy of computed rainfall erosivity. However, the effect of the timeinterval from 1-min to 1-hr on the computed R factor has not yetevaluated. Most of the previous researches linearly related eventEI30 values from the fixed-interval data to those from high-resolution data whereas some others considered R factorsinstead. No research has been reported for the estimation of the 1-in-10-year EI30 values from the fixed-interval data.

The objectives of this study are: (a) to systematically evaluatethe degree of underestimation between R factor calculated from 1-min data (assumed to be the ‘true’ or reference values) anddifferent intervals of rainfall data (5, 6, 10, 15, 20, 30 and 60 min);(b) to develop conversion factors of R factor and the 1-in-10-yearEI30 based on 5, 6, 10, 15, 20, 30 and 60-min data to those basedon the 1-min data. There were 62 stations with 1-min interval data

collected in China and were separated into calibration and valida-tion groups to systematically evaluate the bias and develop con-version factors.

2. Data and methods

2.1. Data

Rainfall data with 1-min temporal resolution from 62 meteo-rological stations in mainland China were collected from the Na-tional Meteorological Information Center of China MeteorologicalAdministration (Fig. 1, Table 1). The data were either digitized byrecording on the pluviograph of siphon rain gauges (No. 1e18), orrecorded by the automatic observation of tipping bucket raingauges (No. 19e62). Data from 26 stations in the south includeddata for the whole year whereas the other 36 stations in the northdid not include winter months for the equipment were stopped inwinter. The data quality was controlled by comparing the annualor seasonal total of rainfall amounts calculated from 1-min datawith those calculated from daily rainfall data observed withsimple rain gauges. The years with a relative error of less than 15%were considered as the effective year, which were used in this

Table 2Percentages of stations (%) whose estimated R factor under different error level.

Relative error less than 5 min 6 min 10 min 15 min 20 min 30 min 60 min

5% 41.9 30.6 21.0 3.2 0.0 0.0 0.010% 82.3 80.6 71.0 33.9 1.6 0.0 0.020% 96.8 93.5 93.5 90.3 80.6 38.7 0.030% 100.0 100.0 100.0 100.0 96.8 91.9 0.040% 100.0 100.0 100.0 100.0 100.0 100.0 3.250% 100.0 100.0 100.0 100.0 100.0 100.0 66.160% 100.0 100.0 100.0 100.0 100.0 100.0 98.4

Fig. 3. Probability distribution of the error of R factor based on 5, 15, 30 and 60 min data.


Fig. 4. Regression models of rainfall erosivities estimated from 5 to 60-min data. The dotted line represents the calibrated linear model. The circles are R factors estimated from 5 to60-min data without conversions. The dots are values multiplied by conversion factors.


Fig. 5. Regression models of 1-in-10-year EI30 estimated from the 5e60 min data. The dotted line represents the calibrated linear models. The circles are 1-in-10-year EI30 valuesestimated from the 5 to 60-min data without conversions. The dots are values multiplied by conversion factors.


study. The length of effective years for stations No. 1e18 rangedfrom 29 to 40, spanning the period of 1961 (1971)-2000. For sta-tions No. 19e62, this ranged from 2 to 12 years, spanning theperiod of 2005e2016. For time intervals with missing data ineffective years, which accounted for no more than 2.68% for eachof the 62 stations, zero rainfall was assumed because it was morelikely to be a dry interval. Numbers of storms varied from 10 to904 events for 62 stations.

2.2. Methods

2.2.1. Calculation of R factor and 1-in-10-year EI30To calculate rainfall erosivity with different time interval data,

the 1-min rainfall data were accumulated into 5, 6, 10, 15, 20, 30,and 60-min rainfall data. Firstly, rainfall periods were divided intoevents when the time of no precipitationwas over 6 h (Wischmeier& Smith, 1978). Rainfall events with amounts of no less than 12mmwere selected as erosive events following Xie et al. (2000) and wereused to calculate the EI30 index and R factor. Note that the criterionfor identification of an erosive event is slightly different from that inWischmeier and Smith (1978), in which an erosive event is definedto have total rainfall amount no less than 12.7 mm or with themaximum 15-min intensity no less than 25.4 mm h-1.

The total energy of a storm (E, MJ ha-1) was estimated using thefollowing equations (USDA-ARS, 2013):

E¼Xl

r¼1

ðer $ PrÞ (1)

er ¼0:29½1�0:72expð�0:082irÞ� (2)

where a rainfall stormwas divided into l periods with the constantintensity in the same period, er (MJ ha-1 mm-1) was the unit energy(energy per unit of rainfall) for the rth period, Pr (mm) was therainfall amount for the rth period and ir (mm h-1) was the intensityfor the rth period. Themaximum contiguous 30-min peak intensity,I30 (mm h-1), can be calculated as:

I30 ¼P300:5

(3)

where P30 (mm) was the maximum contiguous 30-min rainfallamount. For 1, 5, 6, 10, 15, and 30- min rainfall data, P30 was themaximum total amounts of 30, 6, 5, 3, 2, and 1 contiguous records.For 20-min rainfall data, records were linearly interpolated into 10-min data first and P30 was the maximum total amounts of threecontiguous 10-min intervals. For 60-min rainfall data, the P30 washalf of the maximum records. The storm rainfall erosivity EI30 can

be obtained by the product of E and I30:

EI30 ¼ E$I30 (4)

The R factor was defined as the average annual rainfall erosivity,which can be calculated by the sum of the EI30 for all erosive stormsin a year:

R¼ 1N

XN

i¼1

Xm

j¼1

ðEI30Þij (5)

where N was the length of the effective years, mwas the number oferosive storms in the ith year.

Since the lengths of the data from stations with tipping bucketrain gauges were no more than 12 years and annual maximum EI30may not enough to accurately fit an extreme value distribution, the90th percentile of the annual maximum EI30, denote as (EI30)P90,estimated from 1 to 60-min rainfall data of 54 stations (with morethan 10 effective years) were used to develop linear functions forconversion factors of the 1-in-10-year EI30 values from 5 to 60-minrainfall data to those from 1-min data.

2.2.2. Calculation of conversion factorsThe underestimation of rainfall erosivity from t-min resolution

data was characterized by the ratio of rainfall erosivity estimatedfrom t-min data to that calculated from 1-min data, which wasconsidered as the ‘true’ value in this study.

Since both EI30 and R factor estimated from the fixed-intervalrainfall data and the values from using high-resolution data havea good linear relationship (Williams & Sheridan, 1991; Yin et al.,2007). Linear models of R factors estimated from different timeinterval data were developed to obtain the conversion factors forrelating the values from these data to those from 1-min data:

R1min ¼CFt$Rtmin (6)

where R1min and Rtmin were the R factor values calculated by 1-minand t-min rainfall data from 62 stations, respectively, and CFt wasthe calibrated conversion factor for the rainfall erosivity from t-mindata to 1-min data.

To assess the accuracy of the conversion factors in equation (6),data from 62 stations were divided into two groups. Thirty-three ofthem, marked as dots in Fig. 1, were classified as Group A, whereasthe remaining 29 stations marked as circles in Fig. 1 were classifiedas Group B. At first stations in Group A were used to develop con-version factors and those in Group B were used to evaluate theperformance of conversion factors. This process was then con-ducted reversely, which meant data from 29 stations in Group Bwere used to calculate conversion factors, and those from 33 sta-tions in Group Awere used to evaluate the performance. Symmetricmean absolute percentage error (sMAPE) and Nash-Sutcliffe modelefficiency coefficient (NSE) were used for the assessment:

sMAPE¼1n

Xn

i¼1

��Fi � Ai

ðFi þ AiÞ=2�� 100% (7)

NSE ¼ 1�

Pni¼1

ðFi � AiÞ2

Pni¼1

ðAi � AÞ2(8)

where nwas the number of stations, Fi was the R factors estimatedfrom 5 to 60-min rainfall data for the ith station (before or afterapplying conversion factors), Ai was the R factor calculated from 1-

Table 3Conversion factors for estimating annual rainfall erosivity.

5 min 6 min 10 min 15 min 20 min 30 min 60 min

CF 1.051 1.058 1.083 1.121 1.192 1.253 1.871

sMAPEa 7.2% 8.0% 10.4% 13.7% 20.6% 24.7% 66.2%sMAPEca 4.1% 4.3% 4.4% 4.4% 4.8% 4.3% 8.8%NSEa 0.999 0.999 0.999 0.999 0.998 0.998 0.990

sMAPEb 6.7% 7.3% 9.5% 12.6% 19.1% 24.5% 63.0%sMAPEcb 3.7% 3.8% 3.7% 3.8% 3.8% 4.1% 4.6%NSEb 1.000 1.000 1.000 1.000 0.999 0.999 0.992

sMAPE: sMAPE of rainfall erosivity estimated from 5-60 min data.sMAPEc: sMAPE of converted rainfall erosivity estimated from 5-60 min data.

a Averaged validation values of Group B stations when using conversion factorscalibrated by Group A stations.

b Averaged validation values of Group A stations when using conversion factorscalibrated by Group B stations.


min rainfall data for the ith station.

3. Results

3.1. Underestimation of R factor from 5 to 60-min rainfall data

The R factors estimated from the 5 to 60-min rainfall data wereall underestimated compared with values calculated from the 1-min data. The longer the time interval of the data, the larger theunderestimated of R factors. Fig. 2 shows the relationship betweentime interval and the reduction in the R factor, as the time intervalincreases, the ratio of the R factor over that using 1-min datadecreases.

The averaged underestimations were -6.6, -7.2, -9.3, -12.2, -17.9,-21.8 and -48.7% for R factors estimated from 5, 6, 10, 15, 20, 30 and60-min rainfall data, respectively. Table 2 showed that 51, 50 and 44stations out of 62 stations with relative errors less than 10% for 5, 6and 10-min interval data, respectively. However, when it comes tothe data with the temporal resolutions more than 15 min, the biasincreased rapidly (Fig. 2). For the 15-min data, the estimated R fac-tors for 66.1% of stations were underestimated by > 10% and for the30-min data, 61.3% of stations underestimated by > 20% (Table 2).Fig. 3 showed the distribution of the relative error for R factor basedon 5,15, 30 and 60-min data as examples. Therewere 4 stationswithrelative errors more than 15% when 5-min data were used. Thesestations were located in the western part of China with dry climate

and low rainfall erovisity. The relative error tends to be large whenrainfall erosivity is low.

3.2. Conversion factors for the R factor

The scatterplots of the linear models based on equation (6) areshown in Fig. 4 and it indicated a very good regressionwas obtainedwith the coefficients of determination ranging from 0.994 to 0.999.As the time interval of the data increases from 5 to 60 min, theconversion factor rises from 1.051 to 1.871 (Fig. 4). The results alsoshowed that for stations with R factor less than 5,000 MJ mm ha-1

h-1 a-1, the values obtained from the data of the time interval within10 min were not apparently underestimated, but for those withhigher rainfall erosivity, the R factors estimated from the data of theinterval within 10 min showed an obvious underestimation, whichprobably resulted in significant differences when paired-sample t-test was conducted to compare the R factors based on 5 to 60-mindata with those 1-min data.

Conversion factors and validation results can be found in Table 3.The Nash-Sutcliffe model efficiency coefficients (NSE) of themodelsare all above 0.99, suggesting that conversion factors calibrated canbe quite effective for converting R factors estimated from 5 to 60-min data to those based on 1-min data. After the conversion, thesMAPE decreased by about 3% for the 5-min data, 4% for the 6-mindata, 6% for the 10-min data, 9% for the 15-min data, 15% for the 20-min data, 20% for the 30-min data, 60% for the 60-min data.

3.3. Conversion factors for the 1-in-10-year EI30

The efficiency of models for the 1-in-10-year EI30 is slightlylower compared with those for the R factor, with the coefficients ofdetermination ranging from 0.958 to 0.999. Conversion factors forthe 1-in-10-year-frequency EI30 were smaller than those for the Rfactor (Fig. 5). As the time interval increases from 5 to 60 min, theconversion factor rises from 1.034 to 1.489.

Conversion factors for the 1-in-10-year-frequency EI30 and thevalidation results were listed in Table 4. The Nash-Sutcliffe modelefficiency coefficients (NSE) of the models are all above 0.93, andthose from 5 to 30-min rainfall data are all above 0.98. After theconversion, the MAPE decreased by about 1% for the 5-min data, 2%for the 6-min data, 3% for the 10-min data, 5% for the 15-min data,8% for the 20-min data, 9% for the 30-min data, 37% for the 60-mindata. It is noticed that there was a difference of sMAPEc (20.8% and10.1%, respectively) after applying conversion factors for the 60-

Table 4Conversion factors for the 1-in-10-year EI30.

5 min 6 min 10 min 15 min 20 min 30 min 60 min

CF 1.034 1.044 1.053 1.090 1.101 1.180 1.489

sMAPEa 3.6% 4.6% 5.5% 9.5% 15.4% 18.3% 55.5%sMAPEca 2.5% 3.3% 2.7% 5.1% 7.1% 9.6% 20.8%NSEa 0.999 0.998 0.998 0.990 0.995 0.981 0.938

sMAPEb 3.9% 4.4% 5.9% 9.0% 14.6% 17.0% 49.4%sMAPEcb 2.6% 2.8% 3.4% 4.3% 6.3% 8.8% 10.1%NSEb 0.999 0.998 0.998 0.995 0.988 0.991 0.953

sMAPE: sMAPE of 1-in-10-year EI30 values estimated from 5-60 min data.sMAPEc: sMAPE of converted 1-in-10-year EI30 values estimated from 5-60 mindata.

a Averaged validation values of Group B stations when using conversion factorscalibrated by Group A stations.

b Averaged validation values of Group A stations when using conversion factorscalibrated by Group B stations.

Table 5Conversion factors (CFs) for converting EI30 (R factor) values from 5 to 60-min rainfall data to values estimated from high-resolution data.

Study Area 5 min 6 min 10 min 15 min 30 min 60 min Resolution of datacompared

Index REF.

Western Oregon,US

e e e e e 1.193e1.378 15 min EI30 Istok et al. (1986)

SoutheasternGeorgia, US

1.0416 1.0405 1.09 1.1209 1.3127 1.837 Breakpoint EI30 Williams andSheridan (1991)

US e e e e e 1.08e3.16 15 min EI30 Renard (1997)Sicily, Italy e e e e e 1.33 15 min EI30 Agnese et al. (2006)China 1.014 (1.010

e1.020)e 1.044 (1.023

e1.069)1.078 (1.039e1.114)

1.161 (1.094e1.257)

1.73 (1.568e1.814)

Breakpoint EI30 Yin et al. (2007)

Europe 0.7984 (0.7514e0.8568)

e 0.8205 (0.7986e0.8951)

0.8716 (0.8072e0.9233)

1 1.5597 (1.2974e1.6995)

30 min R factor Panagos et al., 2015;2016

Southern Italy 0.92 e 0.97 e 1.12 1.53 15 min EI30 Porto (2016)China 1.051 1.058 1.083 1.121 1.253 1.871 1 min R factor This studyChina 1.034 1.044 1.053 1.090 1.180 1.489 1 min 1-in-10-

year EI30This study


min data from Group A and B. There were no obvious differencesfound for other intervals from two groups.

4. Discussion

Experiments were conducted to detect if conversion factorsvaried by climate types by dividing stations into different regionsand it was found that conversion factors for EI30 in the western partof China were significantly larger than those in the eastern part.However, no significant differences among the regression equa-tions for the R factor and 1-in-10-year EI30 for eastern and westernregions were found. Therefore, all stations were pooled together inthis study when conversion factors were developed for the R factorand 1-in-10-year EI30. Table 5 lists conversion factors for differentparts of the world. As these studies were not undertaken using thesame criteria for defining erosive events, it is not appropriate toconduct a rigorous comparison among these conversion factors.Further complication occurs because these conversion factors wererelated to different erosivity indicators such as EI30, R factor and 1-in-10-year EI30, and they were derived using rainfall of differenttypes, and at different resolutions.

Most previous studies on conversion factors of rainfall erosivityrelated erosive event EI30 index from the coarser fixed-interval datato those from the breakpoint data or finer fixed-interval data.Panagos et al. (2015; 2016) regressed R factor from the coarserresolution of data to the finer data (Table 5). The results of thisstudy showed that conversion factors for different index (R factorand the 1-in-10-year EI30) were quite different. The conversionfactors for the R factor, ranging from 1.051 to 1.871, were higherthan those for the 1-in-10-year EI30, ranging from 1.034 to 1.489. Itsuggested that the corresponding conversion factors should beapplied in the calculation of different indexes of rainfall erosivity(event EI30, R factor, or 1-in-10-year EI30).

The resolution of data used as the compared baseline can alsolead to differences in conversion factors. Table 5 showed that twostudies using the breakpoint data as baselines (Williams& Sheridan,1991; Yin et al., 2007) obtained similar conversion factors for the Rfactor in this study. Four studies used the 15-min data (Agnese et al.,2006; Istok et al., 1986; Porto, 2016; Renard, 1997) and two studiesused 30-min data (Panagos et al., 2015; 2016) as baselines in Table 5.Results in this study showed that the underestimation ranging from4.5% to 28.2% with an average of 12.2% for the 15-min data and from13.4% to 37.4% with an average of 21.8% for the 30-min data, whichindicated that it was better to obtain the breakpoint data or the datawith intervals less than 5min as the baseline compared. If those datawere not available, a further conversion may be required as done byRenard (1997), in which a constant conversion factor of 1.0667(Weiss, 1964) between EI30 from the 15-min data and the breakpointdatawere adopted. Nearing et al. (2017) reminded us of themeaningof the underestimation of the rainfall erosivity. The erosivity anderodibility factors are two of the six factors in USLEwhich have units.The remaining factors including the slope steepness, slope length,cropping management, and control practice are all dimensionless.Erodibility factor was defined as the ratio of actual soil loss dividedby the erosivity in a unit plot, which indicated that if rainfall erosivitywas biased, soil erodibility would need to be re-determined oradjusted to make estimates of soil loss that fit the measured data.

Breakpoint or hyetograph data were used to compute ‘true’values of rainfall erosivity as defined by Wischmeier and Smith(1978) before. Along with the development of automatic tippingbucket observation, fixed-interval data are growing more commonand 1-min data have begun to replace the breakpoint data gradu-ally for easier acquisition, storage, and processing. 1-min intervaldata should have higher or at least the same resolution comparingwith the breakpoint data.

Noted that rainfall erosivity values estimated from 1 to 60-mintime interval data could be converted to each other in addition toconverting values from intervals no less than 5 min to values from1-min data. For example, the R factor estimated from 30-min datacan be converted to that from 5-min data, by applying a conversionfactor of 1.192 (1.253 divided by 1.051, Table 5).

5. Conclusions

This studyanalyzed1-minrainfall data from62stations inChina toevaluate the degree of underestimation between R factors calculatedfrom the 5, 6,10,15, 20, 30 and 60-mindata and those from the 1-mindata, and to develop conversion factors for the R factor and 1-in-10-year EI30 based on coarse resolution data to those based on 1-mindata. The following conclusions can be drawn as follows:

(1) R factors estimated from 5 to 60-min data were under-estimated by 6.6 to 48.7%. For 5, 6 and 10-min data, theunderestimation for more than three-quarters of stationswas less than 10%, whereas, for data with time intervals ofmore than 15 min, the underestimation increased rapidly.For about two-thirds of the stations erosivity was under-estimated by more than 10% using the 15-min data and for61.3% of the stations, the underestimation was over 20% us-ing 30-min data.

(2) Conversion factors for the R factor estimated from 5, 6, 10, 15,20, 30 and 60-min were 1.051, 1.058, 1.083, 1.121, 1.192, 1.253and 1.871, respectively. Validation results showed that afterapplying conversion factors, symmetric mean absolute per-centage error decreased by 3.0%, 3.6%, 5.9%, 9.0%,15.5%, 20.3%and 58.0% for the 5 to 60-min data.

(3) Conversion factors for the 1-in-10-year EI30 were 1.034,1.044, 1.053, 1.090, 1.101, 1.180 and 1.489, respectively, whichwere generally smaller than those for the R factor. Symmetricmean absolute percentage error decreased by 1.3%,1.6%, 2.9%,4.9%, 8.2%, 8.4% and 36.2% for the 5 to 60-min data.


The authors declared that they have no conflicts of interest tothis work.

Acknowledgements

This work was supported by the Second Tibetan Plateau Scien-tific Expedition and Research Program (STEP) (No. 2019QZKK0306),the National Key R&D Program (No. 2018YFC0507006) and Na-tional Natural Science Foundation of China (No. 41877068). Theauthors declared that they have no conflicts of interest in this work.

References

Agnese, C., Bagarello, V., Corrao, C., D’Agostino, L., & D’Asaro, F. (2006). Influence ofthe rainfall measurement interval on the erosivity determinations in theMediterranean area. Journal of Hydrology, 329(1e2), 39e48.

Angulo-Martínez, M., & Beguería, S. (2009). Estimating rainfall erosivity from dailyprecipitation records: A comparison among methods using data from the Ebrobasin (NE Spain). Journal of Hydrology, 379(1e2), 111e121.

Auerswald, K., Fiener, P., Gomez, J. A., Govers, G., Quinton, J. N., & Strauss, P. (2015).Comment on “Rainfall erosivity in Europe” by Panagos et al.(Sci. Total Environ.,511, 801e814, 2015). The Science of the Total Environment, 532, 849e852.

Diodato, N., Borrelli, P., Fiener, P., Bellocchi, G., & Romano, N. (2017). Discoveringhistorical rainfall erosivity with a parsimonious approach: A case study inWestern Germany. Journal of Hydrology, 544, 1e9.

Fiener, P., & Auerswald, K. (2009). Spatial variability of rainfall on a sub-kilometrescale. Earth Surface Processes and Landforms, 34(6), 848e859.

Fiener, P., Neuhaus, P., & Botschek, J. (2013). Long-term trends in rainfallerosivityeanalysis of high resolutionprecipitation time series (1937e2007) fromWestern Germany. Agricultural and Forest Meteorology, 171e172(8), 115e123.


Hanel, M., M�aca, P., Ba�sta, P., Vlnas, R., & Pech, P. (2016). The rainfall erosivity factorin the Czech Republic and its uncertainty. Hydrology and Earth System SciencesDiscussions, 20, 1e24.

Hollinger, S. E., Angel, J. R., & Palecki, M. A. (2002). Spatial distribution, variation, andtrends in storm precipitation characteristics associated with soil erosion in theUnited States. ISWS Contract Report CR-2002-08.

Istok, J. D., McCool, D. K., King, L. G., & Boersma, L. (1986). Effect of rainfall mea-surement interval on EI calculation. Transactions of the ASAE, 29(3), 730e734.

Klik, A., Haas, K., Dvorackova, A., & Fuller, I. C. (2015). Spatial and temporal distri-bution of rainfall erosivity in New Zealand. Soil Research, 40(6), 887e901.

Liu, B. Y., Zhang, K. L., & Xie, Y. (2002). An empirical soil loss equation. Procee-dingseProcess of Soil Erosion and Its Environment Effect, 2, 21e25.

Lo, A., El-Swaify, S. A., Dangler, E. W., & Shinshiro, L. (1985). Effectiveness of EI30 asan erosivity index in Hawaii.

Mannaerts, C. M., & Gabriels, D. (2000). Rainfall erosivity in Cape Verde. Soil andTillage Research, 55(3), 207e212.

Martins, S. G., Avanzi, J. C., Silva, M. L. N., Curi, N., Norton, L. D., & Fonseca, S. (2010).Rainfall erosivity and rainfall return period in the experimental watershed ofAracruz, in the coastal plain of Espirito Santo, Brazil. Revista Brasileira de Cienciado Solo, 34(3), 999e1004.

Mcfarlane, D., Westcott, T., & Davies, J. (1986). Rainfall erosivity in western Australia.Meusburger, K., Steel, A., Panagos, P., Montanarella, L., & Alewell, C. (2012). Spatial

and temporal variability of rainfall erosivity factor for Switzerland. Hydrologyand Earth System Sciences, 16, 167e177.

Nearing, M. A., Yin, S., Borrelli, P., & Polyakov, V. O. (2017). Rainfall erosivity: Anhistorical review. Catena, 157, 357e362.

Neuhaus, P., Fiener, P., Botschek, J., & Schneider, K. (2010). Longterm and seasonaltrends in rainfall erosivity - analysis of a set of high-resolution precipitationtime-series (1937-2007) from Western Germany. In Egu general assembly.

Panagos, P., Ballabio, C., Borrelli, P., Meusburger, K., Klik, A., Rousseva, S., …

Alewell, C. (2015). Rainfall erosivity in Europe. The Science of the Total Envi-ronment, 511, 801e814.

Panagos, P., Borrelli, P., Spinoni, J., Ballabio, C., Meusburger, K., Beguería, S.,Rousseva, S., et al. (2016). Monthly rainfall erosivity: Conversion factors fordifferent time resolutions and regional assessments. Water, 8(4), 119.

Porto, P. (2016). Exploring the effect of different time resolutions to calculate therainfall erosivity factor R in Calabria, southern Italy. Hydrological Processes,30(10), 1551e1562.

Renard, K. G. (1997). Predicting soil erosion by water: A guide to conservation planningwith the revised universal soil loss equation (RUSLE). United States GovernmentPrinting.

Renard, K. G., & Freimund, J. R. (1994). Using monthly precipitation data to estimatethe R-factor in the revised USLE.

Richardson, C. W., Foster, G. R., & Wright, D. A. (1983). Estimation of erosion indexfrom daily rainfall amount. Transactions of the ASAE, 26(1), 153e156.

Santosa, P. B., Mitani, Y., & Ikemi, H. (2010). Estimation of RUSLE EI30 based on 10min interval rainfall data and GIS-based development of rainfall erosivity mapsfor Hitotsuse basin in Kyushu Japan. In International conference on

geoinformatics (pp. 1e6).Schmidt, S., Alewell, C., Panagos, P., & Meusburger, K. (2016). Regionalization of

monthly rainfall erosivity patternsin Switzerland. Hydrology and Earth SystemSciences, 20(10), 4359e4373.

Shamshad, A., Azhari, M. N., Isa, M. H. al, Hussin, W. M. A. W., & Parida, B. P. (2008).Development of an appropriate procedure for estimation of RUSLE EI30 indexand preparation of erosivity maps for Pulau Penang in Peninsular Malaysia.Catena, 72(3), 423e432.

USDA-ARS. (2013). Science documentation: Revised universal soil loss equation version2 (RUSLE2). Washington, D.C.: USDA-Agricultural Research Service.

Verstraeten, G., Poesen, J., Demar�ee, G., & Salles, C. (2006). Long-term (105 years)variability in rain erosivity as derived from 10-min rainfall depth data for Ukkel(Brussels, Belgium): Implications for assessing soil erosion rates. Journal ofGeophysical Research Atmospheres, 111(D22).

Weiss, L. L. (1964). Ratio of true to fixed-interval maximum rainfall. Journal of theHydraulics Division, 90(1), 77e82.

Williams, R. G., & Sheridan, J. M. (1991). Effect of rainfall measurement time anddepth resolution on EI calculation. Transactions of the ASAE, 34(2), 402e406.

Wischmeier, W. H. (1959). A rainfall erosion index for a universal soil-loss equation1. Proceedings - Soil Science Society of America, 23(3), 246e249.

Wischmeier, W. H. (1976). Use and misuse of the universal soil loss equation. Journalof Soil and Water Conservation, 31(1), 5e9.

Wischmeier, W. H., & Smith, D. D. (1958). Rainfall energy and its relationship to soilloss. Transactions - American Geophysical Union, 39(2), 285e291.

Wischmeier, W. H., & Smith, D. D. (1965). Predicting rainfall-erosion losses fromcropland east of the Rocky Mountains: Guide for selection of practices for soil andwater conservation. US Department of Agriculture.

Wischmeier, W. H., & Smith, D. D. (1978). Predicting rainfall erosion losses: A guide toconservation planning. Department of Agriculture, Science and EducationAdministration.

Xie, Y., Liu, B. Y., & Zhang, W. B. (2000). Study on standard of erosive rainfall. Journalof Soil and Water Conservation, 14(4), 6e11.

Xie, Y., Yin, S. Q., Liu, B. Y., Nearing, M. A., & Zhao, Y. (2016). Models for estimatingdaily rainfall erosivity in China. Journal of Hydrology, 535, 547e558.

Yin, S.-Q., Xie, Y., Liu, B., & Nearing, M. A. (2015). Rainfall erosivity estimation basedon rainfall data collected over a range of temporal resolutions. Hydrology andEarth System Sciences Discussions, 12(5).

Yin, S., Xie, Y., Nearing, M. A., & Wang, C. (2007). Estimation of rainfall erosivityusing 5- to 60-minute fixed-interval rainfall data from China. Catena, 70(3),306e312.

Yu, B. (1998). Rainfall erosivity and its estimation for Australia’s tropics. SoilResearch, 36(1), 143e166.

Yu, B., Murphy, B., Vaze, J., & Rawson, A. (2010). A first-order assessment of climatechange effects on rainfall erosivity and soil erosion in New South Wales,Australia. Egu General Assembly, 12, 3859.

Yu, B., & Rosewell, C. J. (1996). Rainfall erosivity estimation using daily rainfallamounts for South Australia. Soil Research, 34(5), 721e733.



The use of remote sensing to detect the consequences of erosion ingypsiferous soils

Maria Jose Marques a, *, Ana Alvarez a, Pilar Carral a, Blanca Sastre b, Ram�on Bienes b

a Geology and Geochemistry Department, Autonomous University of Madrid, Calle Francisco Tom�as y Valiente, 7, 28040, Madrid, Spainb Department of Applied Research and Agrarian Extension, IMIDRA, Finca El Encín. Carretera A-2, Km 38.8. Alcal�a de Henares, 28800, Madrid, Spain


Article history:Received 13 February 2020Received in revised form25 September 2020Accepted 8 October 2020Available online 16 October 2020

Keywords:TillageErosionNDVIGypsumRemote sensingSoil color

a b s t r a c t

Tillage practices on sloping ground often result in unsustainable soil losses impairing soil functions suchas crop productivity, water and nutrients storage, and soil organic carbon (SOC) sequestration. A slopingolive grove (10%) was planted in shallow gypsiferous soils in 2004. It was managed by minimum tillage;the most frequent management practice in central Spain. The consequences of erosion were studied insoil samples (at 0e10, 10e20, and 20e30 cm depths) by analyzing SOC, available water and gypsumcontent, and by detecting spectral signatures using an ASD FieldSpecPro® VIS/NIR-spectroradiometer.The Brightness index (BI), Shape index (FI), and Normalized Difference Vegetation Index (NDVI) werederived from the ASD spectral signatures and from remote sensing (Sentinel-2 image) data. The devel-opment of these young olive trees was estimated from the measured diameter of the trunks (17 ± 18 cmdiameter). In 20e30 cm of the soil, the carbon stock (38 ± 18 Mg ha�1) as well as the available watercontent (12 ± 6%) was scarce, affecting the productivity of the olive grove. The above-mentioned indicesobtained from the laboratory samples and the pixels of the Sentinel-2 image were significantly (p < 0.01)correlated, with a correlation coefficient of around 0.4. The BI was related to the gypsum content and theslope of the plot. The FI was related to the carbon and water contents. The NDVI derived from the satelliteimage identified the influence of soil degradation on the trees and the carbon content. The spatial-temporal changes of the indices might help in tracking soil changes over time.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction

Soil degradation is considered one of the most importantenvironmental topics in recent years (Rojas et al., 2016) and hasbeen studied under very different approaches and methodologies(Borrelli et al., 2017). Without a doubt, remote sensing is one of themost important tools to monitor soil changes, being increasinglyused as image resolution has been improved. Since the initialresearch on the relationships between soil properties and spectralreflectance (Condit, 1970), many different studies have gathereddifferent physical-chemical characteristics to build spectral li-braries (Ben-Dor et al., 2008; Dematte et al., 2004; Shepherd &Walsh, 2002), in order to establish links between spectral

signatures and different soil parameters such as texture (Castaldiet al., 2016), moisture (Whiting et al., 2004) or contents ofdifferent elements or compounds, like mineral composition(Sabins, 1999), salt concentration (Metternicht & Zinck, 2003), ororganic carbon (Gomez et al., 2008). The VIS, NIR, and SWIRhyperspectral airborne data has been even used for the study of ageand soil formation (Ben-Dor et al., 2006; Galv~ao et al., 2008). Animportant aspect of remote sensing or earth observation is itsusefulness for land monitoring and environmental assessment(Dubovyk, 2017), as this information can be used to assist in thedecision-making process to achieve global environmental protec-tion and sustainable development.

This paper is addressing one of the problems of degradation ofagricultural soils, erosion, and its repercussions on crop produc-tivity, whose consequences are of particular importance (Panagoset al., 2015). The research involving earth observation systemsand soil erosion is based on aerial photos and remote sensing sat-ellite imagery. Different assessments are grounded on changes invegetation cover (Dwivedi & Ramana, 2003; Symeonakis & Drake,

* Corresponding author.E-mail addresses: [email protected] (M.J. Marques), anamaria.

[email protected] (A. Alvarez), [email protected] (P. Carral), [email protected] (B. Sastre), [email protected] (R. Bienes).






2004); changes in topography (Lee & Liu, 2001) and mass changesrelated to gully processes (Bennett & Wells, 2019; Bocco &Valenzuela, 1993). The different characteristics between erodedand depositional areas (Beaulieu & Gaonac’h, 2002) can be alsoused as indicators of erosion processes, for example, the relativeabundance of different soil particle size, considered as a proxy ofsoil loss, has produced highly accurate results at the local scale (Hillet al., 1995). A wide review of factors involving erosion and remotesensing, mapping techniques and validation methods can be foundin Vrieling (2006).

When soils are shallow and show well-defined horizons withdifferent colors, erosion is clearly indicated by color changes in theground surface. Color information can be retrieved from the VISregion of the electromagnetic spectrum, and in agricultural soils, itis usually related to soil organic carbon (SOC), textural properties(Castaldi et al., 2016; McCarty et al., 2002) and presence of salts, e.g.outcrops of subsurface horizons enriched with carbonates or gyp-sum (Escadafal, 1994). Color changes indicate a dual process; su-perficial horizons are lost at the same time that the layersunderneath, with contrasting characteristics, are graduallyemerging. As already mentioned, different physical-chemicalcharacteristics can be detected by changes in reflectance andspectral signatures of soils. Remote observation of soils does notallow to calculate erosion rates, but it does allow to estimatedifferent states of degradation of soils when color changes indicatethat C horizons of soils approach the surface. This process usuallyentails changes in SOC, soil bulk density or nutrients that harm soilfunctions, especially primary productivity, edaphic biodiversity,and hydraulic properties. This causes problems in agriculturalproduction and can lead to gradual land abandonment. Eventually,when these soils are definitely abandoned, spontaneous vegetationwill also find difficulties to grow and land experiences a shift toirreversible degradation (D’Odorico et al., 2013).

In this study, differences in soil reflectance observed from sat-ellite imagery were used to estimate soil degradation and loss ofproductivity in a sloping area used for olive production under semi-arid environmental conditions. Traditional olive orchards havebeen installed in sloping and poor gypsiferous soils and conse-quently have experienced high rates of soil erosion (Fleskens &Stroosnijder, 2007) being in conflict with environmental sustain-ability and hampering its conservation as a traditional agriculturalsystem (de Graaff et al., 2008). More sustainable land use practiceshave been proposed (G�omez et al., 2003; Hernandez et al., 2005;Sastre et al., 2017), but farmers are reluctant to change theirmanagement practices (Marques et al., 2015). There is an urgentneed to involve farmers in this effort to protect soil and maintain

sustainable production. In order to do so, the consequences of soilerosion on agronomic productivity can be more effective than fig-ures focused on soil erosion rates.

This paper was motivated by the need to highlight the influenceof erosion on elements that are important for land users like agri-cultural productivity, water availability, and the ability to adapt toclimate change in semi-arid areas. Soil loss studies in the region ofcentral Spain (Marques et al., 2007; Bienes et al., 2009; Sastre, 2017)observed that erosion resulted in the exposure of the deeper soilhorizon and arguably, the differences in soil properties could bedetected by remote sensing. Based on plot-scale measurements, thepresent study intended to demonstrate the relationships betweensoil reflectance and i) soil characteristics: i.e., soil organic carboncontent, water availability, and gypsum content, ii) plot slope, andiii) olive tree development.

2. Material and methods

2.1. Study area

The study area is located in a semi-arid Mediterranean envi-ronment in the Center of the Iberian Peninsula. In the last 20 years,the accumulated mean annual rainfall yields 380 mm (SourceMeteorological Climate Agency). The area shows a rolling landscapewhere agricultural use is predominant, especially for growing olivegroves and vineyards, and for grazing cattle. The tops of hills arecolonized by species of scrubland well adapted to drought andgypsic soils such as the Poacea Stipa tenacissima (Fig. 1a).

The agricultural plot under study (Datum ETRS89, latitude:40�402500N; longitude: 3�3102000W) has gypsiferous soils. Accordingto the FAO soil classification (WRB, 2015), two main soils can bedistinguished: Haplic Gypsisols usually found in flat areas, andGypsiric Regosols being shallow and located in sloping areas. Fig. 1(b) shows these regosols with Cy1 horizon found by 35e40 cmdepth having high gypsum contents. These horizons are very un-favorable for vegetation establishment. In previous research, soilstudy analyses demonstrated high susceptibility to erosion due toits silty loam texture (Sastre et al., 2017). This plot was previouslyused to grow vines, but in 2004, vines were unearthed and olivetreeswere planted (7� 7m). The plot has beenmonitored since thisimplementation and several episodes of intense erosion have beenrecorded (Fig. 1 a). The plot covers 3.7 ha, with 10e12% of slope. Thealtitude above sea level ranges between 540 and 560m. Soil is tilledtwo or three times a year using chisel plow up to 20 cm depth.

2.2. Soil sampling and soil variables

In the plot under study, 30 georeferenced samples were selectedaccording to a randomized systematic design (Fig. 1 a, c). Aroundthese soil samples, trees vigor was estimated by trunk diameter.The diameter of olive treeswas estimated at each sampling point byaveraging the tree’s diameters of the four olive trees immediatelyaround each point; the height of measurement was 1 ± 0.1 m,variations were due to branches impedingmeasurements. The olivetrees were planted in 2004. Dead or defective trees were replacedin the next two years, therefore, at the time of sampling olive treeswere between 12 and 15 years old.

Soil samples were taken in late summer of 2017, at threedifferent depths (0e10; 10 to 20 and 20e30 cm). Soil samples wereair-dried and sieved (2 mm). Two groups of subsamples wereestablished, one group to measure physical-chemical variables andanother one to obtain the spectral signatures using an analyticalspectral device (ASD) spectroradiometer described below.

The first group of subsamples was used to measure availablewater capacity (AWC) using the pressure plate method (Richards,

List of abbreviations

ASD analytical spectral device (ASD)spectroradiometer

AWC available water capacityBI brightness indexFC field capacityFI shape indexNDVI normalized difference vegetation indexNIR near infra-redPWP permanent wilting pointS-2 Sentinel 2SOC soil organic carbonSqr Gyp: square root of gypsum contentVIS visible

M.J. Marques, A. Alvarez, P. Carral et al. International Soil and Water Conservation Research 8 (2020) 383e392

384

1941), and gypsum content was estimated using X-ray diffractionanalysis (Klug& Alexander, 1974). These subsamples were also usedto measure soil organic carbon stock (Equation (1)).

Soil:Organic:C:Stock:�Mg:ha�1

�:¼ :SOC:ð%Þ:

� :Bulk:density:�Mg m�3

�:� :depth:ðmÞ:x:100 (1)

The SOC (%) was estimated by the Loss On Ignition Method(Schulte & Hopkins, 1996); soil bulk density was estimated as themean density of 5 macroaggregates (1e3 cm diameter) obtained bythe mercury-displacement method (Franklin, 1977). This methodwas accurate to measure the bulk density of irregular rocks usingnominal diameters ranging from 2 to 3 cm (Franzini & Lezzerini,2003). Large macroaggregates (0.8e1.9 cm) have been found tostore and protect labile SOC and are good indicators of potential Cresponses to land use management changes (Tivet et al., 2013). Thedepth considered was 30 cm, following recommendations of theIntergovernmental Panel on Climate Change (IPCC, 2006). Therewere no gravel or stones in the study area.

The sieved dry samples of the second subset were placed on5 cm diameter by 1 cm depth soil dishes to measure soil spectralsignatures across 350e1100 nm using an ASD FieldSpecPro® VIS/NIR spectroradiometer (Boulder, CO, USA). The procedure wascarried out in a dark room using the ASD contact probe (halogenbulb 2900K color temperature). Each scan was the 10 internal scanaverage. Noisy portions from 350 to 450 nm and 950e1100 nmwere removed prior to statistical analysis. Awhite Spectralon™wasused every 9 measurements to optimize the spectroradiometermeasurements. Hereinafter the indices, explained in section 2.4,calculated with laboratory ASD measurements will be qualified as“ASD”.

Soil reflectance was obtained by two methods: laboratoryspectroscopic measurements of soil samples (ASD) and spectraderived from the Sentinel 2 (S-2) multispectral image.

2.3. Images

One cloud-free Sentinel-2 image (Level 2) was downloadedfrom ESA Sentinels Scientific Data Hub, this image was obtained onthe November 15, 2017, it was selected as the closest date fromground sampling without rain events in the previous three weeks.Fig. 3 depicts a detailed airborne image (pixel size 50 cm; plani-metric precision 1m rmse), it was downloaded from the Nationalaerial Orthophotography (Instituto Geogr�afico Nacional, 2014), theflight was made in the 2017 campaign. Images were analyzed withOpen Source Geospatial Information System (QGIS., n.d.) and theSentinel Application Platform v7.0 (SNAP-ESA. Sentinel ApplicationPlatform v7.0.2, n.d.).The remote sensing indices, explained in

Fig. 1. Airborne photography of the study plot with sampling points a few days after a high-intensity rain event, the top of the hills are covered by spontaneous vegetationdominated by Stipa tenacissima species; throughout the cultivated slopes, rill erosion is clearly visible in different small watersheds on the plot (a). Representative soil profile (b),and thirty core samples from 0 to 30 cm depth (c).

Fig. 2. Volumetric soil moisture at field capacity (FC vol) and Permanent Wilting point(PWP vol). The difference between these two measures results in the available water(AW vol). Changes according to the gypsum content in 90 soil samples.


385

section 2.4, were derived by ESA’s SNAP Sentinel-2 Toolbox soft-ware. Hereinafter these indices will be qualified as “S-2”.

2.4. The thematic indices

The visible and near infra-red (NIR) part of the electromagneticspectrum was used to compute three different thematic indices.The Brightness Index (BI; Equation (2)) is related to the brightnessof soils, which in turn is influenced by soil moisture, presence ofsalts and organic matter content of soil surface (Escadafal, 1989).The Shape index (IF; Equation (3)) is similar to a coloration index, itwas proposed by Escadafal (1994) for the identification of degradedcalcareous and gypsiferous soils in Tunisia. The normalized differ-ence vegetation index (NDVI; Equation (4)) indicates the photo-synthetic capacity or the energy absorbed by plant canopies, hence,the amount of healthy vegetation. This index can distinguish be-tween soil and vegetation,minimizing topographic effect; however,the influence of soil background is significant (Huete et al., 1994). Inarid and semi-arid environments with scarce vegetation, the NDVIindex shows spectral properties of soils, and considering the visiblerange, it can be a proxy of environmental degradation (Escadafalet al., 1994, pp. 253e259); it varies between �1.0 and þ 1.0.

BI¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�R2 þ G2

�.2

r(2)

FI , ¼ ,ð2R � G� BÞ=ðG� BÞ; (3)

NDVI , ¼ ,ðNIR � RÞ=ðNIRþRÞ (4)

where the letters refer to the reflectance values acquired in the blue(B, Band 2, 490 nm); green (G, Band 3; 560 nm); red (R, Band 4;665 nm); and near-infrared (NIR, Band 8; 842 nm)wavebands (ESA,2015).

2.5. Statistical analyses

The Gaussian distribution of soil variables was checked; only thegypsum content was transformed using the square root of thevariable to approximate to a Normal distribution. Parametric ana-lyses were used to calculate descriptive statistics and correlationsbetween variables. Significant differences between soil layers wereestimated by one way ANOVA and Kolmogorov-Smirnov test forparametric and non-parametric distributions respectively. A prin-cipal component analysis (PCA) was used to reduce the dimensionsof data sets and find out Components or Factors with the highestvariance. The SPSS software was used to perform statistical analysis(SPSS-Inc., 2009).

3. Results

3.1. Soil organic carbon, water, and gypsum

Table 1 shows significant differences across the three soil layersstudied. The first two layers: 0 to 10 and 10e20 cm are similar, butthey are different from the deeper layer (20e30 cm), which showsless carbon, higher bulk density, and less available water content.Gypsum content was also significantly higher at the deepest layer.Gypsum values have been expressed with the median and quartilesand the significant differences established by non-parametric tests.

Soil bulk density and gypsum concentration are increasingsimultaneously with depth. Any increase in these variables, whichmay be found in deeper soil layers (Tables 1, 20e30 cm depth), havesignificant effects onwater content; however, only gypsum contentshows significant negative correlationwith plant available water inthis soil. The relationship between these two variables and watervolume in soil can be observed in Table 2.

The influence of gypsum on water availability can be seen inFig. 2. In these soils, higher gypsum content fundamentally affects

Fig. 3. Location of sampling points at the study plot. On the left aerial orthogonal image obtained in 2017 (latest images PNOA https://pnoa.ign.es/). On the right Sentinel-2 imageLevel 2. Brightness Index (“SNAP BI”, Soil Radiometric Index. SNAP software). Below, four examples (samples No. 3, 14, 1 and 28) of spectral reflectance curves performed with theASD spectroradiometer of samples taken at three different soil depth; 0 to 10 (red), 10 to 20 (green) and 20e30 cm (blue). The Y-axis shows Reflectance, the X-axis showswavelength (nm).


386

the volume of water available at the Permanent Wilting Point(PWP). As a result, the difference between Field Capacity (�33 kPa)and PWP (�1500 kPa) is getting smaller, and therefore the AWC issignificantly affected. Arguably, the increase of gypsum content intopsoils, as a consequence of erosion processes, will influencewateravailability.

3.2. Soil reflectance and indices

It was hypothesized that erosion processes lead to the exposureof the deeper soil horizons, having low SOC and high gypsumcontent, and this different composition can be detected by soilreflectance. In these sloping plots, ASDmeasured reflectance allowssoils to be classified into four general types, concerning theirspectral signature. A group of soils having high reflectance, indi-cating low SOC content and high gypsum content, had spectralsignatures like those located at the bottom left side of Fig. 3; amongthem, several soils had very similar reflectance for the three layersstudied, such as the sample 3; other soils had the deepest layer(20e30 cm) very close to the Cy horizon, so that with high gypsumcontent, and very light colors, as was the case of sample 14 withhigh reflectance for the 20e30 cm layer. Another group of soilsshowed lower surface reflectance, indicating higher SOC and lowgypsum content, in Fig. 3 they are located at the bottom right side,and once again, there were soils with the upper 30 cm very ho-mogeneous, like sample 1; and other soils had less reflectance atthe 20e30 cm layer, this would be the case of soils with a highercarbon content throughout the whole depth studied (from 0 to30 cm).

Following the equations mentioned above, BI, FI and NDVIindices were calculated for ASD and S-2 reflectance. The figuresobtained for these indices were different. Considering the signifi-cant differences between the two upper layers (0e10 and10e20 cm) representing tilled soils, and the deeper one (from 20 to30 cm depth) untilled, a correlation analysis between the variablesconsidered in this study was divided into two blocks, surface soil,(0e20 cm) and the whole profile (0e30 cm). Table 3 shows thecorrelations between indices obtained by ASD measurements andthe corresponding indices obtained from the reflectance of pixels inthe S-2 image.

The pairs of indices obtained from S-2 and ASD are positivelyand significantly correlated, especially the FI_S-2 and FI-ASD,

whose correlation is 0.49 for tilled layers. It is also noteworthythat for the indices calculated in soil samples from 0 to 30 cm depththe significance of correlation persisted, although diminished forNDVI and FI, but not so for BI.

3.3. Relationships between soil reflectance and soil properties

Again, considering the significant differences between the twofirst layers and the at 20e30 cm depth, two principal componentanalysis (PCA) were performed separately, one of themwith valuesobtained from 60 samples taken at 0 to 10 and 10e20 cm depth,and the other PCA analysis conducted with 30 samples taken from20 to 30 cm depth. The slope at the sampling point position, theSOC estimated as C Stock, the volumetric AW and the gypsumcontent, transformed by square root (Sqr Gyp) were used as groundindicators. The olive tree trunk diameter, ranging from 7.6 to22.5 cm, was a biological indicator of the influence of soil condi-tions. In this study, trunk diameter showed a positive correlationwith C stock in soils (0e30 cm) (r ¼ 0.57, p < 0.01), and negativewith gypsum content (r ¼ �0.53, p < 0.01). Finally the set of indi-rect indicators of soil condition: BI, FI, and NDVI, derived by S-2, andcalculated from ASD measurements was also included. The objec-tive was to determine the information that can be retrieved fromthese indices and the relationship between these indices obtainedfrom different sources. Considering the topsoil or soil that wasinfluenced by tillage (0e20 cm depth), Fig. 4 shows the projectionof variables on the plane set by Factors 1 and 3 which extract 70% ofthe variance; it depicts the relationships between soil and vegeta-tion variables and the group of Indices. Both brightness indices areon the right side of the factor plane, close to gypsum concentrationand slope, with which they are related. On the opposite side, themagnitude of tree diameter, diametrically opposite to the Slope ofthe plot, is located close to the NDVI index obtained in the S-2image. On the same side of the factor plane, we can find the C stock,close to both Shape indices (FI) and close to them the AWC, which isin turn, opposite to the gypsum content.

The contributions of variables on the PCA Factor coordinatesbased on correlations are shown in Table 4. The first PCA Factor isinvolving C stock, AWC, Olive trunk diameter, and NDVI and Shapeindices, all of them are inversely related to gypsum content. Ac-cording to this Factor, higher terrain slope tends to coincide withsoils having higher gypsum content, less C stock and less wateravailability. These results are valid for both PCA analysis, 0 to 20 and20e30 cm depth. In both cases, Factor 1 extracts 55% of the variance(Table 4). The indices obtained from ASD showed stronger corre-lations with this Factor than those obtained from S-2 images.

The second Factor of PCA is specifically related to the slope of theplot, with a weak relationship with the tree diameter, especially forthe set of 20e30 cm depth. The third Factor is of special interestbecause it is made up by the simultaneous variations of treediameter and water availability, being the other factors negligible,

Table 1Mean and standard deviation of soil parameters. FC ¼ volumetric soil moisture at Field Capacity; PWP ¼ volumetric soil moisture at Permanent Wilting Point. Gypsumconcentration is described with the median and quartiles (Q25 and Q75) due to the lack of Normal distribution of this variable. Different letters indicate significant differences,p < 0.05.

Soil parameters at different depth 0e10 cm (n ¼ 30) 10e20 cm (n ¼ 30) 20e30 cm (n ¼ 30) Average 0e30 cm (n ¼ 90)

SOC (g kg�1) 12.8 ± 4.2 a 10.9 ± 4.6 a 7.4 ± 4.8 b 10.4 ± 5.1Bulk density (Mg m�3) 1.1 ± 0.2 a 1.2 ± 0.1 a 1.4 ± 0.2 b 1.2 ± 0.2C Stock (Mg ha�1) 14.5 ± 5.3 a 12.6 ± 6.1 a 9.7 ± 6.4 b 37.8 ± 18.3Water content at FC (%) 32 ± 5 a 30 ± 6 a 34 ± 7 a 32 ± 6Water content at PWP (%) 17 ± 5 a 18 ± 6 a 24 ± 8 b 20 ± 7Available Water (%) 14 ± 4 a 12 ± 5 a 10 ± 6 b 12 ± 6Gypsum median (%) 80.7 (70e86) a 81.6 (61e89) a 89.3 (46e94) b 82 (59e90)

Table 2Correlations between bulk density and gypsum concentration (squared root) andwater content of soils (FC: volumetric water content at Field Capacity; PWP: volu-metric water content at Permanent Wilting Point).

Soil Parameters Bulk density (n ¼ 90) Sqr Gypsum (n ¼ 90)

Water content at FC (%) R ¼ 0.59; p < 0.001 R ¼ 0.34; p ¼ 0.001Water content at PWP (%) R ¼ 0.47; p < 0.001 R ¼ 0.64; p < 0.001Plant Available Water (%) R ¼ 0.07; p ¼ 0.52 R ¼ � 0.44; p < 0.001


387

especially at the deepest layer. The information carried out by thisthird Factor is of concern in this water-scarce environment, so thatFig. 4 shows the spatial distribution of sampling points according toFactors 1 and 3 of the PCA. The figures show the two shallow layersseparately from the deepest layer. First, we can observe the similarlocation of samples in the plane formed by these factors whenlayers are alike since the points usually appear very close. In red, thesoils corresponded to 0e10 cm depth, and in green, soils were from10 to 20 cm depth; this was expected given that there were nosignificant differences between them (Table 1). Factor 1, whichextracts the maximum variance, is indicating the carbon content,the water content that is maximum to the left of the created spaceand is minimal on the right. Similarly, on this axis of Factor 1, on theleft are the samples with the lowest gypsum content, and on theright, those with the highest gypsum content, which coincidedwith areas with greater slope. Factor 3, on the Y-axis, represents thehealth of the trees, which is measured indirectly through thediameter of the trunks, which is also related to the plant availablewater. This distribution of the PCA indicates that themost degraded

soils are in the lower right quadrant.When comparing charts in Fig. 5, we can observe that several

soils appeared at approximately the same location in both analyses(charts a and b), this means that for these soil samples, the threelayers were similar, for example, samples 4 or 17, always in thelower right quadrant, or samples 8 and 27 always in the lower leftquadrant. When samples were placed in a different quadrant, therewere more differences between the variables in surface and depth,as was the case of sample 14, whose spectral signature is alsoobserved in Fig. 1, which shows more reflectance in the deepestlayer, and therefore with greater gypsum content and less SOC.

It can also be noted that soil samples of the upper layers (0e10and 10e20 cm depth) on the upper chart of Fig. 5 (a), are mainlylocated at the confluence of the axes, that is, this is a sample pop-ulation of similar data, while in the deeper layer (20e30 cm) on thebottom chart of Fig. 5 (b), soil samples tend to appear scattered, andthere are none in the central zone.

According to the PCA analysis, there is a link between soilbrightness and gypsum content, similarly the FI index is related tothe C stock and NDVI appears close to the variable of diameter ofolive trees. Fig. 6 depicts soil and the tree variables studied in thefirst 20 cm, grouped according to regular intervals of the indicesobtained from the pixels of the S-2 image, with which they can bestbe monitored. From left to right some soils have progressively moregypsum content (Fig. 6 a), less SOC (b) and a smaller diameter of theolive trees (c). Brightness values greater than 0.3 are an indicator ofsoil problems, which are reflected in a lower development of trees.The FI index, represented versus the C stock values, shows signifi-cant differences in SOC content from values of 2.4. The NDVI is theindex that best represents the diameter of the olive trees, althoughsignificant differences appear only for values greater than 0.18.

4. Discussion

Erosion in the study area has been studied in recent years. Theestimated soil loss was reported as 30 Mg ha�1 yr�1 (Sastre, 2017);moreover, losses of 93 Mg ha�1 were recorded after a single high-intensity precipitation event of 55 mm h�1 for 10 min (Bienes &Marques, 2008) leading to an average soil loss of 0.6 cm. Alarm-ing amounts of annual erosion rates were noticed for other slopingolive groves managed by conventional tillage practices, forexample, 10.4 Mg ha�1yr�1 (Martinez et al., 2006), 21.5 Mgha�1yr�1 (G�omez & Gir�aldez, 2007), 41.4 Mg ha�1yr�1 (Bruggemanet al., 2005) or 50 Mg ha�1yr�1 (Karamesouti et al., 2015). Sucherosion rates should be reduced to a tolerable limit of around1 Mg ha�1 yr�1 (Verheijen et al., 2009), especially for degradedshallow soils.

One of the effects of soil loss is the low C Stock; on average,38 Mg ha�1 of C stock was observed in the 30 cm of soil depth inthis study. It concurs with the range of C Stock from the study of 45different soil profiles in the olive groves in Spain that is39.9 ± 28.3Mg ha�1 over 1m depth (Rodríguez-Murillo, 2001). Thisrange is considered low according to the global estimates

Table 3Correlations between indices. The remote sensing indices were calculated in the Sentinel-2 image (date 2017-11-15) with SNAP software. Brightness Index (BI_S-2), ShapeIndex (FI_S-2); Normalized Difference Vegetation Index (NDVI_S-2). The same indices were calculated from the blue, green, red and NIR wavelengths obtained with theanalytical spectral device, BI_ASD; FI_ASD and NDVI_ASD. (*) Marked correlations are significant.

Indices Depth 0e20 cm (tilled layers)N ¼ 60

Depth 0e30 cm (both tilled and untilled)N ¼ 90

NDVI_S-2 BI_S-2 FI_S-2 NDVI_S-2 BI_S-2 FI_S-2NDVI_ASD 0.44 * p < 0.001 0.31 * p < 0.005BI_ASD 0.38 * p < 0.005 0.37 * p < 0.005FI_ASD 0.49 * p < 0.001 0.37 * p < 0.001

Fig. 4. Projection of variables on the plane formed with Factors 1 and 3. Physical-Chemical soil variables (tilled soil 0e20 cm depth) were carbon stock in Mg ha�1 (C-Stock); Available water capacity (AWC), Gypsum (Sqr Gyps). The slope and Tree Di-ameters were also considered. The indices calculated in the Sentinel-2 image (date2017-11-15) with SNAP software were: Brightness Index (BI_S-2), Shape Index (FI_S-2);Normalized Difference Vegetation Index (NDVI_S-2). The same indices were calculatedfrom the blue, green, red and infra-red wavelengths obtained with the analyticalspectral device, BI_ASD; FI_ASD and NDVI_ASD. Factor 1 is related to slope and gypsumcontent. Factor 3 is related to olive tree diameter and water scarcity.


388

elaborated by the FAO (2017). For a 30 cm depth, the C stock rangesfrom 20 to 120 Mg ha�1. Concurrently, the available water content(AWC) declines. Olive trees are not water-demanding plants, and

they can grow in different types of soils, producing fruits in areaswith less than 200 mm of annual rainfall (Gucci & Fereres, 2012),although the optimal condition is around 600 mm of annual pre-cipitation. However, water shortages and soil quality impact theplant vigor and survival rate (Sibbett & Osgood, 2004). Olive pro-duction in these soils, considering the average of good and badyears, is between 12 and 15 kg per tree (data not published). Pro-duction depends on the climate, soil, age, and tree variety. How-ever, these figures are comparable with the olive production ofdifferent varieties growing in other drylandswith less than 250mmof annual rainfall with a maximum production of around 10.6 kgper tree (Aïachi Mezghani et al., 2019). The production is, however,lower than the yields of olive trees in highly productive irrigatedareas and better soils that reach an average of 50 kg per tree (FAO,n.d.).

Several issues are damaging the development and production ofolive trees, and all are interrelated: for example, low SOC hamperssoil structure and decreases soil’s ability to hold water (Castro et al.,2008; Palese et al., 2014); a high gypsum content results in an in-crease of water volume unavailable to the plants at the PermanentWilting Point. Reported publications also describe the negativeimpact of gypsum on water availability (Moret-Fernandez &Herrero, 2015). The shallow soils, technically not very suitable forgrowing crops, becomemarginal when the first centimeters of soilsare lost. In addition, when gypsum content exceeds 35%, wateravailability drops under 10% (Boyadgiev, 1974). Similar results werenoticed in the soils of this study; soil layers under 20 cm depthshowed AWC of 10%. Furthermore, over this depth, the massive Cyhorizons were preventing further water availability and rootdevelopment, which are usually extended only on the upper layersof soils (Parra-Rinc�on et al., 2002). The complex topography of theplot in this study created three micro-watersheds experiencingsediment loss and sediment deposition at different sites, leading toan accumulation of soil organic carbon (SOC), fine particles (silt andclay), and water in certain areas. In these sediment sinks, we couldfind better soil conditions (Hill & Schütt, 2000) and, therefore,favorable conditions for olive tree development. These areasshowed lower reflectance compared to the upslope areas.

The described processes and changes can be monitored usingremote sensing indices (Dorigo et al., 2007). The correlations be-tween the indices derived from the S-2 image spectral data and theASD laboratory-measured spectral data were significant (between

Table 4Factor coordinates of the variables, correlations with three main factors and variance explained. On the left the principal component analysis (PCA) performed with samplesbetween 0 to 10 and 10e20 cm depth; on the right, the PCA performed only with samples taken at 20e30 cm depth. Physical-Chemical soil variables were carbon stock in Mgha-1 (C-Stock); Available water capacity (AWC), Gypsum (Sqtr.Gyps). The slope and tree’s Diameters were also considered. The Indices calculated in the Sentinel-2 image (date2017-11-15) with SNAP software were Brightness Index (BI_S-2), Shape Index (FI_S-2); Normalized Difference Vegetation Index (NDVI_S-2), and finally, the same indices werecalculated from the blue, green, red and infra-red wavelengths obtained with the analytical spectral device, BI_ASD; FI_ASD and NDVI_ASD.

Variables0 to 20 cm depthFrequently tilled

20 to 30 cm depthUntilled

Factor 1 Factor 2 Factor 3 Factor 1 Factor 2 Factor 3

Slope% 0,58 0,72 �0,28 0,44 0,83 0,18Tree Diameter �0,70 0,43 0,54 �0,76 0,05 0,60C Stock Mg/ha �0,86 0,21 �0,07 �0,88 0,32 0,04AWC m3/m3 �0,79 0,23 �0,38 �0,71 0,40 �0,53Sqr Gyps 0,77 0,32 0,23 0,82 0,29 0,04BI_S-2 0,14 �0,26 �0,16 0,20 �0,20 �0,18FI_S-2 �0,27 0,46 0,03 �0,22 0,27 0,30NDVI_S-2 �0.38 0.22 0.18 �0.38 0.08 0.22BI_ASD 0,48 �0,05 0,13 0,25 �0,20 �0,06FI_ASD �0,68 0,16 0,01 �0,75 0,11 0,08NDVI_ASD �0.72 0.34 �0.08 �0.66 0.15 0.32

% Total variance 55.3 18.0 11.6 54.5 20.9 13.7Cumulative % 55.3 73.3 84.9 54.5 75.4 89.1

Fig. 5. Projection of soil samples on the plane formed by PCA Factors. Samples taken at0e10 and 10e20 cm depth are represented at the upper figure (a). Samples taken at10e30 cm depth are represented at the bottom figure (b).


389

0.39 and 0.49). Their contributions to the PCAwas similar, as can beobserved from Table 4. In sparse or low vegetative areas, like thisolive grove, the influence of soil on reflectance values is very strong.From the PCA results, it was concluded that BI-ASD was morerelated to the gypsum content. Given the influence of the topog-raphy on the presence of gypsum at the topsoil, it was not sur-prising that the BI_S-2 was related to the slope of the plot.

The gypsiferous soils show a wide range of colors, moderatelydark such as 10 YR 5/3 with less gypsum or light colors such as 2.5Y8/1, which is due to the high gypsum content of Cy1 horizon. Thesecolors are easily observed in the satellite imagery, especially whengypsum forms crusts, whose high reflectance is noticed if soils havemore than 70% gypsum content (Escadafal et al., 1994, pp.253e259). As a result, high gypsum content in the topsoil can be an

indicator of soil degradation and desertification by erosion(Khanamani et al., 2017). In this study, the BI was related to thegypsum content, and the FI was related to the C stock, especiallyFI_ASD; studies have established the relationship between theshape of the spectral signatures and SOC (Henderson et al., 1992;Summers et al., 2011) which influences, in turn, the hue of soils.

The NDVI has been used as a proxy of net primary production;therefore, it is used to assess the health and coverage of vegetation(Higginbottom & Symeonakis, 2014). This index can be influencedby soil background reflectance in scarcely vegetated areas (Tuckeret al., 1985); however, it has also been used to estimate soiltexture and water content (Lozano-García et al., 1991), moisture inshallow soils (Wang et al., 2007), nitrogen and C content in soils(Sumfleth & Duttmann, 2008), and soil color, especially hue (Singhet al., 2006). From Fig. 3 it can be observed that the NDVI_S-2 isrelated to the tree diameter that indicates the plant vigor. However,these soils had a vegetation cover of around 20%, so the effect of soilreflectance was also prominent. However, the NDVI_ASD, calcu-lated from soil samples measured at the laboratory, could notrepresent the plant vigor, and this index was more related to theSOC and AWC (Fig. 3). In general, the results of PCAwere similar forthe two depths of study (0e20 and 20e30 cm). Reflectance gath-ered by the S-2 images refer to the upper millimeters of soil;however, the results obtained in this study showed that soil char-acteristics over the 30 cm depth had a general influence on the soilsurface and could be indicated by the vigor of trees, which wasgreater in areas with lower reflectance. These areas received sedi-ment deposition from upper sites and had more SOC and AWC.

Tillage practices mix soil horizons; therefore, the gypsumaccumulated in the C horizon was progressively more abundant inthe upper layers, especially in the most eroded areas of the plot.Fig. 5 (a) shows the more homogeneous characteristics of soilsamples in the upper layers (0e20 cm); these samples were,therefore, homogeneously distributed in the plane formed by thePCA Factors. On the contrary, the analysis performed with samplesof the deepest layer (20e30 cm), as shown in Fig. 5 (b), resulted in aheterogeneous distribution of samples in the plane of Factors, i.e.,none of the samples were located at the origin of the axes. This factindicated that in the deeper layers (untilled), soil characteristicswere more different between samples. In general, the PCA analysisseparates different types of samples influenced by their position atthe slope of the plot. Some samples showed lighter colors at theupper layers, but are darker in the layer underneath. This might bebecause they were located in areas receiving eroded sedimentsfrom higher topographic positions with lighter colors. However,high SOC and AWC in the deeper layers facilitated large trunk di-ameters, for example; this was the case with samples 12 and 21.Other areas showed soils with high reflectance in all the threelayers considered. They are shown at the lower right quadrant ofFig. 5. The areas had trees with smaller trunk diameter as they weregrowing on layers close to the massive Cy horizon with waterlimitations. This Figure shows that soil changes according to theposition on the connectivity system of this plot modeled by erosionprocesses depending on the topographic position and the degree ofslope (Marques et al., 2020). In the present study, there wasconsiderable variation between samples in the same plot; however,three thematic indices could establish the different influences ofwater erosion globally evidenced in tree development. Thedegraded soils evidenced less growing of olive trees that could bedetected by an NDVI value of less than 0.18. The impact of soildegradation on the size of trees was also visible in the aerial pho-tographs (Fig. 3).

The analysis of temporal variations of the thematic indicesderived from temporal remote sensing images can be used formonitoring degradation or recovery of this type of soil if changes

Fig. 6. From left to right the effect of soil erosion on the variables. Mean and standarderror of Gypsum content (a), Carbon Stock (b) and Tree diameter (c) and the corre-sponding remote sensing indices derived by Sentinel 2 image: Brightness Index (BI),Shape Index (FI) and Normalized Difference Vegetation Index (NDVI).


390

are made in the land management practices. Agronomic produc-tivity is related to water availability and SOC (Reeves, 1997); how-ever, there are not many studies to relate agronomic productivityand erosion (Pimentel et al., 1995). The consequences have beenrevealed in this study; this aspect of erosion can be more effectiveto get the involvement of farmers who, according to surveys in theregion, think that tillage prevents erosion (Marques et al., 2015) andare more worried about water and production then they are aboutsoil erosion (Barbero-Sierra et al., 2018).

5. Conclusion

In the shallow soils of the study area with markedly differinghorizons, tillage practices exposed different colors or hues at thesurface that are indicative of the erosive processes. Certain hues atthe surface indicate soil horizons that are not suitable for plantgrowth and can be observed by remote sensing. According to thePCA results, in these gypsiferous soils, the Brightness Index wasrelated to the gypsum content, the increasing concentration ofwhich decreases the amount of available water in the soil. Therelationship between water content and soil organic carbon wasalso verified. The latter could be observed with the Shape Index.The NDVI, normally used as a vegetation index, was an indicator ofsoil and crop degradation in these sparsely vegetated soils of thestudy area. These three indices are recommended for monitoringchanges and can be related to loss of water availability and cropproductivity. The information can be used to increase awarenessamong land users, instead of figures of tons per hectare of soil loss.

Acknowledgements

This research was funded by the Project ACCION, and is part ofthe Operative Group Le~nosot, supported by the Rural DevelopmentProgramme 2014e2020 of the Comunidad de Madrid Government(Spain). We thank the cooperation of the Finca La Chimenea staffand the work of undergraduate students from the AutonomousUniversity of Madrid.

References

Aïachi Mezghani, M., Mguidiche, A., Allouche Khebour, F., Zouari, I., Attia, F., &Provenzano, G. (2019). Water status and yield response to deficit irrigation andfertilization of three olive oil cultivars under the semi-arid conditions ofTunisia. Sustainability, 11(17). https://doi.org/10.3390/su11174812

Barbero-Sierra, C., Perez, M. R., Perez, M. J. M., Gonzalez, A. M. A., & Macein, J. L. C.(2018). Local and scientific knowledge to assess plot quality in Central Spain.Arid Land Research and Management, 32(1), 111e129. https://doi.org/10.1080/15324982.2017.1377781

Beaulieu, A., & Gaonac’h, H. (2002). Scaling of differentially eroded surfaces in thedrainage network of the Ethiopian Plateau. Remote Sensing of Environment,82(1), 111e122. https://doi.org/10.1016/S0034-4257(02)00028-7

Ben-Dor, E., Levin, N., Singer, A., Karnieli, A., Braun, O., & Kidron, G. J. (2006).Quantitative mapping of the soil rubification process on sand dunes using anairborne hyperspectral sensor. Geoderma, 131(1e2), 1e21. https://doi.org/10.1016/j.geoderma.2005.02.011

Bennett, S. J., & Wells, R. R. (2019). Gully erosion processes, disciplinary fragmen-tation, and technological innovation. Earth Surface Processes and Landforms,44(1), 46e53. https://doi.org/10.1002/esp.4522

Bienes, R., & Marques, M. J. (2008). Rill and interrill erosion produced by a single-storm event in an olive grove in central Spain. EUROSOIL.

Bocco, G., & Valenzuela, C. R. (1993). Integrating satellite-remote sensing andgeographic information systems technologies in gully erosion research. RemoteSensing Reviews, 7(3e4), 233e240. https://doi.org/10.1080/02757259309532179

Borrelli, P., Robinson, D. A., Fleischer, L. R., Lugato, E., Ballabio, C., Alewell, C.,Meusburger, K., Modugno, S., Schuett, B., Ferro, V., Bagarello, V., Van Oost, K.,Montanarella, L., & Panagos, P. (2017). An assessment of the global impact of21st century land use change on soil erosion. Nature Communications, 8. https://doi.org/10.1038/s41467-017-02142-7

Boyadgiev, T. G. (1974). Contribution to the knowledge of gypsiferous soils. AGON/SF/SYR/67/522. Rome: FAO.

Bruggeman, A., Masri, Z., Turkelboom, F., Z€obisch, M., & El-Naheb, H. (2005). Stra-tegies to sustain productivity of olive groves on steep slopes in the northwest of

the Syrian Arab Republic. In J. Benites, M. Pisante, & F. Stagnari (Eds.), Integratedsoil and water management for orchard development: Role and importance (Vol.10, pp. 75e87). FAO. Land and Water Bulletin.

Castaldi, F., Palombo, A., Santini, F., Pascucci, S., Pignatti, S., & Casa, R. (2016).Evaluation of the potential of the current and forthcoming multispectral andhyperspectral imagers to estimate soil texture and organic carbon. RemoteSensing of Environment, 179, 54e65. https://doi.org/10.1016/j.rse.2016.03.025

Castro, J., Fern�andez-Ondo~no, E., Rodríguez, C., Lallena, A. M., Sierra, M., & Aguilar, J.(2008). Effects of different olive-grove management systems on the organiccarbon and nitrogen content of the soil in Ja�en (Spain). Soil and Tillage Research,98(1), 56e67. https://doi.org/10.1016/j.still.2007.10.002

Condit, H. R. (1970). The spectral reflectance of American Soils. PhotogrammetricEngineering, 36(9), 955e966.

Dematte, J. A. M., Campos, R. C., Alves, M. C., Fiorio, P. R., & Nanni, M. R. (2004).Visible-NIR reflectance: A new approach on soil evaluation. Geoderma,121(1e2), 95e112. https://doi.org/10.1016/j.geoderma.2003.09.012

Dorigo, W. A., Zurita-Milla, R., de Wit, A. J. W., Brazile, J., Singh, R., &Schaepman, M. E. (2007). A review on reflective remote sensing and dataassimilation techniques for enhanced agroecosystem modeling. InternationalJournal of Applied Earth Observation and Geoinformation, 9(2), 165e193. https://doi.org/10.1016/j.jag.2006.05.003

Dubovyk, O. (2017). The role of remote sensing in land degradation assessments:Opportunities and challenges. European Journal of Remote Sensing, 50(1),601e613. https://doi.org/10.1080/22797254.2017.1378926

Dwivedi, R. S., & Ramana, K. V. (2003). The delineation of reclamative groups ofravines in the Indo-Gangetic alluvial plains using IRS-ID LISS-III data. Interna-tional Journal of Remote Sensing, 24(22), 4347e4355. https://doi.org/10.1080/0143116031000116994

D’Odorico, P., Bhattachan, A., Davis, K. F., Ravi, S., & Runyan, C. W. (2013). Globaldesertification: Drivers and feedbacks. Advances in Water Resources, 51,326e344. https://doi.org/10.1016/j.advwatres.2012.01.013

ESA. (2015). Sentinel-2 user handbook. In ESA standard document (p. 64). EuropeanSpace Agency. https://sentinel.esa.int/documents/247904/685211/Sentinel-2_User_Handbook.

Escadafal, R. (1989). Remote sensing of arid soil surface color with Landsat thematicmapper. Advances in Space Research, 9(1), 159e163. https://doi.org/10.1016/0273-1177(89)90481-X

Escadafal, R., Belghith, A., & Ben Moussa, H. (1994). Indices spectraux pour lad�egradation des milieux naturels en Tunisie aride. 6�eme Symp. Int. Mesures Phy-siques et Signatures En T�el�ed�etection. ISPRS-CNES.

FAO. (2017). Global soil organic database (at 30 arcsec). FAO and Euro-MediterraneanCenter on Climate Change Foundation. http://www.fao.org/3/a-i7292e.pdf.

FAO. (n.d.). Data bases & software. Olive. Yield. Land & water. Retrieved February 3,2020, from http://www.fao.org/land-water/databases-and-software/crop-information/olive/en/.

Fleskens, L., & Stroosnijder, L. (2007). Is soil erosion in olive groves as bad as oftenclaimed? Geoderma, 141(3e4), 260e271. https://doi.org/10.1016/j.geoderma.2007.06.009

Franklin, J. A. (1977). Suggested methods for determining water content, porosity,density, absorption and related properties (Part 1). In I. Pergamon Press (Ed.),Suggested methods for determining water content, porosity density, absorption andrelated properties and swelling and slake-durability index properties (pp.141e151). International Society for Rock Mechanics. Commission on Stan-dardization of Laboratory and Field Tests.

Franzini, M., & Lezzerini, M. (2003). A mercury-displacement method for stonebulk-density determinations. European Journal of Mineralogy, 15, 225e229.https://doi.org/10.1127/0935-1221/2003/0015-0225

Galv~ao, L. S., Formaggio, A. R., Couto, E. G., & Roberts, D. A. (2008). Relationshipsbetween the mineralogical and chemical composition of tropical soils andtopography from hyperspectral remote sensing data. ISPRS Journal of Photo-grammetry and Remote Sensing, 63(2), 259e271. https://doi.org/10.1016/j.isprsjprs.2007.09.006

G�omez, J. A., Battany, M., Renschler, C. S., & Fereres, E. (2003). Evaluating the impactof soil management on soil loss in olive orchards. Soil Use & Management, 19(2),127e134. https://doi.org/10.1111/j.1475-2743.2003.tb00292.x

G�omez, J. A., & Gir�aldez, J. V. (2007). Soil and water conservation. A Europeanapproach through ProTerra projects (Proceedings of the European Congress onAgriculture and the Environment).

Gomez, C., Viscarra Rossel, R. A., & McBratney, A. B. (2008). Soil organic carbonprediction by hyperspectral remote sensing and field vis-NIR spectroscopy: AnAustralian case study. Geoderma, 146(3e4), 403e411. https://doi.org/10.1016/J.GEODERMA.2008.06.011

de Graaff, J., Duran Zuazo, V.-H., Jones, N., & Fleskens, L. (2008). Olive productionsystems on sloping land: Prospects and scenarios. Journal of EnvironmentalManagement, 89(2, SI), 129e139. https://doi.org/10.1016/j.jenvman.2007.04.024

Gucci, R., & Fereres, D. (2012). Olive. In P. Steduto, T. C. Hsiao, D. Fereres, & D. Raes(Eds.), Crop yield response to water. FAO irrigation and drenage paper. No. 66 (pp.300e313). FAO.

Henderson, T. L., Baumgardner, M. F., Franzmeier, D. P., Stott, D. E., & Coster, D. C.(1992). High dimensional reflectance analysis of soil organic matter. Soil ScienceSociety of America Journal, 56, 865e872. https://doi.org/10.2136/sssaj1992.03615995005600030031x

Hernandez, A. J., Lacasta, C., & Pastor, J. (2005). Effects of different managementpractices on soil conservation and soil water in a rainfed olive orchard. Agri-cultural Water Management, 77(1e3), 232e248. https://doi.org/10.1016/


391

j.agwat.2004.09.030Higginbottom, T. P., & Symeonakis, E. (2014). Assessing land degradation and

desertification using vegetation index data: Current frameworks and futuredirections. Remote Sensing, 6(10), 9552e9575. https://doi.org/10.3390/rs6109552

Hill, J., Megier, J., & Mehl, W. (1995). Land degradation, soil erosion and desertifi-cation monitoring in Mediterranean ecosystems. Remote Sensing Reviews,12(1e2), 107e130. https://doi.org/10.1080/02757259509532278

Hill, J., & Schütt, B. (2000). Mapping complex patterns of erosion and stability in drymediterranean ecosystems. Remote Sensing of Environment, 74(3), 557e569.https://doi.org/10.1016/S0034-4257(00)00146-2

Huete, A., Justice, C., & Liu, H. (1994). Development of vegetation and soil indices forMODIS-EOS. Remote Sensing of Environment, 49(3), 224e234. https://doi.org/10.1016/0034-4257(94)90018-3

Instituto Geogr�afico Nacional. (2014). Instituto Geogr�afico nacional. Plan nacionalde Ortofotografía a�erea PNOA. In Ministerio de Fomento de Espa~na. http://centrodedescargas.cnig.es/CentroDescargas/index.jsp.

IPCC. (2006). Intergovernmental Panel on climage change guidelines for nationalgreenhouse gas inventories. National Greenhouse Gas Inventories Programme.Institute for Global Environmental Strategies (IGES) https://www.ipcc-nggip.iges.or.jp/public/2006gl/.

Karamesouti, M., Detsis, V., Kounalaki, A., Vasiliou, P., Salvati, L., & Kosmas, C. (2015).Land-use and land degradation processes affecting soil resources: Evidencefrom a traditional Mediterranean cropland (Greece). Catena, 132, 45e55.https://doi.org/10.1016/j.catena.2015.04.010

Khanamani, A., Fathizad, H., Karimi, H., & Shojaei, S. (2017). Assessing desertifica-tion by using soil indices. Arabian Journal of Geosciences, 10(13). https://doi.org/10.1007/s12517-017-3054-5

Klug, H. P., & Alexander, L. E. (1974). X-ray diffraction procedures for polycrystallineand amorphous materials (2nd ed.). Wiley.

Lee, H., & Liu, J. G. (2001). Analysis of topographic decorrelation in SAR interfer-ometry using ratio coherence imagery. IEEE Transactions on Geoscience andRemote Sensing, 39(2), 223e232. https://doi.org/10.1109/36.905230

Lozano-García, D. F., Fern�andez, R. N., & Johannsen, C. J. (1991). Assessment ofregional biomass-soil relationships using vegetation indexes. IEEE Transactionson Geoscience and Remote Sensing, 29(2), 331e339. https://doi.org/10.1109/36.73676

Marques, M. J., �Alvarez, A. M., Carral, P., Esparza, I., Sastre, B., & Bienes, R. (2020).Estimating soil organic carbon in agricultural gypsiferous soils by diffusereflectance spectroscopy. Water, 12(1), 261. https://doi.org/10.3390/w12010261

Marques, M. J., Bienes, R., Cuadrado, J., Ruiz-Colmenero, M., Barbero-Sierra, C., &Velasco, A. (2015). Analysing perceptions attitudes and responses of wine-growers about sustainable land management in Central Spain. Land Degradation& Development, 26(5). https://doi.org/10.1002/ldr.2355

Martinez, J. R. F., Zuazo, V. H. D., & Raya, A. M. (2006). Environmental impact frommountainous olive orchards under different soil-management systems (SESpain). The Science of the Total Environment, 358(1e3), 46e60. https://doi.org/10.1016/j.scitotenv.2005.05.036

McCarty, G. W., Reeves, J. B., Reeves, V. B., Follett, R. F., & Kimble, J. M. (2002). Mid-infrared and near-infrared diffuse reflectance spectroscopy for soil carbonmeasurement. Soil Science Society of America Journal, 66(2), 640e646. https://doi.org/10.2136/sssaj2002.0640

Metternicht, G. I., & Zinck, J. A. (2003). Remote sensing of soil salinity: Potentialsand constraints. Remote Sensing of Environment, 85(1), 1e20. https://doi.org/10.1016/s0034-4257(02)00188-8

Moret-Fernandez, D., & Herrero, J. (2015). Effect of gypsum content on soil waterretention. Journal of Hydrology, 528, 122e126. https://doi.org/10.1016/j.jhydrol.2015.06.030

Palese, A. M., Vignozzi, N., Celano, G., Agnelli, A. E., Pagliai, M., & Xiloyannis, C.(2014). Influence of soil management on soil physical characteristics and waterstorage in a mature rainfed olive orchard. Soil and Tillage Research, 144, 96e109.https://doi.org/10.1016/j.still.2014.07.010

Panagos, P., Borrelli, P., Meusburger, K., Alewell, C., Lugato, E., & Montanarella, L.(2015). Estimating the soil erosion cover-management factor at the Europeanscale. Land Use Policy, 48, 38e50. https://doi.org/10.1016/j.landusepol.2015.05.021

Parra Rinc�on, M. A., Fern�andez-Escobar, R., Navarro García, C., & Arquero, O. (2002).Los suelos y la fertilizaci�on del olivar cultivado en zonas calc�areas. Mundi Prensa.

Pimentel, D., Harvey, C., Resosudarmo, P., Sinclair, K., Kurz, D., Mcnair, M., Crist, S.,Shpritz, L., Fitton, L., Saffouri, R., & Blair, R. (1995). Environmental and economiccosts OF soil erosion and conservation benefits. Science, 267(5201), 1117e1123.https://doi.org/10.1126/science.267.5201.1117

QGIS. (n.d.). Geographic information system. OpenSource geospatial foundation

Project. v2.3.0.Chugiak.Reeves, D. W. (1997). The role of soil organic matter in maintaining soil quality in

continuous cropping systems. Soil and Tillage Research, 43(1e2), 131e167.https://doi.org/10.1016/S0167-1987(97)00038-X

Richards, L. A. (1941). A pressure-membrane extraction apparatus for soil solution.Soil Science, 51(1), 377e386. https://doi.org/10.1097/00010694-194105000-00005

Rodríguez-Murillo, J. C. (2001). Organic carbon content under different types ofland use and soil in peninsular Spain. Biology and Fertility of Soils, 33(1), 53e61.https://doi.org/10.1007/s003740000289

Rojas, R. V., Achouri, M., Maroulis, J., & Caon, L. (2016). Healthy soils: A prerequisitefor sustainable food security. ENVIRONMENTAL EARTH SCIENCES, 75(3). https://doi.org/10.1007/s12665-015-5099-7

Sabins, F. F. (1999). Remote sensing for mineral exploration. Ore Geology Reviews,14(3e4), 157e183. https://doi.org/10.1016/S0169-1368(99)00007-4

Sastre, B. (2017). Estimating soil redistribution rates in an agricultural hillside by137Cs: Case study in Central Spain on gypsiferous soil. In U. de Alcal�a (Ed.),Empleo de Cubiertas vegetales en olivar, repercusi�on sobre el suelo, la erosi�on y lacalidad del aceite de oliva virgen (pp. 40e61). Universidad de Alcal�a.

Sastre, B., Barbero-Sierra, C., Bienes, R., Marques, M. J., & García-Díaz, A. (2017). Soilloss in an olive grove in Central Spain under cover crops and tillage treatments,and farmer perceptions. Journal of Soils and Sediments, 17(3). https://doi.org/10.1007/s11368-016-1589-9

Schulte, E. E., & Hopkins, B. G. (1996). Estimation of organic matter by weight loss-on-ignition. In F. R. Magdoff, M. A. Tabatabai, & E. A. Hanlon, Jr. (Eds.), Soilorganic matter: Analysis and interpretation (pp. 21e31). SSSA Special Publication.

Shepherd, K. D., & Walsh, M. G. (2002). Development of reflectance spectral librariesfor characterization of soil properties. Soil Science Society of America Journal,66(3), 988e998.

Sibbett, G. S., & Osgood, J. (2004). Olive production manual. In G. S. Sibbett, &L. Ferguson (Eds.) (2nd., pp. 27e34). Oakland, Calif: University of California.Agriculture and Natural Resources.

Singh, D., Meirelles, M. S. P., Costa, G. A., Herlin, I., Berroir, J. P., & Silva, E. F. (2006).Environmental degradation analysis using NOAA/AVHRR data. Advances inSpace Research, 37(4), 720e727. https://doi.org/10.1016/j.asr.2004.12.052

SNAP-ESA. Sentinel application Platform v7.0.2. http://setp.esa.int.SPSS-Inc. (2009). PASW statistics for windows. SPSS Inc.Sumfleth, K., & Duttmann, R. (2008). Prediction of soil property distribution in

paddy soil landscapes using terrain data and satellite information as indicators.Ecological Indicators, 8(5), 485e501. https://doi.org/10.1016/j.ecolind.2007.05.005

Summers, D., Lewis, M., Ostendorf, B., & Chittleborough, D. (2011). Visible near-infrared reflectance spectroscopy as a predictive indicator of soil properties.Ecological Indicators, 11(1), 123e131. https://doi.org/10.1016/j.ecolind.2009.05.001

Symeonakis, E., & Drake, N. (2004). Monitoring desertification and land degradationover sub-Saharan Africa. International Journal of Remote Sensing, 25(3),573e592. https://doi.org/10.1080/0143116031000095998

Tivet, F., S�a, J. C., de, M., Lal, R., Briedis, C., Borszowskei, P. R., dos Santos, J. B.,Farias, A., Eurich, G., da Cruz Hartman, D., Junior, M. Nadolny, Bouzinac, S., &S�eguy, L. (2013). Aggregate C depletion by plowing and its restoration bydiverse biomass-C inputs under no-till in sub-tropical and tropical regions ofBrazil. Soil and Tillage Research, 126, 203e218. https://doi.org/10.1016/j.still.2012.09.004

Tucker, C. J., Vanpraet, C. L., Sharman, M. J., & Van Ittersum, G. (1985). Satelliteremote sensing of total herbaceous biomass production in the senegalese sahel:1980-1984. Remote Sensing of Environment, 17(3), 233e249. https://doi.org/10.1016/0034-4257(85)90097-5

Verheijen, F. G. A., Jones, R. J. A., Rickson, R. J., & Smith, C. J. (2009). Tolerable versusactual soil erosion rates in Europe. Earth-Science Reviews, 94(1e4), 23e38.https://doi.org/10.1016/j.earscirev.2009.02.003

Vrieling, A. (2006). Satellite remote sensing for water erosion assessment: A review.Catena, 65(1), 2e18. https://doi.org/10.1016/J.CATENA.2005.10.005

Wang, X., Xie, H., Guan, H., & Zhou, X. (2007). Different responses of MODIS-derivedNDVI to root-zone soil moisture in semi-arid and humid regions. Journal ofHydrology, 340(1e2), 12e24. https://doi.org/10.1016/j.jhydrol.2007.03.022

Whiting, M. L., Li, L., & Ustin, S. L. (2004). Predicting water content using Gaussianmodel on soil spectra. Remote Sensing of Environment, 89(4), 535e552. https://doi.org/10.1016/j.rse.2003.11.009

Wrb, I. (2015). Reference Base for Soil Resources 2014, update 2015. International soilclassification system for naming soils and creating legends for soil maps. No. 106.FAO.


392


Assessment of deforestation impact on soil erosion in loess formationusing 137Cs method (case study: Golestan Province, Iran)

Mohammadreza Gharibreza a, *, Mohammad Zaman b, Paolo Porto c, Emil Fulajtar b,Lotfollah Parsaei d, Hossein Eisaei d

a Department of River and Coastal Engineering, Soil Conservation and Watershed Management Research Institute, Agricultural Research, Education andExtension Organization (AREEO), Tehran, 1389847611, Tehran, Iranb Department of Soil and Water Management & Crop Nutrition, Joint FAO/IAEA Division of Nuclear Techniques in Food & Agriculture, Vienna, A-1400,Austriac Department of Agriculture, Mediterranean University, Localit�a Feo di Vito, Reggio Calabria, 86124, Italyd Department of Watershed Management, Golestan Agricultural and Natural Resources Research and Education Center, AREEO, Gorgan, 491567755, Iran


Article history:Received 21 January 2020Received in revised form20 July 2020Accepted 25 July 2020Available online 14 August 2020

Keywords:Hircanian forestsGolestan provinceNuclear techniqueSoil erosionDeforestationAfforestationDry-farming land

a b s t r a c t

Golestan, a province in the North-East of Iran, is characterized by high coverage of loess deposits. Since1963, the area has experienced approximately 200,000 ha deforestation due to land-use changes inagriculture and increasing demand for wood. Approximately, 110,000 ha of the clear-cut lands are underdry-farming, mainly for wheat cropping, and about 86,000 ha have been reforested. This IAEA fundedproject is the first attempt to use nuclear techniques in the East of Hircanian Forest for determination ofon-site impacts of deforestation due to two land-use changes (i.e. dry farming and reforestation).Practicing long-term dry-farming led to 60% soil losses with a mean rate of 2 mm per year. The neterosion rate of croplands on loess deposits in the study area was 32.27 t ha�1 yr�1. Reforestation,cultivation of even-aged Cypress trees since 1993, showed 13 to 60 percent effectiveness in soil con-servation. Dry-farming land use resulted in the loss of 95 t ha�1 soil organic carbon (SOC) stock at a meanrate of 1.7 t ha�1 over 54 years. Cultivating Cypress trees successfully restored the SOC content by 100%compared with the SOC in original forests. The conversion of dry-farming lands to orchards of olive treessince 2004, brought more income for farmers but were less effective in soil conservation because of lowcanopy cover of olive trees. Our data provide key information and guidance for land users and decision-makers about implementing strategic and sustainable conservation practices to restore degraded land.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction

Land-use changes such as clear-cutting of forests hasten soildegradation (erosion) in catchment areas worldwide (Gharibrezaet al., 2013a, 2013b, , 2013c). Deforestation is known as one of themost unsustainable land-use changes that lead to awide range of on-site and off-site negative impacts. These include soil and nutrientlosses, increased emission of greenhouse gases (GHGs), loss ofbiodiversity, loss of habitats, migration of animals, and increased in-cidents of the landslide, debris flow, and floods which are wellrecognized on-site impacts of deforestation. Besides, increasing of

sediment transport and sedimentation in reservoirs, wetlands, lakes,shoaling of coastal sedimentaryenvironments, severe eutrophicationand pollution of water and sediment resources, climate change andsocio-economic impacts are the main results among off-site impacts(Mohamed Abd Elbasit et al., 2013; FAO, 2015, p. 253; Indarto, 2016;TejaswiandMutaqin,2007,p.49;Gharibreza&Ashraf, 2014). Besides,the short and long-term socio-economic effects of deforestationaffected incomes which were dependent on the forest.

Land development projects, agriculture, mining, logging, ur-banization and rural settlements, cattle ranching, transport, powerplants, commercial oil palmplantations, and utilization ofwood fuelare knownas themaindriving forces fordeforestation (Kaimowitz&Angelsen, 1998, p. 153; Gharibreza& Ashraf, 2014; Gharibreza et al.,2013a, 2013b, b). Burning and clear-cutting are the most commondestructive methods that have been applied during deforestation

* Corresponding author.E-mail addresses: [email protected], [email protected]

(M. Gharibreza).






over the last decades. Alternative methods such as strip, shelter-wood, and selective cutting were also used for wood harvestingaround theworld.On theotherhand, silvicultural systemshavebeenused for reforestation of cleared cut lands for industrial wooddemands.

These systems reflect the type of forest structure remaining afterthe initial harvest. Two classes of silvicultural systems can be recog-nized; even-aged and uneven-aged. Even-aged systems generallycreate standswhere the trees are approximately the same age or oneage class. However, sometimes, even-aged systems can result in twodistinctageclassesof treeswhensomeolder treesare left behindafterharvesting. Clear cutting, seed tree, and shelterwood are generallyconsidered to be even-aged silvicultural systems. Uneven-aged sys-tems, on the contrary, aimat creatingormaintaining, by the selection,stands to consist of trees with several distinct age classes.

Degradation of the forest by converting forest land to dry-farmingand silvicultural systems is common in Iran but their on-site impactson soil redistribution are still unknown. The effects of these types ofland use on soil redistribution rates were evaluated in this study.

Evaluation of empirical, conventional, and nuclear-basedmethods (Evans et al., 2017; IAEA, 1998) has resulted in the appli-cation of the nuclear technique of 137Cs to estimate soil redistribu-tion rates on deforestation-induced land uses in the study area.Therefore, the Technical Cooperation (TC) project between Iran andthe International Atomic Energy Agency (IAEA), IRA 5013 wasestablished to find out the effects of deforestation and afforestationon soil erosion in theNorth of Iran using fallout radionuclides (FRNs)techniques. The study method was supervised by the IAEA experts.

Cesium-137 (137Cs) is afissionproduct and atomic bomb-derivedradioisotope that has a half-life of 30.02 years and emits gamma rayswith an energy of 661.6 keV (Poreba, 2006). This radionuclide wasfirst used byYamagata et al. (1963) andRogowski andTamura (1970)to estimate the rates of soil erosion. Analytical methods andmodelsfor estimating soil erosion using 137Cs have remarkably improvedover the last four decades. Ritchie (2007) reported that scientificpublications on the 137Cs technique started in 1961 and reached itsmaximum number in 1999. Newmodels for estimating soil erosionhave also been introduced by Robbins (1978), IAEA (1995, 1998),Rogowski and Tamura (1970), Walling and Quine (1990), Wallingand He (1999), Walling et al. (1999), Zapata and Agudo (2000), andPoreba (2006). Mabit et al. (2008) have evaluated different modelsandnoted the advantages and limitations of using 137Cs and 210Pb forassessing soil erosion. Therefore, 137Cs and 210Pb-based techniquescan be applied both quickly and efficiently to estimate soil loss andredeposition. A PC-compatible software package by Walling et al.(1999) contains improved models, based on several empirical andtheoretical approaches that are applicable to both cultivated andnon-cultivated areas. The package includes the Proportional Model,the Mass Balance I, II, and III Models, the Profile Distribution Model,and the Diffusion andMigration, Models. Specific requirements are,however, needed in applying the models for the underlying as-sumptions, descriptions, and representation of temporal variation(Walling et al., 1999).

The objective of this research was to determine the soil redis-tribution pattern induced by deforestation along the representativetransects of the dry-farming and afforested lands in the GolestanProvince of Iran using the 137Cs technique.


2.1. The study area

The Hircanian Forests (HF) are located along the South of theCaspian Sea, in theNorth of Iran. Their distribution along the northernslopes of theAlborzMountains, covers ca.1.85million hectares in Iran.

The HF includes an important refugium of temperate broad-leavedtrees. The HF stretch across a large area that comprises the threeprovinces of Guilan, Mazandaran, and Golestan (Fig. 1). GolestanProvincehasapproximately452,000haof forest coverage, fromwhich66% are productive, 18% are conserved, and 16% are degraded. A rapidincrease in the humanpopulation associatedwith the developmentofwood industries led to extensive deforestation and a decline incoverage of HF between 1942 and 2005. Between 1976 and 1991,deforestation throughwood harvesting occurred at amean rate of 2.3million m3 per year. Afterward, this deforestation rate decreased to300,000 m3 per year; while since 1981, approximately 75,527 ha ofcleared cut areas have been reforested.

The study area has a semi-humid climate with a mean tem-perature of 16.4 �C and annual precipitation between 700 and1000 mm which mainly occurs in October and sometimes in July.

2.2. Site selection and sampling

A sampling strategy is a critical step for the success of the 137Csmethod in soil erosion and sedimentation studies (Fulajtar, 2017, p.63). In this study, based on the research objectives and distributionof specific land-use changes across the Golestan province, an accu-rate sampling design was established. Field observations showedthat the hillslopes located between 150m and 500m elevations andwith a slope lower than 30% have been experiencing maximumdeforestation over the last decades. The deforestation affectedmainly the loess areas of Golestan Province. Loess is a sedimentarydeposit composed largely of silt-size grains that are looselycemented by calcium carbonate. Such deposits cover older geolog-ical formations and rockunitswithdifferent thicknesses in the studyarea. Muhs (2007) reported that the loess deposits have beendeposited during the Quaternary interglacialeglacial cycles. There-fore, the development of loess deposits in the Golestan Province inthe eastern part of the Caspian Sea (CS) is correlated with sea-levelvariation periods and with the availability of a huge amount ofsilty and sandy sediments, eventually redistributed as aeolian sed-iments layers (loess). Kakroodi et al. (2012) recognized five CS levelcycles after 3260 cal yr BP. Therefore, two major periods of coolingphases occurred, including the 2600e2300 cal yr BP high stand andthe Little Ice Age, which had a strong impact on the CS water level(Gharibreza et al., 2018). Such cooling ages provided a suitableenvironment for aeolianmechanisms to accumulate loess at the toeof the Eastern AlborzMountain range. This idea has been supportedby a newly published work (Rahimzadeh et al., 2019) on the for-mation of aeolian deposits in the Eastern Caspian Sea. Sedimento-logical and dating information has stressed on the accumulation ofloess and fossil dunes in the Early Holocene, probably following theso-called Mangyshlak regression. High erodibility of loess depositsespecially in China has been proved by some authors (Yan et al.,2018; Mohamed Abd Elbasit, 2013) using conventional models.Therefore, another strategy of sampling in the present study was toestimate deforestation-induced erosion rates on loess deposits inthe North-East of Iran.

The comprehensive field observations across the Golestanprovince resulted in the selection of five representative transects(Fig. 1). Site selection was conducted based on different land-usechanges across the study area. Intensive deforestation and con-verting of the forest to dry-farming started in 1963 (Act 41, 1962) byrelaxing the law for the nationalization of forests and sharing landwith laborers and new farmers. Driving forces such as willingnessto create new land ownership, agriculture, cattle ranching, and newdemands for wood by industries have facilitated the deforestation.

A sampling strategy was designated to find out the properreference sites inwestern and easternparts of the Golestanprovinceto estimate variations in the 137Cs radioisotope inventory.

M. Gharibreza, M. Zaman, P. Porto et al. International Soil and Water Conservation Research 8 (2020) 393e405

394

Accordingly, two reference sites in the Kuhmian and Galikeshwatershed were identified with the cooperation of an IAEA expert(IAEA, 1998) in August 2017. Both the reference sites are located inconservative parts of the watershed where original forests are pro-tected by the Forest, Rangelands, and Watershed Management Or-ganization of Iran. The Galikesh reference site was located 1.3 kmfromthe studysite at anelevationof 728m, in theoriginal forest. TheKuhmian reference site was located 1.1 km from the study site at anelevation of 420 m, in the original forest. A sampling of referencesites was carried out and 24 bulk samples and two incrementaldepth samples were collected. A scraper plate with a cutting edgeand having a rectangular metal frame (surface area ¼ 1292.71 cm2)served to collect 2 sectioned cores samples (one for each site) at 2-cm intervals at a total depth of 30 cm. Both the Kuhmian and

Galikesh watersheds included conservative and trading forests anddeforested-induced land uses that gave opportunities to the projectteam to find out appropriate locations for reference sites and sam-pling transects.

Sampling strategy for such studies (IAEA,1998) implies selectingthe nearest location to the reference sites where a variety of sam-pling patterns could be implemented. Conducting sampling pat-terns with the catchment scale was not applicable because of theextensive sub-catchments and limitations in the number of sampleanalyses. Therefore, a transect approach was followed for samplingalong hillslopes with representative land uses.

Accordingly, three of the transects were chosen to represent theeffect of converting forest to dry-farming lands in two separatewatersheds in the western and eastern parts of the Golestan

Fig. 1. Location of the study areas and profiles of sampled transects in representative sites of Golestan Province, Iran: a) Eastward Dry-farming transect, Azadshahr, b) WestwardDry-farming transect, c) Reforested by Cypress, Olive trees and Dry-farming transect Azadshahr, d) Dry-farming transect, Galikesh, and e) Reforested by even-aged Cypress trees,Galikesh.


395

Province (Sites Azadshahr and Galikesh, Table 1 and Fig. 1). Tworepresentative transects of dry-farming lands, located at the Kuh-mian Watershed and west of Azadshahr (AZ) city, where clear-cutting of original forest focused on hilly and loess formations,between 150 m and 400 m elevations (Fig. 1). These sites have auniform slope (22%), Southeast and Southwest aspect, and 340 mand 635 m 168 in length, respectively (Fig. 1 c, d).

In the case of transect “e” (Fig. 1), the presence of hedges (madeby bushes and wood bunches) between the different farm ownersrepresents a discontinuity along the slope, which was considered inthe sampling activity. Therefore, the Southeast direction dry-farming land use was selected in such hillslopes with four farmsto identify soil redistribution patterns.

Sampling in the study area consisted of 93 bulk samplescollected along the five transects (Table 1 and Fig. 1). In this case, ahandy core sampler of 11-cm diameter was used to a depth of25 cm. Also, single samples frommoved sediment at the catchmentoutlet (where deposition occurred) were collected to gain infor-mation on the particle size factor. In total, 21 and 24 samples werecollected with about 20 m intervals from the Azadshahr westwardand eastward dry-farming transects, respectively.

The third dry-farming land use transect with 1000 m length(Fig. 1 d) was located in the Galikesh Watershed, eastern GolestanProvince. This transect has similar topographic conditions andcropping systems as compared to the transects of the Azadshahrcity (Fig. 1a and b). Accordingly, 21 samples, 25 cm deep, werecollected along this transect approximately 50 m from each other.

On the other hand, reforestation of converted forest lands wasanother land-use change in the study area. Cultivation of Cypresstrees using an even-aged method was envisaged by the environ-mental Act71 (1992) in the Golestan Province in 1993. This lawsettled by the Parliament of the Islamic Republic of Iran aimed toperform appropriate practices to preserve natural resources.Therefore, two transects were chosen to explore the effectiveness ofsilvicultural practices in decreasing soil erosion in the study area.Transects with 500 m and 350 m length (Fig. 1 b, e), represent theimplementation of even-aged silvicultural systems located near todry-farming landsatAzadshahrandGalikesh, respectively.19 and10samples were collected from the Azadshahr (Fig. 1 c) and Galikeshsilvicultural (Fig. 1 e) transects with intervals of 25 m and 37 m,respectively. Therefore, 19 and 10 samples were collected from theAzadshahr andGalikesh silvicultural transectswith intervals of 25mand 37m, respectively. The vicinity of transects provides conditionsto compare the rate of soil erosion in different land uses.

2.3. Sample preparation and analysis

The soil samples were oven-dried at 105 �C and ground usingmortar. Then, the sampleswere homogenized, weighed, and sieved.Portions finer than 2 mm were packed in special containersweighing 250 g. The 137Cs activity was measured using a well-calibrated gamma-spectrometry based on high-purity germanium(HPGe) detectors in the laboratories of the Nuclear Science & Tech-nology Research Institute (NSTRI), in Iran. The gamma-spectrometermodel EGPC 80-200-R, (EURISYS MESURES) detector at NSTRI has a

relative efficiencyof 80% and fullwidth at halfmaximum(FWHM)of2.5 keV for 60Co gamma-energy line at 1332 keV. The gamma-spectrometer was calibrated using multi-nuclides standard (POLA-TOMMIXSOURCE) solutions dispersed in soil homogeneously in thesame sampleedetector geometry (250 g Marinelli beaker). Thelower limit of detection depends on efficiency, FWHM, and countingtime, and thus, it differs from sample to sample. The IAEA referencematerials were used for quality control of the gamma-detector.

A total of 119 samples were analyzed for finding out grain sizedistribution and bulk density values. The distribution of course andfine particle size portions was estimated using the ASTM 422 stan-dard and the ASTM D7928 methods, respectively. Besides, to esti-mate the land degradation induced by deforestation, the total soilorganic carbon (SOC) was measured using the Walkley and Black(1934) method. The soil organic carbon stock (SOCs) was esti-mated with the formula proposed by Hiederer and Martin (2011):

SOCs¼ SOCc� BD��1� VS

100

�� LD� 102 (1)

where:SOCS: the total amount of soil organic carbon to a given depth (t

ha-1).SOCC: soil organic carbon content for given depth (%)BD: dry bulk density (kg m�3).VS: volume of stones (%)LD: depth of soil layer (m).

2.4. The conversion model to estimate soil redistribution rates from137Cs measurements

Some conversion models able to convert 137Cs loss (or gain) intovalues of soil erosion (or deposition) are available in the literature(Walling et al., 2007). Considering the type of land-use changes andthe current cultivation along the representative transects, the MassBalance Model 2 (MBM2) was an appropriate choice (Walling et al.,2014). This model considers both the temporal variation in 137Csfallout input and the initial distribution of fresh fallout on thesurface soil. In addition, transects with dry-farming land use havebeen plowed for more than 50 years, while two other selectedtransects have been plowed at least for 30 years before afforesta-tion. The equation for the MBM2 can be written following the formproposed by Walling and He (1999 a, b):

AðtÞ¼Aðt0Þ e�ðt

t0

ðPR=Dþ lÞ dt0

þðt

t0

ð1�GÞIðt0Þ e�ðPR=DþlÞ ðt�t0Þdt0

(2)

where:

R ¼ erosion rate (kg m�2 yr�1);D ¼ cumulative mass depth representing the average plowdepth (kg m�2);

Table 1Characteristics of the representative transects in the study area.

No. Sample number Slope % Transect length (m) Aspect Land uses Area

Transect a 23 22 580 East Dry-farming AzadshahrTransect b 21 22 350 West Dry-farming AzadshahrTransect c 19 22 480 Northwest Even-aged Cypress trees, Olive trees and Dry-farming AzadshahrTransect d 20 20 1000 Northwest Dry-farming GalikeshTransect e 10 20 380 Northeast Even-aged Cypress trees Galikesh


396

l ¼ decay constant for 137Cs (yr�1);I(t’) ¼ annual 137Cs or 210Pbex deposition flux (Bq m�2 yr�1);

G ¼ percentage of the freshly deposited 137Cs fallout removed byerosion before being mixed into the plow layer;P ¼ particle size correction factor;t0 (yr.) ¼ year when cultivation started;A(t0) (Bq m�2) ¼ 137Cs inventory at t0.

If A(t) is greater than the local reference inventory Aref, at asampling point, deposition may be assumed. In this case, the meansoil deposition rateR0 can be calculated from the followingequation:

R0 ¼

ðtt0R0Cdðt0Þe�lðt�t0Þdt0

ð

S

RdS

PP0ðtt0dt0

ð

S

�Iðt0Þg

�1� e�R=H

�.Rþ Aðt0Þ

.D�dS

(3)

where:

His the relaxation mass depth of the initial fallout input;Cd(t0) reflects the radionuclide content of sediment mobilizedfrom all the eroding areas that converge on the aggrading point.Generally, Cd(t0) can be assumed to be represented by theweighted mean 137Cs activity of the sediment mobilized fromthe upslope contributing area S (m2);P0 is a furtherparticle sizecorrection factor reflectingdifferences ingrain size composition between mobilized and deposited sedi-ment; g is the proportion of the annual fallout susceptible toremoval by erosion before incorporation into the soil profile bytillage. According to field observations, a tillage depth ofD¼ 0.25mhas been assumed. The following valueswere used fortheother inputparameters:H¼4,g¼1 (basedon the relationshipbetween the timing of cultivation and the rainfall regime), and

calculatedvaluesof theP factorbasedonparticle sizeanalyses.ThePC-compatible Excel Add-in, developed by Walling et al. (2007)was used to apply the model described above.


3.1. Soil physical properties and organic matter

Particle size analyses confirmed the research hypothesis that theconversion of forest to dry-farming and silvicultural operations hasled to an increase in erosion of loess deposits (see Table 2).

Soils with silty texture are the majority in the study area - up to65% on average. In addition, at reference sites and dry-farmingtransects, the clay content is between 30 and 40 percent (Fig. 2 a, b).

According to Smalley and Smalley (1983), loess layers hadgenerally unimodal, well-sorted, leptokurtic, and negatively-skewed particle-size population. The present research supportsdominating palaeoclimate events in the study area and theconsequent development of paleosols on the basis of granulometricresults. The data showed a significant correlation to unimodal,well-sorted, and clayey loess deposits. This composition pointed toclimate change andmeteoric diagenesis bywhich loamification andsoil formation on the loess developed. All evidence points out avery high sensitivity of altered loess and paleosols to soil erosion,especially when degraded by deforestation.

3.2. 3.2.137Cs inventories at the reference sites

The 137Cs depth distribution of the two sectioned cores collectedfrom the two reference sites (one for each location), showed anexponential decline with depth (Fig. 3 a, b). The shape of theseprofiles confirmed that 70e80% of the 137Cs inventory were presentin the top 12 cm. The 137Cs inventory gained in the reference sites ofAzahshahr and Galikesh was 3849 Bq m�2 and 4062 Bq m�2,

Table 2Physical properties of representative transects in the study area.

Study area Bulk Density (kg m�3) SOC (%) Clay (%) Silt (%) Sand (%) Texture

Transect a (n ¼ 23)Average 1494.6 1.24 35.67 59.63 4.67 Si - C - La

STDEV 163 0.45 4.97 4.12 4.57CV 11 36 14 7 98Transect b (n ¼ 21)Average 1431.7 1.32 33.76 60.67 5.57 Si - C - LSTDEV 119.28 0.25 2.43 3.31 3.76CV 8 19 7 5 68Transect c (n ¼ 19)Average 1385.37 2.32 31.37 61.37 7.26 Si - C - LSTDEV 125 0.73 2.71 2.61 3.35CV 9 32 9 4 46Transect d (n ¼ 20)Average 1429 2.2 33.14 60.24 6.62 Si - C - LSTDEV 112 0.6 2.03 2.81 3.25CV 8 30 6 5 49Transect e (n ¼ 10)Average 1355 3.58 30.10 67.6 2.3 Si-Lb

Si - C - LSTDEV 96 0.83 5.86 4.5 4.69CV 7 23 19 7 200Reference Site1-Azadshahr (n ¼ 12)Average 1400.8 3.7 40.38 56.62 2.09 SeCc

Si - C - LSTDEV 92 0.53 7.7 4.61 0.83CV 6.6 25 19 8 40Reference Site2-Galikesh (n ¼ 12)Average 1307 3.57 29 64.5 6.5 Si-L

Si - C - LSTDEV 101 1.1 3.5 6.6 4.25CV 7 30 11 0.1 11

a Silty-Clayey Loam.b Silty Loam.c Silty Clay.


397

Fig. 2. Particle size distribution along the incremental samples collected in the references sites of a) Azadshahr, b) Galikesh.

Figs. 3. 137Cs depth distribution from the sectioned cores collected at two reference sites, a) Galikesh, b) Azadshahr.


398

respectively. A very low difference (ca. 10%) between the 137Cs in-ventory of the two reference sites, testifies a similar distribution ofthe 137Cs fallout in the study area. A recent work by Tagami et al.(2019), assessing the geographical distribution of the 137Cs in-ventory in India and other areaswith similar latitude and longitude,reports 137Cs inventories of ca 4000 Bq m�2 and confirms the reli-ability of the reference values found in our study areas.

3.3. 137Cs inventories within the study area

The 137Cs inventory along the representative transects by dis-tance from the top of the hillslope is represented in Fig. 4, togetherwith the measured distances from the top hill. More specifically,the 137Cs measurements in the two characterized dry-farming landuses at Azadshahr area (Fig. 4 a and b), represented the mean 137Csinventory of ca. 1421.77 Bq m�2 with single values ranging from271.2 Bq m�2 to 2896.4 Bq m�2 with a CV about ca. 60%.

Further, theGalikesh representative dry-farming landuse transectshowed the mean value of the 137Cs inventory of ca. 1600.91 Bq m�2

with single values ranging from833.6 Bqm�2 to 2340.4 Bqm�2 and a

CV equal to ca. 31%. As mentioned in Table 1, the Azadshahr repre-sentative mixed land use transect consists of cultivation of Cypresstrees under even-aged silviculture, cultivation of olive trees, and dry-farming section.Under these landuses, themean137Cs inventory is ca.1703.04 Bq m�2 with single values ranging from 401.26 Bq m�2 to5608.95 Bqm�2 and a CV equal to ca. 71% (Fig. 4 c). There are not anyconvex and concave forms along these transects and variation in 137Csinventory represents a redistribution pattern. Further, the 137Cs in-ventory along the Galikesh silviculture land use transect representedan approximately 3544.31 Bq m�2 with single values ranging from1119.7 Bqm�2 to 8265.45 Bqm�2 and a CV equal to ca. 65% (Fig. 4 e).Variations of 137Cs inventory in single points of samples in bothdisturbed and undisturbed forest areas can be expected because ofdifferences in canopy covers and ages of trees. All single inventoryvalues in both sites of the dry-farming land are lower than the cor-responding reference value.

The soil redistribution trend shown in Fig. 4 provides useful in-formation for farmers and decision-makers to understand whichpart of the farmneeds to bemodified for soil and nutrient resources.

Variability on the ground and land use (see Fig. 1) have caused

Figs. 4. 137Cs inventory variation along the transects related to the dry-farming land use of (a) Eastward Dry-farming transects, Azadshahr, b) Westward Dry-farming transect, c)Reforested by Cypress and Olive trees transect Azadshahr, d) Dry-farming transect, Galikesh) and e) Reforested by even-aged Cypress trees in Galikesh.


399

variations in the trend of the 137Cs single inventory along the hill-slope (Fig. 4). Correlation between the 137Cs single values and dis-tance from the hilltop of the Azadshahr westward transect of dry-land farming (Fig. 4 b), using the Mann-Kendall test showed a cor-relation coefficient of 0.733 which is significant at the 0.01 level.Dividing hillslopes to small farms is themost common feature in thestudy area. Therefore, the Azadshahr eastward dry-farming transectwas selected along such a hillslope where it has been divided intofour farms by stakeholders. Accordingly, there is no significant cor-relationbetween the 137Cs and thedistance fromthehilltop (Fig. 4 a).Instead, statistics of each four sections point to meaningful corre-lations. For instance, at the first farm (0e163 m), the correlationcoefficient of 0.667 was gained which is significant at the 0.05 level.The second farm (163e363 m), the correlation coefficient of 0.714which is significant at the 0.05 level was obtained. Similarly, thereare two consecutive farms along the third dry-farming transect atthe Galikesh area (Fig. 4 d). Correlation between the 137Cs singlevalues and distance from the hilltop at the first section (0e640m) oftransect “d” using the Mann-Kendall test showed a correlation co-efficient of 0.538,which is significant at the 0.01 level. The statisticaltest showed no significant correlation between 137Cs value and thedistance from the top of the hill from 640 m onward.

The multiple land use along the Azadshahr transect (Fig. 4 c)resulted in a low correlation between the 137Cs inventory and thedistance from the hilltop. There are two unexpected downwarddecreases in the 137Cs inventory of silviculture even-aged Cypresstrees. Meanwhile, the correlation coefficient of 0.905 was obtainedwhich is significant at the 0.01 level for this section. The Mann-Kendall test showed a correlation coefficient of 1.000 which is sig-nificant at the0.01 level forolive trees that points to a realdownwardincreasing in the 137Cs values because of soil redistribution.

Clear variability between the 137Cs sing values and distance fromthe hilltop was observed in the Galikesh silviculture transect (Fig. 4e), which points to unexpected correlation. This transect is locatedbeneath the dry-farming land - some portions of eroded soil movedto the first 100mof this transect and this caused an increase in 137Csinventory.As a result, thefirst 100mof the transect enrichedby137Csin comparison with the local reference value, while the rest of thetransect shows a downward increase of inventory because of soilredistribution.

3.4. Estimating erosion and deposition rates at the sampling points

Asmentioned above, the rates of soil loss or deposition associatedwith the individual sampling points in comparison with the localreference value, have been obtained using the Mass Balance Model II(Eq. (2) and Eq. (3)). The overall results showed that soil erosion wasthe dominant process compared with the deposition along the hill-sides since the beginning of 137Cs fallout in the mid-1950s. The land-use history of the study area (Bobek, 2005) documents that exten-sive deforestation has occurred before the maximum radionuclidefallout (1954e1963). This activity reflects the lower values of 137Csinventory in the study areas if compared with the correspondingreferencevalues.Therefore, the trendof soil erosionand theamountofsoil and nutrients losses can be considered as the main indicators ofanthropogenic activities and susceptibility of loess deposits to landdegradation.

Maximum, minimum and mean values of soil erosion ratesalong the AZ-East direction transect were 81.25, 7.63 and35.86 t ha�1 yr�1 (Table 3; transect a) respectively, while the similarvalues for the AZ-West direction transect were calculated to be69.68, 8.66 and 34.87 t ha�1 yr�1, respectively (Table 3; transect b).The maximum, minimum and mean values of soil erosion ratesalong the Galikesh-Northwest direction transect were 48.4, 15.8,and 29.19 t ha�1 yr�1 (Table 3; transect d). Besides, the hillslope

under silviculture and horticulture practices in the Azadshahr areashows a mean value of erosion rate equals to 31.7 t ha�1 yr�1, withsingle values ranging from 71.9 t ha�1 yr�1 (erosion) to 11.6 t ha�1

yr�1 (deposition) (Table 3; transect c). Such situations in the Gali-kesh area indicate a net erosion rate of 10.01 t ha�1 yr�1 with singlevalues ranging from 39.8 t ha�1 yr�1 (erosion) to 20.9 t ha�1 yr�1

(deposition) (Table 3, transect e).The net erosion rate of dry-farming lands of the East and West of

Golestan Province was calculated to be 35.86 t ha�1 yr�1 and29.19 t ha�1 yr�1, respectively (Table 3; transects a, b). Therefore, themean rate of soil erosion of the whole province for loess soil type is32.27 tha�1 yr�1, onaverage.Results pointedout that ca.160.6kgm�2

(orca.10.97cm)andca.150.6kgm�2 (orca.10.29cm)of soilhavebeeneroded from the AZ-Westward and the AZ-Eastward transects,respectively. Resultantdatahasbeenobtainedbasedonthemeanrateof soil loss along the Azadshahr transects (64.24% and 60.24%), themeanbulkdensityof 1463kgm�3, and the tillagedepthof 250kgm�2

(or ca. 25 cm). Considering the elapsed time of 54 years betweendeforestation and the present research, soil erosion along thesetransects has occurred with the rate of 2 mm yr�1 and 1.9 mm yr�1,respectively.

Results suggest 151.47 kg m�2 (or ca. 10.59 cm) of soil erodedfrom dry-farming land use (transect d). Similarly, these data werecalculated based on the mean rate of soil loss along the Galikeshtransect (60.59%), the mean bulk density of 1429 kg m�3, and thetillage depth of 250 kg m�2 (or ca. 25 cm). Therefore, a long-termrate of soil erosion of 1.96 mm yr�1 has been obtained along thistransect by considering the elapsed time (54 years) betweendeforestation and the present research.

Soil loss based on the amount of the reduction in 137Cs inventoryfrom the silvicultural, horticultural, and dry-farming parts of theAzadshahr mixed land use transect (c) was calculated to be 52.8%,58.62%, and 61.67%, respectively. In such a situation, these sectionserodedby themean rate of 1.8mmyr�1,1,8mmyr�1, and2.1mmyr�1,respectively. Although the upper part of this transect has experiencedthe dry-farming land use for about 30 years, the silvicultural practiceusing even-aged Cypress trees resulted in 15% effectiveness in soilconservation since 1993. Similarly, the cultivation of olive trees in thesecond section (z150 m) of the transect, caused - since 2004 - a 5%soil loss in comparison to the remaining dry-farming land use.

In general, afforestation of dry-farming land inGalikesh results in a

Table 3Soil loss and net erosion along the representative transects.

Transects Soil loss (%) Net Erosion (t ha�1 y�1)

Transect a (n ¼ 23)Average 60.24 35.86Max 92.95 81.25Min 24.75 7.63Transect b (n ¼ 21)Average 64.28 34.87Max 90.78 68.68Min 28.16 8.66Transect c (n ¼ 19)Average 55.75 31.7Max 89.57 71.9Min 45.75b 11.6a

Transect d (n ¼ 20)Average 60.59 29.19Max 79.47 48.39Min 42.38 15.8Transect e (n ¼ 10)Average 12.74 10.01Max 72.43 39.8Min 102b 20.9a

a Soil Accumulation.b Deposition.


400

highereffectiveness insoilerosioncontrol (ca. 34%)ascompared to theadjacent dry-farming land (Table 3; transect e). A long-term volu-metric soil lossof ca. 31.85kgm�2 (orca.2.38cm)witha rateof0.4mmyr�1 has been recorded along this transect. This rate was obtainedusing key resultant data of soil loss (12.74%), the mean bulk density(1355 kg m�3), and the tillage depth of 250 kg m�2 (or ca. 25 cm).

In Fig. 5, the trend of soil redistribution along the transects arepresented. The severe soil erosionoccurring in theupperpart (z25%of transect b), covered with dry-farming land use, is considerablyhigher than the rest of the transect. Interpretation of satellite images(Sentinel-2A, 10 m resolution) and comprehensive field observa-tions documented very low fertility of such farms because of soil andnutrients loss and confirms the above assumption. On the contrary,the annual range of soil erosion along the rest of transect “b” doc-uments value lower than 20 t ha�1.

Intensive soil erosion in the upper section of the transect withsilvicultural and horticultural practices was also observed (Fig. 5, c).Thedivisionof the slope into sectionswithdifferent landuse seems tobeuseful in reducing soil erosion. In each section, theupper part has a

high erosion rate and the lower part has a much lower erosion rate.Preventing soil erosion and collection of eroded soils under

silvicultural practices (with rates of 10e20 t ha�1 yr�1) is thedominant process in the Galikesh transect (Fig. 5, e), even if severesoil erosion (up to 40 t ha�1 yr�1) can be observed in the middle ofthe transect e. It is difficult to explain such a situation, but it iscertainly related to the complex land-use history as a result of ageneral alternation of clear-cutting and long-term dry-farmingbefore cultivation (Fig. 5, e).

3.5. Relations between soil erosion and soil organic carbon losses

The rates of soil loss provided by the 137Cs techniquemust be seenas the first important step to point out some of the on-site effectscaused by soil erosion. In this respect, loss of nutrients can beconsidered one of the most important consequences of land degra-dation that can be observed after land-use changes. FAO and IPCC(FAO, 2017; IPPC, 2019) have emphasized negative trends in landconditions, caused by direct or indirect human-induced processes

Fig. 5. Trends in soil redistribution along the transects (a) Eastward Dry-farming transects, Azadshahr, b) Westward Dry-farming transect, c) Reforested by Cypress and Olive treestransect Azadshahr, d) Dry-farming transect, Galikesh) and e) Reforested by even-aged Cypress trees in Galikesh.


401

including anthropogenic activities and climate change, expressed as along-term reduction or loss of at least one of the following aspects:biological productivity, ecological integrity or value to humans. Inforested areas, these aspects can be evaluated if the information onsoilorganic carbon isavailable.Normally, soil carbon is in steady-stateequilibrium in the natural forest, but as soon as deforestation (orafforestation) occurs, the equilibriumwill be affected.

Accordingly, the present study highlights the aspects of landdegradation in terms of loss of soil organic carbon (SOC) followed bythe conversion from forest to dry-farming land use.Measurements ofSOC allowed to explore the existing relation between loss of SOC andsoil erosion andprovided useful information on the status of soil aftera change in land use. Azadshahr and Galikesh reference sites pre-sented SOC values of 3.57 ± 1.1% and 465 3.7 ± 0.53%, respectively.According to the bulk density values (1400 kgm�3 and 1300 kgm�3),

sample thickness (30 cm), and coarse grains content (2% and 6.5%) ofthe reference samples, the carbon stock was calculated to be ca147 t ha�1 and 135 t ha�1 for original forests of the study area. TheAzadshahr and Galikesh dry-farming transects (Fig. 6 a, b, d) pre-sented SOC values of 1.24± 0.45% and 1.32 ± 0.25%, and 2.02± 0.62%,respectively, on average. Clear cutting of forest to develop dry-farming lands have contributed mainly to a carbon stock loss of ca.93.55 tha�1, 93.83 tha�1 inAzadshahr transects andca. 54.14 tha�1 inthe Galikesh transect over the last five decades. Afforestation,particularly on degraded soils with low organic matter contents, canbe an important way to mitigate this impact by long-term carbonsequestration both in biomass and in soil (FAO, 2017). In the GolestanProvince, for example, silvicultural practices through the cultivationof Cypress trees since 1993 in highly eroded parts of the hillside haveresulted in the restoration of SOC content by up to 100% effectiveness

Fig. 6. Correlation between the rate of soil and nutrient loss along the transects (a) Eastward Dry-farming transects, Azadshahr, b) Westward Dry-farming transect, c) Reforested byCypress and Olive trees transect Azadshahr, d) Dry-farming transect, Galikesh) and e) Reforested by even-aged Cypress trees in Galikesh.


402

in comparison to the adjacent dry-farming lands. Accordingly, thispractice has increased the SOC content from 1.4% to 2.82% whichcorresponds to a carbon stock restored from53.28 t ha�1 to 114 t ha�1.

The present study has highlighted the remarkable effectivenessof silvicultural practices in the Galikesh area (Fig. 6 e), where thecarbon stock increased to 143 t ha�1 by 5% in comparison to the SOCstock of the corresponding reference site (135 t ha�1).

In this respect, it is important to note the clear relation betweenrates of soil erosion and loss of SOC content under dry-farming landuse and silvicultural areas (Fig. 6). In transects a, b and d, the highvalue of R2 coefficients (R2 z 0.81, R2 z 0.70, and R2 z 0.58) (Fig. 6a, b, d) show a significant correlation between soil erosion and SOCloss along the representative transects under dry-farming land use.This is likely due to plowing and the consequent decompositionand oxidation of soil organic matter as compared to the originalforest where the soil remained undisturbed. Besides, in the dry-farming lands, a very small amount of biomass is left in the fieldsafter harvesting. In addition, a high R2 coefficient (R2 z 0.69) wasalso found between the SOC amount and 137Cs inventory in theafforested dry-farming in the Galikesh area. These data stressed thehigh affinity of 137Cs with SOC content and clay particles showing asimilar long-term fate (Fig. 7). Similar to the soil loss trend, anexpected downward decline in the rate of nutrient loss was foundalong all the studied transects.

4. Discussion

The new findings of the present research can be compared withthe results of similar studies carried out in Golestan Province andaround the world. Erodibility of loess deposits in the Golestan Prov-ince has been studied by several authors (see Hosseinalizadeh et al.,2018; Najafinejad et al., 2018; Jafari et al., 2015; Hosseinalizadehet al., 2018; and; Khormali & Ajami, 2011). A specific contributionbySeyed-Alipouretal. (2014)has reportedratesof soil erosion in loessdeposits using the 137Cs technique. These authors document a meanrate of soil loss equal to 10.78 t ha�1 yr�1 that is lower than thoseobtained in our study. However, because of the different samplingstrategyadopted (basedonrandomsampleswithout a specific trend),the reported rate is not comparable with the present research in

method and assumption. The mean rate of soil erosion for loess de-posits of the Golestan Province was estimated by Hosseinalizadehet al., 2018 using the WEPP model. He reported rates of soil lossequal to 27.26 t ha�1 yr�1 and 37.11 t ha�1 yr�1 for a buffer strip withandwithoutvegetationalongthehillsides,which is coherentwithourresults. A recent study by Poreba et al. (2019) documents rates of soilerosion in loess deposits of Poland using radionuclides (137Cs and210Pb). In that case, a similar approach to that used in our study hasbeenemployedand themeansoil erosion rates of 33.5 t ha�1 yr�1 and21.8 t ha�1 yr�1 have been calculated using Mass Balance I and II,respectively. The loess deposits in Poland have been exposed toagriculturebecauseof deforestationandare coherentwithour resultsin terms of geomorphic context and range of soil loss. Another pre-vious work aimed at estimating soil erosion on loess deposits usingthe 137Cs technique has been carried out by Pennockl et al. (1995) inCanada. Their results remarkably stressed the high sensitivity of theloess deposits to erosion with rates of soil loss ranging between20 t ha�1 yr�1 and 40 t ha�1 yr�1.

Additional contributions aimed at characterizing loess deposits oftheGolestanProvincehave alsobeenprovidedbyAyoubi et al. (2011);Jafari Honar et al. (2015); Khormali and Ajami (2011); Khormali et al.(2009);Kiani et al. (2007) in termsof effectsofdeforestation, land-usechanges and soil erosion features on soil properties. Specifically,Ayoubi et al. (2011) stated thatdeforestation-inducedcroplands led toa decrease in SOM by 71.5% over the last five decades. On the otherhand, this work highlighted the effectiveness of horticultural andsilvicultural practices through the cultivation of olive trees and Cy-press trees, which restored the SOM by about 49% and 72%, respec-tively. The overall results provided by Ayoubi et al. (2011) are in goodagreement with the findings of the present research, in which forestclearing followedby thecultivationof the loesshill slopesresulted inageneral decline of the soil quality attributes (SOC content) over 54years with a rate of soil loss equal to 93 t ha�1 in Azadshahr transectsand 54.14 t ha�1 in the Galikesh area. Similar information on soilproperties of deforestation-induced croplands in the Golestan Prov-ince is reported byKhormali et al. (2009). These authors document anincrease of silt particle size up to 65%, and a general decrease of theSOC content and clay particle size to 1% and 21%, respectively. Theseresults compare very well with the findings obtained in our study in

Fig. 7. Correlation between rate of soil and nutrient loss along the dry-farming land use, Azadshahr-West slope direction.


403

terms of the method applied (usage of transect approach), topo-graphical conditions (similar slope), driving forces, and emphasizethe negative effect of deforestation.

5. Conclusions

Soil redistribution following the conversion of original forests tocroplands and croplands to reforested lands in the East of the Hir-canian Forest has been estimated, for the first time in Iran, using the137Cs technique supervised by the IAEA. The research assumptionwas tested and the study objectives were achieved successfully. Theresearch points to almost 110,000 ha of hilly lands, ranging inelevation from 100 m to 500 m, having experienced deforestation,dating back to the 1960s when maximum fallout of radionuclideshas been received in original forests. Therefore, the elapsed time of54 yearswas assumedbetween theperiodofmaximumclear cutting(1963) and the date of sampling (2017). Agriculture, cattle ranching,and a variety of wood demands are the major driving forces ofdeforestation on loess deposits in the North East of Iran, which spedup soil and nutrients resources loss. Previous studies around theworld suggested the mean rate of soil erosion using radioisotopes(137Cs and 210Pb) inducing through deforestation on loess depositsranges from ca 20 t ha�1 yr�1 to 40 t ha�1 yr�1. The highest impactson land degradation appeared on the converted original forest todry-farming lands. Present research concludes that loess hilly slopesin theGolestan Province, North-East of Iran have been eroded by themean rate of 32.27 t ha�1 yr�1 over the last five decades. Under suchconditions, soil and nutrient losses account for ca. 25 cm of super-ficial soil with a rate of z2 mm yr�1. Also, resultant data demon-strated the loss of ca 93 t ha�1 of the carbon stock from croplandsover 54 years in comparison to theoriginalHircanian Forest. Further,the present study has highlighted the effectiveness of soil conser-vationpractices implemented since 1993 and 2004 (for olive groves)and has led to a successful decrease in soil erosion and nutrient loss.Cultivation of even-aged Cypress trees seems to be one of the bestsilvicultural methods for reducing soil erosion. Accordingly, resultssuggest a decline of soil loss from 16% to 60% in comparison withadjacent dry-farming lands, and up to 100% efficiency in the resto-ration of carbon stock in comparison with the original HircanianForests. Although the cultivation of olive trees documented theeffectiveness of only 13% (because of its low canopy cover) in soilerosion control, it has produced a general increase (ca. five times) inthe annual income of farmers. The research outcomes have recom-mended the application of radioisotope techniques as a reliablemethod in measuring some of the on-site impacts of deforestationand specific landuses especially on loess deposits. The present studyhas contributed greatly to highlighting the sensitivity of soil erosionin suchanecologically important regionof Iranandhasprovidedkeyinformation for Iranian and international decision-makers to fulfillsoil conservation practices.

Acknowledgments

We thank the International Atomic Energy Agency (IAEA) Viennafor technical and financial support under national TC project(IRA5013), Soil Conservation andWatershedManagement ResearchInstitute of Iran (SCWMRI) and Nuclear Science & TechnologyResearch Institute (NSTRI), Atomic energy organization of Iran.

References

Act41. (1962). Forest and rangelands nationalization law. Tehran: The nationalparliament of Iran.

Act71. (1992). The law on conservation and protection of natural resources and forestresources of the country. Tehran: The Parliament of Islamic Republic of Iran.

Ayoubi, S., Khormali, F., Sahrawat, K. L., & Rodrigues de Lima, A.c. (2011). Assessing

impacts of land use change on soil quality indicators in a loessial soil in golestanprovince, Iran. Journal of Agricultural Science and Technology A, 13, 727e742.

Bobek, H. (2005). Historical view of natural forests and Iran’s trees (Vol. 1). Tehran:Geographic Organization of the Armed Forces.

Evans, R., Collins, A. L., Zhang, Y., Foster, I. D. L., Boardman, J., Sint, H., & Griffith, B. A.(2017). A comparison of conventional and 137Cs-based estimates of soil erosionrates on arable and grassland across lowland England and Wales. Earth-ScienceReviews, 173, 49e64. https://doi.org/10.1016/j.earscirev.2017.08.005

FAO. (2015). Global forest resources assessment. Rome: Food and Agriculture Orga-nization of the United Nations.

FAO. (2017). Land degradation assessment in small island developing states (SIDS).Rome, Italy: Land and Water Division.

Fulajtar, M., Mabit, L., Renschler, C. S., & Lee Zhi Yi, A. (2017). Use of 137Cs for soilerosion assessment. Rome: Food and Agriculture Organization of the UnitedNations (Soil and Water Management & Crop Nutrition Section, Joint FAO/IAEADivision of Nuclear Techniques in Food and Agriculture, International AtomicEnergy Agency, Vienna, Austria).

Gharibreza, M., & Ashraf, M. A. (2014). Applied limnology. Tokyo, Heidelborg, NewYork, Dordrecht, London: Springer.

Gharibreza, M., Nasrollahi, A., Afshar, A., Amini, A., & Eisaei, H. (2018). Evolutionarytrend of the gorgan bay (southeastern Caspian Sea) during and post the lastCaspian Sea level rise. Catena, 166, 339e348. https://doi.org/10.1016/j.catena.2018.04.016

Gharibreza, M., Raj, J. K., Yusoff, I., Ashraf, M. A., Othman, Z., & Tahir, W. Z. W. M.(2013a). Effects of agricultural projects on nutrient levels in lake bera (tasekbera), peninsular Malaysia. Agriculture, Ecosystems & Environment, 165, 19e27.

Gharibreza, M., Raj, J. K., Yusoff, I., Ashraf, M. A., Othman, Z., & Tahir, W. Z. W. M.(2013b). Effects of agricultural projects on nutrient levels in lake bera (tasekbera), peninsular Malaysia. Agriculture, Ecosystems & Environment, 165, 19e27.https://doi.org/10.1016/j.agee.2012.12.004

Gharibreza, M., Raj, J. K., Yusoff, I., Othman, Z., Tahir, W. Z. W. M., & Ashraf, M. A.(2013c). Land use changes and soil redistribution estimation using 137Cs in thetropical Bera Lake catchment, Malaysia. Soil and Tillage Research, 131, 1e10.

Hiederer, R., & Martin, K. (2011). Global soil organic carbon estimates and theharmonized world soil database (vol. EUR 25225 EN - 2011 (p. 79). Italy: Instituefor environment sustainability, JRS European Commission.

Hosseinalizadeh, M., Kariminejad, N., Campetella, G., Jalalifard, A., & Alinejad, M.(2018). Spatial point pattern analysis of piping erosion in loess-derived soils inGolestan Province, Iran. Geoderma, 328, 20e29. https://doi.org/10.1016/j.geoderma.2018.04.029

IAEA. (1995). Use of nuclear techniques in studying soil erosion and siltation TECDOC-828 (Vol. 828). Vienna: International Atomic Energy Agency.

IAEA. (1998). Use of Cs137 in the study of soil erosion and sedimentation TECDOC-0828(Vol. 1028). Vienna: International Atomic Energy Agency.

Indarto, J., & Mutaqin, D. J. (2016). An overview of theoretical and empirical studieson deforestation. Journal of International Development and Cooperation, 22(1),107e120, 0.15027/39231.

Jafari Honar, A., Kiani, F., & Khormali, F. (2015). Effect of climates different onvariation of loess soil erodibility indices in golestan province. Journal of waterand soil conservation (journal of agricultural sciences and natural resources),22(1), 49e70.

Kaimowitz, D., & Angelsen, A. (1998). Economic models of tropical deforestation Areview. Bogor, Indonesia: Center for International Forestry Research.

Kakroodi, A., Kroonenberg, S., Hoogendoorn, R., Khani, H. M., Yamani, M.,Ghassemi, M., & Lahijani, H. (2012). Rapid Holocene sea-level changes along theIranian Caspian coast. Quaternary International, 263, 93e103.

Khormali, F., & Ajami, M. (2011). Pedogenetic investigation of soil degradation on adeforested loess hillslope of Golestan Province, Northern Iran. Geoderma,167e168, 274e283. https://doi.org/10.1016/j.geoderma.2011.07.030

Khormali, F., Ajami, M., Ayoubi, S., Srinivasarao, C., & Wani, S. P. (2009). Role ofdeforestation and hillslope position on soil quality attributes of loess-derivedsoils in Golestan province, Iran. Agriculture, Ecosystems & Environment, 134(3),178e189. https://doi.org/10.1016/j.agee.2009.06.017

Kiani, F., Jalalian, A., Pashaee, A., & Khademi, H. (2007). Effect of deforestation,grazing exclusion and rangeland degradation on soil quality indices in loess-derived landforms of golestan province. Journal of Water and Soil Science,11(41), 453e464.

Mabit, L., Benmansour, M., & Walling, D. W. (2008). Comparative advantages andlimitations of the fallout radionuclides 137Cs, 210Pbex and 7Be for assessing soilerosion and sedimentation. Journal of Environmental Radioactivity, 99,1799e1807.

Mohamed Abd Elbasit, A., Shekhar Prasad Ojha, C., Jinbai, H., Yasuda, H., Kimura, R.,& Ahmed, Z. (2013). Relationship between rainfall erodibility indicators underarid environments: Case of liudaogou basin in Chinese loess plateau. Journal ofFood Agriculture and Environment, 11(2), 1073e1077.

Muhs, D. R. (2007). Loess deposits, origins and properties quaternary stratigraphy:Lithostratigraphy. USA: U.S. Geological Survey, CO.

Pennockl, D. J., Lemmen, D. S., & de Jong, E. (1995). Cesium-l37 measured erosionrates for soils of five ;parent-material groups in southwestern Saskatchewan.Canadian Journal of Soil Science, 205e210.

Poreba, G. J. (2006). Cesium-137 as a soil erosion treacer: A review. Journal onmethods and applications of Absolut Chronology, 25, 37e46.

Poreba, G., �Snieszko, Z., Moska, P., Mroczek, P., & Malik, I. (2019). Interpretation ofsoil erosion in a Polish loess area using OSL, 137Cs, 210Pbex, dendrochronologyand micromorphology e case study: Biedrzykowice site (s Poland). https://doi.


404

org/10.1515/geochr-2015-0109, 46(1), 57.Rahimzadeh, N., Khormali, F., Gribenski, N., Tsukamoto, S., Kehl, M., Pint, A., &

Frechen, M. (2019). Timing and development of sand dunes in the golestanprovince, northern Irandimplications for the late-PleisSOCene history of theCaspian Sea. Aeolian Research, 41, 100538. https://doi.org/10.1016/j.aeolia.2019.07.004

Ritchie, J. C., Ritchie, C. A., Ritchie, J. C., & Ritchie, C. A. (2007). Bibliography ofpublications of 137Cesium studies related to erosion and sediment deposition.USDAeARS Hydrology and Remote Sensing Laboratory Occasional Paper HRSL-2007-01.

Robbins, J. A., Edgington, D. N., & Kemp, A. L. W. (1978). Comparative 210Pb, 137Cs,and pollen geochronologies of sediments from lakes ontario and erie. Quater-nary Research, 10(2), 256e278.

IPPC. (2019). In J. M. M. Rodriguez, & C. Vera (Eds.), Land degradation (p. 186). theUnited Nations: IPCC SRCCL.

Rogowski, A. S., & Tamura, T. (1970). Erosional behavior of cesium-137. HealthPhysics, 18, 467e477.

Seyed-Alipour, H., Feiznia, S., Ahmadi, H., Zare, M., & Hosseinalizadeh, M. (2014).Comparison of soil erosion by 137CS and RUSLE-3D for loess deposits North-East of Iran (study area: Aghemam catchment). Journal of Water and Soil Con-servation, 21(5), 27e47.

Smalley, I. J., & Smalley, V. (1983). Loess material and loess deposits: Formation,distribution and consequences. In M. E. Brookfield, & T. S. Ahlbrandt (Eds.),Developments in sedimentology (Vol. 38, pp. 51e68). Elsevier.

Tagami, K., Tsukada, H., & Uchida, S. (2019). Quantifying spatial distribution of 137Csin reference site soil in Asia. Catena, 180, 341e345. https://doi.org/10.1016/j.catena.2019.05.009

Tejaswi, G. (2007). Manual on deforestation, degradation, and fragmentation using

remote sensing and GIS. Rome: Forestry Department , Food and AgricultureOrganization of the United Nations.

Walkley, A. I., & Black, A. (1934). An examination of the degtjareff method fordetermining organic carbon in soils: Effect of variations in digestion conditionsand of inorganic soil constituents. Soil Science, 63, 251e263.

Walling, D. E. (1999). Linking land use, erosion and sediment yields in river basins.Hydrobiologia, 410, 223e240, 0.

Walling, D. E., & He, Q. (1999). Using fallout lead-210 measurements to estimate soilerosion on cultivated land. Soil Science Society of America Journal, 63, 1404e1412.

Walling, D. E., & Quine, T. A. (1990). Calibration of caesium-137 measurements toprovide quantitative erosion rate data. Land Degradation and Rehabilitation, 2,161e175.

Walling, D. E., Zhang, Y., & He, Q. (2007). Models for converting Measurements ofenvironmental radionuclide inventories (137Cs, excess 210Pb, and 7Be) to Estimatesof soil Erosion and deposition rates (including Software for model implementation).Exeter: University of Exeter.

Walling, D. E., Zhang, Y., & He, Q. (2014). Conversion models and related software. InGuidelines for using fallout radionuclides to assess erosion and effectiveness of soilconservation strategies (pp. 125e148). Vienna: IAEA. Vol. IAEA-TECDOC-1741.

Yamagata, N., Matsuda, S., & Kodaira, K. (1963). Run-off of caesium-137 ansstrontium-90 from rivers. Nature, 200, 668e669.

Yan, R., Zhang, X., Yan, S., & Chen, H. (2018). Estimating soil erosion response to landuse/cover change in a catchment of the Loess Plateau, China. International Soiland Water Conservation Research, 6(1), 13e22. https://doi.org/10.1016/j.iswcr.2017.12.002

Zapata, F., & Garcia-Agudo, E. (2000). Future prospects for the 137Cs technique forestimating soil erosion and sedimentation rates. Acta Geologica Hispanica,35(3e4), 197e205.


405


Integrated nuclear techniques for sedimentation assessment in LatinAmerican region

Jos�e Luis Peralta Vital*, Reinaldo Honorio Gil Castillo, Yanna Llerena Padr�on,Yusleidy Milagro Cordovi Miranda, Naymi Labrada Arevalo, Leroy Alonso PinoEnvironmental Impact Assessment Group, Center of Radiation Protection and Hygiene, La Habana, Zip Code 10600, Cuba


Article history:Received 11 March 2020Received in revised form5 June 2020Accepted 10 June 2020Available online 17 June 2020

Keywords:Nuclear techniqueSedimentationWater reservoirSoilLandHydric resources

a b s t r a c t

The paper shows the novel techniques introduction in the region of Latin America and the Caribbean inorder to strengthen the surveillance and monitoring of the sedimentation phenomenon in the surfacewater reservoirs. The integrated use of the nuclear techniques allows a more complete assessment ofwater and land sustainability. With the development of a regional project, supported by the InternationalAgency of Atomic Energy, Cuba participates as technical leader for 15 countries, to encourage andimplement the integrated application of 3 nuclear techniques (Fallout Radionuclides (137Cs, 7Be, 210Pb),Compound Specific Stable Isotopes (fatty acids of the v13C) and the Isotopic Hydrology (18O, 2H, 3H). Themethodology used in the last studies referred to RLA 5076 Project is to apply the integration of thesethree techniques allows evaluating, in the hydraulic facilities and surface water reservoirs, the negativesedimentation impacts in natural and anthropic process as environmental and social risk. These tech-niques evaluate the sedimentation process and its synergy, obtaining since the soil quantification in thelandscape, definition the soil origin deposited until the dynamic characterization of the water body assoil receptor. The integrated application of these nuclear techniques in the Project RLA-5076 confirms itas a useful tool to support the decisions makers in the definition of the national strategies and programsrelated to the land and the water resources sustainability. With the support of the IAEA projects in theLatin American and Caribbean region, has been evidenced an important increment in the use of thesenovel techniques.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction, scope and main objectives

The introduction and integration of three nuclear techniques isshown in order to strengthen in the Latin American and Caribbeanregion, the surveillance and monitoring of the negative sedimen-tation phenomenon in the surface water reservoirs (natural andartificial). Better strategies and national’s programs for assessmentthe sedimentation. (Dape~na et al., 2010, 2013), (Díaz et al., 2017),(Gil et al., 2015), (Peralta Vital et al., 2003, 2011, 2013, 2015),(Peralta , 2011), (Schuller et al., 2011).

The development of several regional projects support by theInternational Atomic Energy Agency (IAEA), has allowed a sub-stantial advance in the nuclear techniques uses, as technical tools

that contribute to the evaluation of the water, soil and sediment.These tools allow to decision makers, to definition of better stra-tegies and national environmental programs that assure finally theland and the water resources sustainable. Among other importantregional projects (ARCAL) developed in the last 10 years, we hadtwo projects that have been important to encourage the nucleartechniques use in the Latin American and Caribbean region.

These two projects support the development of the knowledgeon the nuclear techniques (Fallout radionuclide and Compoundspecific stable isotopes) to evaluate the soil redistribution in thelandscape and the On-site and Off-site impacts of thesedimentation.

- Project RLA-5051 (2009e2013). “Using environmental radionu-clides as indicator of land degradation in Latin American,Caribbean and Antarctic ecosystems” (ARCAL C); with theparticipation of these countries (Argentina, Bolivia, Brazil, Chile,

* Corresponding author.E-mail addresses: [email protected], [email protected],

[email protected] (J.L. Peralta Vital).






Cuba, Dominican Republic, El Salvador, Haiti, Jamaica, Mexico,Nicaragua, Peru, Uruguay, Venezuela and IAEA, Spain). 15countries. (IAEA Project RLA-5051, (2009e2013).

- Project RLA-5064 (2014e2016). “Strengthening Soil and WaterConservation Strategies at the Landscape Level by Using Inno-vative Radio and Stable Isotope and Related Techniques” (ARCALCXL); with the participation of these countries (Argentina,Bolivia, Brazil, Chile, Colombia, Costa Rica, Cuba, DominicanRepublic, Ecuador, El Salvador, Guatemala, Haiti, Jamaica,M�exico, Nicaragua, Panam�a, Paraguay, Peru, Uruguay, Venezuelaand IAEA). 21 countries. (IAEA Project RLA-5064, 2014e2016).

Both projects were relevant to initiate the application of nucleartechniques for soil and sediment assessment in the region. Despitethe assessment of water resources by isotopic hydrology were notincluded in the project scope; they showed the advantages to applythese new techniques as assessment tools that contribute withtheir results to support the sustainable development.

Assisting to the results obtained in the region during theimplementation and use of these techniques and the acquiredknowledge about it, the countries advance to more complex in-vestigations at Basin scales. There is the need to study at thelandscape scale, the sedimentation phenomenon that degrade theland and the water resources and for this reason Cuba intends theintegrated use of 3 nuclear techniques to evaluate the negativeimpacts of this phenomenon in surface water reservoirs (naturaland artificial). Responding to this problem andwith the purposes toimprove the results with the integrated approach was proposed in2017 and starting from the year 2018, the new ARCAL project.

- RLA-5076 (2018e2020), belongs to the cycle 2018e2019“Strengthening Surveillance Systems and Monitoring Programsof Hydraulic Facilities Nuclear Using Techniques to AssessSedimentation Impacts as Environmental Social and Risks”(ARCAL CLV). 15 countries. (IAEA Project RLA-5076, 2018e2020).

With the development of this regional project, supported by theInternational Atomic Energy Agency (IAEA), Cuba acts as technicaldirector with 15 countries participants (Argentina, Bolivia, Brazil,Chile, Colombia, CUBA, Honduras, Mexico, Nicaragua, Panama,Paraguay, Peru, Dominican Republic, Uruguay and Venezuela).

The important general goal is to encourage and to implementthe integrated application of three nuclear techniques (Fallout ra-dionuclides (137Cs, 7Be and 210Pb), Compound specific of stableisotopes (fatty acids of the v13C) and the Isotopic Hydrology (18O,2H and 3H). These techniques confirm its validation as effectivetechnical tool to support the strategies definition and nationalprograms linked to support the sustainability the land and thewater resources.

In this ongoing project, the countries of the region incorporatethe necessary knowledge on these 3 nuclear techniques useful tocharacterize the dynamics of the dam to evaluate the sedimenta-tion impact in the surface water reservoirs (IAEA Project RLA-5076,2018e2020).


2.1. Methodology

The integrated application of these 3 techniques allows evalu-ating, in the hydraulic facilities and surface water reservoirs (nat-ural or artificial), the negative impacts of the sedimentation innatural and anthropic process, as environmental and social risk(Iurian et al., 2013), (Mabit et al., 2008, 2011, 2013, 2014),(Meusburger et al., 2013), (Taylor et al., 2013). Each technique

assess the whole process and its synergy, contributing results thatexpose from the quantification of the soil redistribution in thelandscape, the exact definition of the soil origin deposited until thelater dynamics characterization of the body water as final receptorof that soil moves.

Each one of these three nuclear techniques exposes particularresults, which show a synergy in the process of the phenomenonassessment:

- FRN (Fallout Radionuclides): Uses the 7Be; 137Cs and 210Pb, toevaluate the soil redistribution to landscape scale and other scales.It quantifies the redistribution of the soil in a study region, allowingdetermining areas of lost (erosion) and deposition areas (sedi-mentation) in the landscape.

- CSSI (compound specific of the stable isotopes): Uses the fattyacids of the Carbon chain to evaluate the origin of the soil depos-ited, allowing determining with great precision and accuracy, thesediment origin transported and deposited.

- Isotopic Hydrology: Uses the 3H; 2H and 18O, to evaluate thewater dynamics (surface and undergroundwater) in a study region.It allows determining several technical aspects, such as the waterorigin, to identify recharge areas, the relationship surface-waterand underground-water, the aquifers vulnerability to the pollu-tion and the saline intrusion, to classify the sources of renewableand not renewable waters, to identify water mixtures, the waterdating, residence times of water, etc.

3. Results

The integrated application of these techniques is a fundamentaltool to support the sedimentation impact assessment, studying thisphenomenon as a process, each nuclear technique for itself and itssynergy; provides relevant results at the stages in this process.

The process of soil redistribution in the landscape includes, thesoil erosion its transport and the final deposition, when this sedi-mentation occurs in a dam such as a water reservoir, these nucleartechniques due its technical benefits, can be applied in an inte-grated way to evaluate their impacts efficiently. The diagram showsin Fig. 1, the integration of nuclear techniques to apply in each stageof the process.

3.1. Synergy and integration of these three nuclear techniques

The nuclear technique “FRN”, quantifies and identifies the areaswith soil loss (-) and soil gains (þ) in the study region. The “CSSI”technique, identifies possible areas of soil sources (areas -) andpossible mixture areas (areas þ), also contributing potential pointslike hot spot. This result allows to the CSSI, a first approach of thesampling points to collect to evaluate the sediment origin (depos-ited soil). Once quantity the soil redistribution and certain theorigin of the soil deposited, the nuclear technique named “IsotopicHydrology”, plays their role in clarifying the dynamics of the waterbody where appear the soil deposit.

The integrated use of these three nuclear techniques, with theirapplication methodologies, allows finally determining the sedi-ment impacts in surface water reservoirs (natural and artificial).

4. Discussion

The processes of soil redistribution occur to different scales inthe landscape and it is extensive at Basin levels. This redistributionalways has three stages or moments: firstly, the soil degradationappears, after that, the soil moves and then deposited. These stagesin the soil redistribution, can be evaluated by using the integrateduse of the three nuclear techniques (FRN, CSSI and isotopic hy-drology), which describe each stage, analyzing the erosion process

J.L. Peralta Vital et al. / International Soil and Water Conservation Research 8 (2020) 406e409 407

in the areas where the soil loss and deposition processes occur.When the soil transported is accumulated in water surface reser-voirs, then the third nuclear technique (isotopic hydrology), allowscharacterizing the dynamics of this water (for example, dams, lakes,etc.) and this way to clarify the impact of this sedimentationphenomenon.

To evaluate the sedimentation phenomenon in surface waterreservoir, applying these 3 nuclear techniques in an integrated way,we recommend to carry out the interpretation of the results,analyzing the phenomenon like a process, in which the soil isdegraded, after is transported, and it is redistributed in the land-scape. There are areas where the sediment and other areas of soilloss (the soil redistribution rate is assessed with the application ofthe first nuclear technique (FRN)). If these areas are showed in amap, where a sign (-) are those that lose soil and a sign (þ) are theareas where it is accumulated, then it has preliminary informationto the second nuclear technique (CSSI). The areas that contributesoil, which are called “potential sources” (-) and showing the areasthat receive soil, called “mixtures” (þ); keeping in mind it, theorigin of the soil that arrives to the area (þ) is evaluated.

4.1. Important note

In all study applying the integrated use of these nuclear tech-niques, is strongly recommended to take into account the differ-ences that could exist in the temporary scale of the events that weare analyzing; keeping in mind this, is possible to evaluate theorigin of the soil that arrives to the area (þ).

If this soil redistribution in the landscape evidences soil accu-mulated in a water reservoir, then the results obtained are datainputs for the third nuclear technique (isotopic hydrology) inte-grated in order to characterize this water reservoir and so to ach-ieve, finally, to assess the sedimentation impact in this reservoir.The soil accumulation in surface water reservoir, evidence the“sedimentation” phenomenon and the impacts are always nega-tive. The integrated use of these three nuclear techniques (FRN,CSSI, isotopic Hydrology), allows to evaluate the whole impact,responding the uncertainties of the described process stage.

5. Conclusions

1 The new Project RLA-5076 (2018e2020). “Strengthening Sur-veillance Systems and Monitoring Programs of Hydraulic Facil-ities Nuclear Using Techniques to Assess Sedimentation Impactsas Environmental Social and Risks” (ARCAL CLV), allows to 15countries of the region the integrated application of three nu-clear techniques. In addition, the project has a positive impact inthe strengthening in the Latin American and Caribbean region,the strategies and programs of surveillance/monitoring of thesedimentation phenomenon in surface water reservoirs.

2 In the project RLA-5076, the participant countries of the regiontake whole knowledge Isotopic Hydrology, like very useful tool

to characterize the dynamics of the dam to be able to assess theimpact on the surface water reservoir.

3 The integrated application of these nuclear techniques (FalloutRadionuclides (137Cs, 7Be and 210Pb), Compound specific of sta-ble isotopes (fatty acids of the v13C and the Isotopic Hydrology(18O, 2H and 3H) are validated as effective technical tools for thedecisions makers in the definition of strategies and nationalprograms linked to achieve the land and the water resourcesmore sustainable.

4 The integrated application of these three nuclear techniquesconfirm the validation as effective technical tools for the impactassessment of the sedimentation phenomenon in the waterreservoirs. This technical tools support the definition of thestrategies and programs of surveillance/monitoring of thesedimentation phenomenon in water reservoir superficial.


All authors of the paper confirm and declare that there are notactual or potential conflict of interest including any financial, per-sonal or other relationships with other people or organizations thatcould be inappropriately influence, or be perceived to influence,this work.

Acknowledgements

The International Agency of Atomic Energy (IAEA). The supportreceived in Latin American and Caribbean, is fundamental todevelop the nuclear techniques application like important assess-ment tools in the environment degradations.

References

Dape~na, C., Panarello, H. O., Peralta, J. L., & Gil Castillo, R. H. (2010). Environmentalisotopes in a coastal karstic aquifer, Cuba preliminary report. VII south Americansymposium on isotope geology. CD Paper S0647:487-491. Brasilia, Brasil.

Dape~na, C., Peralta, J. L., Panarello, H. O., Castillo, R. G., & Bombouse, D. L. (2013).Isotope and hydrochemical study of seawater intrusion into the aquifers of acoastal zone. Cuba. International Symposium on Isotopes in Hydrology. InMarine ecosystems, and climate change studies (Vol. 1, pp. 349e356). Vienna,Austria: IAEA (CN-186) IAEA CN-082.

Díaz, A. M., Corcho, J. A., Cartas, H., & Pulido-Caraball, A. (2017). 210Pb and 137Cs astracers of recent sedimentary processes in two water reservoirs in Cuba. Journalof Environmental Radioactivity, 177, 290e304, 2017.

Gil, R. H., Peralta, J. L., Carrazana, J., Riverol, M., & Aguilar, Y. (2015). Utilizaci�on det�ecnicas nucleares para estimar la erosi�on hídrica en plantaciones de tabaco enCuba. Revista Cultivos Tropicales, 36(4), 7e13. ISSN 1819-4087.

IAEA. (2009-2013). Project RLA-5051. Using environmental radionuclides as indicatorof land degradation in Latin American, Caribbean and Antarctic ecosystems. ARCALC).

IAEA Project. (2014-2016). RLA-5064. Strengthening soil and water conservationstrategies at the landscape Level by using innovative Radio and stable isotope andrelated techniques. ARCAL CXL).

IAEA. Project RLA-5076. (2018-2020). Strengthening surveillance Systems and moni-toring programmes of hydraulic facilities nuclear using techniques to assess sedi-mentation impacts ace environmental social and risks. ARCAL CLV).

Iurian, A. R., Mabit, L., Begy, R., & Cosma, C. (2013). Comparative assessment oferosion and deposition rates on cultivated land in the Transylvanian Plain of

Fig. 1. Integration of the nuclear techniques to apply in each stage of the process.

J.L. Peralta Vital et al. / International Soil and Water Conservation Research 8 (2020) 406e409408

Romania using 137Cs and 210Pbex. Journal of Environmental Radioactivity, 125,40e49.

Mabit, L., Benmansour, M., & Walling, D. E. (2008). Comparative advantages andlimitations of Fallout radionuclides (137Cs, 210Pb and 7Be) to assess soil erosionand sedimentation. Journal of Environmental Radioactivity, 99(12), 1799e1807.

Mabit, L., Fulajtar, E., & Alewell, C. (2011). The usefulness of 137Cs as a tracer for soilerosion assessment: A critical reply to parsons and foster. Earth-Science Reviews,127, 300e307.

Mabit, L., Meusburger, K., Iurian, A. R., Owens, P. N., Toloza, A., & Alewell, C. (2014).Sampling soil and sediment depth profiles at a fine-resolution with a newdevice for determining physical, chemical and biological properties: The FineIncrement Soil Collector (FISC). Journal of Soils and Sediments, 14(3), 630e636.

Meusburger, K., Mabit, L., & Park. (2013). Combined use of stable isotopes and FRNsas soil erosion indicators in a forested mountain site, South Korea. Bio-geosciences, 10, 5627e5638.

Peralta, J. L. (2011). Radioisotope monitoring network designed for the best hydrody-namic characterization of a karstic basin”. International symposium on isotopes inhydrology, marine ecosystems, and climate change studies oceanographic museum,Monaco.

Peralta Vital, J. L., Gil Castillo, R., Dape~na, C., & Gonz�ales, L. V. (2015). Hidrologíaisot�opica, herramienta nuclear para la gesti�on sostenible del recurso hídrico.Revista Ingeniería Hidraúlica y Ambiental, XXXVI, 1680-2338.

Peralta Vital, J. L., Gil Castillo, R., Le�on, L. F. M., Panarello, O. H., Dape~na, C., &

Gonz�ales, L. V. (2011). Cuban experience in the applications of the nuclear tech-niques as valuable evaluations tools for the sustainable management water re-sources. International Symposium on Isotopes in Hydrology, Marine Ecosystems,and Climate Change studies oceanographic museum, Monaco.

Peralta Vital, J. L., Gil Castillo, R., Molerio Le�on, I. F., Dape~na, C., & Acosta, J. (2013).Improving a radioisotope monitoring network for the hydrodynamic charac-terization of a karstic basin International Symposium on Isotopes in Hydrology.In , Vol. 1. Marine ecosystems, and climate change studies (pp. 263e272). Vienna,Austria: IAEA (CN-186) IAEA CN-232.

Peralta Vital, J. L., Gil Castillo, R., Molerio Le�on, I., Leyva Bombuse, D., Dape~na, C.,Panarello, H. O., Vera, M. C., & Ibarra, E. D. (2003). Isotope hydrology applicationin Cuba for assessment of water resource management in the most importantbasin of Havana city. Advances in isotope hydrology and its role in sustainablewater resources management (HIS-2007). CN151-3. In , 2. IAEA Proceedings (pp.245e253). Vienna, Austria.

Schuller, P., Walling, D. E., Iroum�e, A., Quilodr�an, C., Castillo, A., & Navas, A. (2011).Using 137Cs and 210Pbex and other sediment source fingerprints to documentsuspended sediment sources in small forested catchments in south-centralChile. Journal of Environmental Radioactivity.

Taylor, A., Blake, W. H., Smith, H. G., Mabit, L., & Keith-Roach, M. J. (2013). As-sumptions and challenges in the use of fallout beryllium-7 as a soil and sedi-ment tracer in river basins. Earth-Science Reviews, 126, 85e95.

J.L. Peralta Vital et al. / International Soil and Water Conservation Research 8 (2020) 406e409 409


Use of geochemical fingerprints to trace sediment sources in anagricultural catchment of Argentina

Romina Torres Astorga a, Yanina Garcias a, Gisela Borgatello a, Hugo Velasco a, *,Rom�an Padilla b, Gerd Dercon c, Lionel Mabit c

a IMASL, UNSL, CONICET. Ej, de los Andes 950, San Luis, Argentinab Nuclear Science and Instrumentation Laboratory, Division of Physical and Chemical Sciences, IAEA, Seibersdorf, Austriac Soil and Water Management & Crop Nutrition Laboratory, Joint FAO/IAEA Division of Nuclear Techniques in Food and Agriculture, Seibersdorf, Austria


Article history:Received 12 March 2020Received in revised form25 September 2020Accepted 19 October 2020Available online 29 October 2020

Keywords:FingerprintingGeochemical elementsEnergy dispersive X-ray fluorescenceSoil erosionMixing models

a b s t r a c t

Soil erosion and associated sediment redistribution are key environmental problems in CentralArgentina. Specific land uses and management practices, such as intensive grazing and crop cultivation,are considered to be significantly driving and accelerating these processes. This research focuses on theidentification of suitable soil tracers from hot spots of land degradation and sediment fate in an agri-cultural catchment of central Argentina with erodible loess soils. Using Energy Dispersive X-Ray Fluo-rescence (EDXRF), elemental concentrations were determined and later used as soil tracers forgeochemical characterization. The best set of tracers were identified using two artificial mixturescomposed of known proportions of soil sources collected from different lands having contrasting soiluses. Barium, calcium, iron, phosphorus, and titanium were identified for obtaining the best suitablereconstruction of source proportions in the laboratory-prepared artificial mixtures. Then, these elements,as well as the total organic carbon, were applied for pinpointing critical hot spots of erosion within thestudied catchment. Feedlots were identified to be the main source of sediments, river banks and dirtroads together are the second most important source. This investigation provides key information foroptimizing soil conservation strategies and selecting land management practices and land uses which donot generate great contribution of sediment, preventing pollution of the waterways of the region.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction

In the Dry Chaco, South America, one of the main semiaridwoodlands on Earth, the native vegetation is quickly being con-verted to pastures and croplands (Baldi & Jobb�agy, 2012; le Polainde Waroux et al., 2016). This region has been depleted as aconsequence of the global increase in food demand and theincorporation of new agricultural technologies, among other fac-tors, over the last 30 years. The depletion rate in native woodlandsreaches 2.2%, on average, per year (Gasparri & Grau, 2009; Zaket al., 2004).

Additionally, in the southern limit of this region, in centralArgentina, the agricultural frontier continues to expand from thewet Pampas towards arid and semi-arid environments despite

higher water shortage. In many cases, land practices adoptsimilar agricultural strategies to those applied in the more humidregions, which increase the risk of environmental deterioration(Viglizzo et al., 2011). After decades of changes in the use of land,suitable indicators of the impact of these practices on the soilstatus and water quality are necessary. These indicators shouldprovide reliable information for effective decision making thatcould lead to sustainable development. In this way, it could bepossible to contribute to reduce the existing tensions betweenproductive development and environmental protection. Themagnitude of soil erosion is one of the most evident indicators ofenvironment degradation. In this region with high vulnerability,erosion can significantly increase by inappropriate land usemanagement, which results in a reduced productivity of crop-lands and increases pollution of streams, rivers and water res-ervoirs. Effective strategies for controlling the increased flux ofsediments are necessary to determine both the nature and the* Corresponding author.

E-mail address: [email protected] (H. Velasco).






region of the main sources of sediments at catchment scale andits relationship with the land uses.

The method o f geochemical fingerprinting has been widelyused to determine sediment provenance (Hardy et al., 2010).Elemental concentrations in the areas where the sediments origi-nate and accumulate allow us to identify and to quantify the rela-tive contribution of different sources. These concentrations mainlydepend on the soil type, the geological substratum and the landuses fromwhich they originate (Blake et al., 2012). Applying mixingmodels (MM) allows us to derive the relative contributions of eachsource to the sediment mixtures in the fate places.

In this paper, we applied a geochemical fingerprint approach tocharacterize the temporal sediment apportionment in a smallbasin located in the Province of San Luis, in the central region ofArgentina. In this relatively small mountain catchment, differentland uses have been incorporated at the expense of native vege-tation since the 1960s, with greater intensity in the last 20 years.Soils are currently being used for agriculture (no tillage croprotation), livestock, and some fields used to exploit fruit trees.Original vegetation occupies important extensions of the region.To evaluate the impact of different land uses in the sedimentcontribution, the source samples were collected in the region of ElDurazno sub-catchment, where loessoid material soils are domi-nant (i.e. Quaternary deposits). Thus, no differences in lithologieswere studied.

In a previous study (Torres Astorga et al., 2018) on the sameregion, a set of 5 geochemical elements were identified as finger-prints to reconstruct the proportion of two artificial mixture sam-ples (MIX 1 and MIX 2). These were prepared in a laboratorycombining 4 sedimentary sources from different sectors of thecatchment. In the reconstruction processes of the sources propor-tion two mixing models were independently used: CSSIAR v2.00(de los Santos-Villalobos et al., 2017) and IsoSource (Phillips &Gregg, 2003). In both cases a very good approximation was ob-tained. The selected elements were: Ba, Ca, Fe, P and Ti.

In this second part of the investigation, we used the 5 finger-prints identified from the artificial mixtures analysis. The totalorganic carbon (TOC) was added as a sixth tracer in order toimprove the accuracy of the results without changing the resultingproportions.

The main objective of this research is to use the previouslyselected geochemical fingerprint elements and TOC to describe thetemporal sediment apportionment in different locations in thehydrographic network that includes streams of stationary char-acter, rivers with very variable flows and artificial bodies of waterthat serve for their regulation. The results achieved are highlyrelevant. They make it possible to have specific indicators of theimpact ofthe new productive uses of the land on natural resourcesin a highly vulnerable region.

2. Study area

The selected study site is El Durazno Sub-catchment (previouslycalled Estancia Grande Sub-catchment), covering 1235 ha, which islocated in the centre of Argentina 23 km north east of San Luis city(S 33� 100; W 66� 080) at 1100 m.a.s.l. The studied sub-catchment ispart of theVolc�an river Sub-catchment (Fig. 1). The Volc�an riverSub-catchment presents 5 different lithological units: granites,gneisses, micaschists, mafic and ultramafic rocks, and quaternarydeposits (Morosini et al., 2017). The mean annual temperature is17 �C. In summer (December to March), the mean temperature is23 �C. Annual rainfall ranges from 600 to 800 mm, with a tendencyto increase. At the same time, the frequency of extreme rainfallevents during the last decades has been amplified (de la Casa &Ovando, 2014; Penalba & Vargas, 2004). Precipitation regimes

vary seasonally, following a normal distribution (maximum rainfallin midsummer). The rainy season is from November to April, andthe dry season is fromMay to October, with almost no precipitation,but some occasional drizzles. The studied sub-catchment is char-acterized by highly erodible Eutric Fluvisol soils. These soils origi-nated from silty sand material and possess a high level of organicmatter in their upper 25 cm. The studied catchment belongs to theloess belt of North East Argentina (Teruggi, 1957). There is no rockyoutcrop in the studied area, but rather secondary loessoid deposits.Fig. 1 displays that El Durazno sub-catchment has 2 different lith-ological units, mainly Quaternary deposits and a small portionwithGneiss. The soil is composed of silt-sandy materials of river reworkorigin (Torres Astorga et al., 2018).

The region is currently being used mainly for agriculture(rotation of 2 or 3 crops). Livestock (rangelands, pastures andfeedlots) is another important land use. Likewise, some of theagricultural fields are used for growing nut orchards (walnuts andalmonds) (Fig. 2). Furthermore, native vegetation is found in be-tween the agriculture lands and in the upper part of the sub-catchment. In relation to agriculture and its soil managementpractices, since early 2000 the local farmers have been using adirect seeding mulch-based system as the main practice for theircrop cultivation. This system has increased cultivation yields andreduced soil erosion. Crops include soybean, maize, and wheat.Direct seeding mulch-based system requires the use of herbicidesand fertilizers. The most applied herbicides are atrazine andglyphosate. Nevertheless, fertilizers are not used with the sameconsistency on all the agricultural fields. Some of the mostwidely used ones are NePeK-based fertilizers, such as ureaammonium nitrate (UAN) 32-0-0, monoammonium phosphate(MAP) 11-52-0, triple superphosphate 0-46-0; and biologicalgrowth promoters which depend on the species of the cultivatedcrop. Feedlot cattle are fed with different grain: maize, oats,sunflower meal, and grazing hay. Besides, the cattle are admin-istered mineral supplements of calcium, magnesium, phosphorusand sodium chloride.


3.1. Sampling

The sampling methodology included removing leaves andother organic material found in the place; and collecting a soillayer of 100 cm2, 2 cm thick of exposed soil with a stainless-steelflat spatula. At each sampling area, 5 to 6 subsamples from asurface of about 150e200 m2 were collected in a plastic containerto put together a composite sample representative of that landuse, i.e. the source sample. Sediment mixture samples (from theriver courses) were collected at the top 2 cm of the accumulationzones on small floodplains where deposition of sediments wasobserved. Photos of different land uses are shown in Fig. 3. For theriver bank source the samples were collected scraping verticallythe wall of them, as most of themwere wide high and along 10 m(or less for shorter extensions), Fig. 3(f). Sediment mixture com-posite samples were collected in smaller areas, maximum 5 mapart in between the subsamples at 5 sites along the channel asobserved in Fig. 3(h).

The sediment samples (mixtures) were taken during threedifferent periods: (I) at the end of the rainy season (II) at the end ofthe dry season, and (III) at middle of the rainy season. The locationof the sediment sampling points is presented in Fig. 2. In the firstperiod (I) sediment sample collection in the northern part of theriver (M4 and M5) was not possible.

A total of 71 samples were collected from source soils andmixture sediments. The number of samples was chosen based on

R. Torres Astorga, Y. Garcias, G. Borgatello et al. International Soil and Water Conservation Research 8 (2020) 410e417

411

the dimension of each land use.

3.2. Analytical methods

3.2.1. Samples of artificial mixtures: fingerprints identificationIn the first stage, a group of samples were used to create artificial

mixtures for fingerprint selection. This samples were first dried at50 �C, later disaggregated and sieved through a 2-mm sieve at theGEA-IMASL Laboratory.

To perform an EDXRF spectrometry analysis, the samples wereground into fine powder, and pressed pellets of 25 mm diameterand 2.5 g weight were shaped. The pellets were measured at theIAEA Nuclear Science and Instrumentation Laboratory using asoftware-controlled EDXRF spectrometer which uses five second-ary targets for accuracy improvement (SPECTRO X-LAB 2000). Thespectral fitting and following quantification was carried out usingthe software and instrument calibration supplied by the manu-facturer for the analysis of geological samples. The accuracy of theresults was verified by analysing three different reference mate-rials: IAEA-Soil7 (soil sample), IAEA-433 (marine sediment) andIAEA-405 (sediment). The concentration of 51 elements for eachsample was obtained from them. Eight elements showed concen-tration levels below detection limits for some samples and werediscarded.

As described in a previous publication (Torres Astorga et al.,2018), suitable fingerprinting elements were identified from thetwo artificial mixed samples (MIX 1 and MIX 2). They werecomposed using known source samples following the below

proportions:

MIX 1 ¼ 0.40 S3 þ 0.25 S2 þ 0.25 S4 þ 0.10 S1MIX 2 ¼ 0.45 S2 þ 0.32 S4 þ 0.20 S3 þ 0.03 S1

where S1 represents the source riverbank. S2 and S4 were two soilsamples collected in agricultural fields with crop rotation. Duringthe sampling campaign, one of these sources was under corn andthe other one with soybean cultivation. Source S3 corresponds tosoil samples collected in a feedlot. The proportions were chosen torepresent possible distributions of sediment origins and, at thesame time, to ensure that the mixing model yields results outsidethe uncertainty margins.

A three-stepped procedure was applied for fingerprint selec-tion: 1) Kruskal Wallis H test was performed to dismiss fingerprintproperties that were redundant. This test is a nonparametric pro-cedure equivalent to analysis of variance (ANOVA); 2) DiscriminantFunction Analysis (DFA) was used to test the power of the proper-ties that passed the former test to classify the source samples intothe correct groups and 3) Bi-plots exam that consists of a visualinspection of bi-plots of the elements that were previously selected.All possible combinations of element pairs were plotted. Then, weconsidered the fact that if a mixture lies outside the polygon shapedby the sources, one or both of the element pairs should not be used(Torres Astorga et al., 2018).

The resulting elements were validated using the two artificialmixtures in two MMs: CSSIAR v2.00 (de los Santos-Villalobos et al.,2017) and IsoSource (Phillips & Gregg, 2003). In Fig. 4 it can be

Fig. 1. Location of the Volc�an sub-catchment in San Luis, Argentina and geological map of this sub-catchment presenting 5 lithological units. Study area is highlighted with slantedblack lines.


412

observed the difference in the use of these two MM in the artificialmixtures. After validation, CSSIAR v2.00 was then applied foridentifying critical hot spots of erosion using the selectedgeochemical elements and TOC data as fingerprints in the collectedmixture samples from the studied sub-catchment. An uninforma-tive prior was used, as there was no previous information of thesediment transport related with the land uses in the region.

3.2.2. El Durazno Sub-catchment sample analysisThe sampling was extended to the complete region of study.

The total amount of samples (sources and mixtures) were ana-lysed for total organic carbon (TOC) and EDXRF spectrometry. TOCwas measured to assess its addition to the set of fingerprintproperties. The TOC value of each sample was determined at theIAEA Soil and Water Management & Crop Nutrition Laboratory.Soil samples were dried at 40 �C, carbonates were removed by acidfumigationwith hydrochloric acid and TOC content was measuredusing an elemental analyser (Vario Isotope Select, Elementar,Langenselbold, Germany) coupled to an isotope ratio mass spec-trometer (Isoprime 100, Elementar UK, Manchester, UnitedKingdom) followingmethods proposed by (Harris et al., 2001) and(IAEA, 2019).

To perform the EDXRF spectrometry analysis, the samples wereprepared as explained above (in Samples of artificial mixtures:fingerprints identification). The pellets were measured at the IAEANuclear Science and Instrumentation Laboratory using a software-controlled EDXRF spectrometer which uses different secondarytargets for accuracy improvement (Epsilon 5, Malvern Panalytical).The spectral fitting and subsequent quantification was carried outusing a calibration established by using 15 different certifiedreference materials of geological origin. The accuracy of the results

was also verified by analysing three different reference materials:IAEA-Soil7 (soil sample), IAEA-433 (marine sediment) and IAEA-405 (sediment). The correspondence between the two equipmentused was assured prior use of the data.

4. Results

After validation (Torres Astorga et al., 2018), the selected ele-ments (Ba, Ca, Fe, P and Ti) were used as tracers in the catchment torecognize the main sediment sources. TOC was used in the sedi-ment mixture 4 (M4) and in sediment mixture 5 (M5) as the sixthfingerprint improve the accuracy of the results without changingthe resulting proportions. Table 1 is a summary of the fingerprintelement concentration of source samples. In five points along theriver channel the sedimentmixture samples were collected in threedifferent seasonal times, end of rainy season, end of dry season andin middle of the rainy period of the year. The MM used for thisprocedure was CSSIAR v2.00. The unmixing procedure was madetaking into a count the sources that contribute to each mixture, inaccordancewith the flow direction of the river. Results on sedimentapportionment in channel mixtures are reported in Fig. 5.

5. Discussion

The selected elemental tracers can be used as suitable finger-prints due to specific features of the land uses of the exploredcatchment. Calcium content is lower on the top soils of agriculturalfields as compared to soil samples collected in native vegetationareas and stream banks, which have no human intervention. Cacontent is also high in feedlot soil. Likewise, iron shows differentconcentrations, the lower content of Fe may be due to the constant

Fig. 2. Land uses map of Durazno sub-catchment and sediment mixtures location tagged from M1 to M5.


413

manuring in the feedlot soils (Whalen et al., 2000). The trees topsoil(walnuts) are also expected to have a lower Fe content in thannative vegetation top soils (Little & Bolger, 1995). Feed lot soils areexpected to have the highest content of phosphorus due to cattlemanure. Phosphorus content in bank and dirt road soils is notablylower than in the rest of the sources. The lowcontent of phosphorusin cropping soils is in agreement with results of Roger et al., 2014,

who found higher concentrations of this element in grasslands andmountain pastures than in crop lands. Titanium content could beinherited from the parent material, and its variability may showdifferences because of the loess materials’ provenance. The feedlotsource has the lowest value of all sources, and it is different frombanks, rotation crops and roads, considering their standard de-viations. TOC content is strongly higher in feedlots soils, as

Fig. 3. Activities and land uses identified in the study area. (a) rotation crops, (b) nut orchards, (c) dirt roads, (d) native vegetation, (e) grazing, (f) river banks, (g) feedlot and (h)river channel in summer.


414

expected, because of cattle’s manure. Native soils present highvalues as well, due to the abundant vegetation of the area. As ex-pected, the lowest contents of TOC are banks and dirt roads, due tothe lack of vegetation. A descriptive statistics (mean and standarddeviation) of the sources soils fingerprint properties can be foundin Table 1.

The analysis of the channel mixture sediments revealed thatfeedlot soils were one of the major sources of the sediments thatreached the watercourses, and in one of the mixtures the propor-tion of this source was 90% (Sediment Mixture 5 - M5). River banksand dirt roads together are the second most important source ofsediments, particularly at the end of the dry season (period II) whenvegetation coverage is limited. Both sources, which consist ofsubsoil material, are the main source of sediments in all three

downstream mixtures at the end of the dry season. These uncov-ered soils reach 43% in Sediment Mixture 1 (M1), in the period oftime II, i.e. after the dry season. In some cases, rangelands andpasture lands (treated as grazing) are considered as the main sourcein two channel sediment mixtures. Moreover, where grazing is themajor contributor, the proportions are high (76% and 60%, in thethird period in Sediment mixture 2 (M2), and in the second periodin Sediment mixture 4 (M4)). This might be explained by a largernumber of animals living in that area and their proximity to thewater channel. Another important outcome is the low contributionof the native vegetation and nut orchards sources. This is not sur-prising, as no major soil removal is expected to occur any in thesezones.

Analysing the temporal changes in the proportions, we can only

Fig. 4. Comparison between actual (in red) and calculated soil proportion in the artificial mixed samples for (a) MIX 1 and (b) MIX 2. Blue bars represent soil proportions calculatedby CSSIAR model and green bars soil proportions calculated by IsoSource model. The error bars of actual proportions depicted the associated uncertainty of preparing the artificialmixtures. For the calculated soil proportions of CSSIAR, the standard deviation provided by MM was included as error bars (Adapted from Torres Astorga et al., 2018).

Table 1Source samples fingerprint mean concentrations and standard deviations.

Sample P (ppm) Ca (ppm) Fe (ppm) Ti (ppm) Ba (ppm) TOC (%) Number of samples

Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD

Bank 16 21 24,000 1700 38,600 3100 4710 260 526 49 0.3 0.3 6Feedlot 2600 1000 24,300 3400 32,400 3100 4120 320 452 48 5.3 2.2 5Rotation 140 100 18,300 2200 36,900 2300 4800 180 507 17 1.9 0.6 10Grazing 280 270 21,800 3700 38,500 5000 4690 310 472 42 3.1 2.0 8Trees 260 140 20,000 2800 35,200 980 4560 170 503 30 2.9 1.3 5Native 300 200 26,600 2200 41,000 4700 4560 280 436 27 5.1 4.3 5Roads 63 90 27,100 4700 41,400 3000 4870 240 489 32 0.7 0.0 2


415

observe a clear relationship, in most of the channel mixtures, be-tween the contributions of sediments from subsoils and the dryperiod of the year. For the rest of the sources, there is no relation-ship between the land uses and the different periods of the year.

Soil erosion is a key degradation factor in highly vulnerableenvironments, such as the one explored in this research. This po-tential loss in the environmental quality is not only due togeological (i.e. topography, lithology) or climatological factors (i.e.rain regime, singular events of intense precipitation) but also due tothe introduction of new land uses and changes in vegetation cover.Agricultural development leads to the rapid transformation ofnative vegetation into pasture and farmland. In this context, it isimperative to have suitable indicators of the environmental con-sequences derived from new land uses and other activities associ-ated with them.

In this investigation we have shown that geochemical elementsused as fingerprints of the sediment transport allow us to identifyhot spot areas of erosion. In these areas, it would be necessary to actwith specific strategies in order to reduce or to avoid soil

deterioration processes. Taking into account the identified mainsediment sources, some possible measures could be:

a. To control the slow but permanent expansion of the gullies andriver banks, preserving the natural vegetation or planting inblocks or strips, in order to have an appropriate buffer zonearound the streams.

b. To consider dirt roads as important sources of sediment. It isvery likely that their contribution will grow due to the increasein population in the region and the use of machines and othermeans of transportation employed in agricultural activities. Afeasible alternative to reduce this contributionwould be to pavethe main access roads and all the sides a suitable plantingshould be established. This actionwould provide a living filter tosoil particles.

c. As demonstrated, during the annual cycle the main feedlots areactive sources of sediment. In this case, an additional problem,besides sediment mobilization, is the increase of organic andinorganic pollutants towards effluents. The size of these

Fig. 5. Sediment mixtures (fromM1 to M5) collected in 5 sites along the main channel. The samples were collected at three times of the year, period I: at the end of the rainy season(after harvesting), period II: at the end of dry season and period III: at the middle of the rainy season. In the sites M4 and M5 (subfigures (d) and (e)) sediment collection was notpossible, for logistic issues, in period I.


416

facilities, the maximum number of animals and the locationaway from streams and rivers should be critically analysed. Thepermanent control of soil movement inside the feedlot and thedrainage channels should be strictly controlled.

On a prospective basis, the degree of impact of these measuresof environmental preservation could be annually tested using thesame identified fingerprints, with a low cost and proven efficiency.

6. Conclusions

In this study, the geochemical fingerprints approach has beenused to explore sediment sources in a small mountain catchment ina semiarid region. Element signatures allow for discriminatingsources based on different land uses in the same lithology (qua-ternary deposits). The most relevant results obtained are:

a. The same set of geochemical elements (Ba, Ca, Fe, P and Ti)allowed us to approach the source contribution in the laboratoryartificial mixtures;

b. These tracers, used as sources signatures in the whole catch-ment, made it possible to identify the feedlots as the mainsource of sediments in most of the analysed channel sedimentmixtures;

c. Together, river banks and dirt roads are the second mostimportant source of sediments. Indeed, the limited vegetationcover during the dry season favours sediment movement;

d. Rangelands and pasturelands can be a relevant source ofsediments;

e. The area of native vegetation presents one of the lowest con-tributions to soil erosion.

The identification of the main sources of sediments using thegeochemical signature allows the monitoring of the watershedgiving relevant information in a relatively quickly and cost-effectiveway. Nevertheless, given the complexity of the problem and thelimitation of the technique, the method should be applied as acomplement to other more conventional approaches.


The authors that they declare have no known competingfinancial interests personal relationships that could have appearedto influence the work reported in this paper.

Acknowledgements

The authors would like to thank the community of farmers thatallowed us to collect the samples. The collaboration of EstanciaGrande’s mayor was important and much appreciated. The authorsthank the IAEA Nuclear Science and Instrumentation LaboratoryTeam (Seibersdorf, Austria) for providing insight and expertise inthe EDXRF sample preparation and measurements. The first authorwould like to acknowledge the STEP (Sandwich Training Educa-tional Programme) fellowship financed by IAEA/ICTP, whichallowed the training, measurement and discussion of results atIAEA Seibersdorf Laboratories, Austria. The authors would like tothank GAECI team for the English review of this article.

This investigation has been conducted in the framework of the

IAEA-funded regional Latin American Technical Cooperation Pro-jects [ARCAL RLA 5064 and ARCAL RLA 5076]. This research wassupported by CONICET, Argentina [PIP 112 201501 00334] andUniversidad Nacional de San Luis, Argentina [PROICO 22/F41].

References

Baldi, G., & Jobb�agy, E. G. (2012). Land use in the dry subtropics: Vegetationcomposition and production across contrasting human contexts. Journal of AridEnvironments, 76, 115e127. https://doi.org/10.1016/j.jaridenv.2011.08.016

Blake, W. H., Ficken, K. J., Taylor, P., Russell, M. A., & Walling, D. E. (2012). Tracingcrop-specific sediment sources in agricultural catchments. Geomorphology,139e140, 322e329. https://doi.org/10.1016/j.geomorph.2011.10.036

de la Casa, A. C., & Ovando, G. G. (2014). Climate change and its impact on agri-cultural potential in the central region of Argentina between 1941 and 2010.Agricultural and Forest Meteorology, 195e196, 1e11. https://doi.org/10.1016/j.agrformet.2014.04.005

Gasparri, N. I., & Grau, H. R. (2009). Deforestation and fragmentation of Chaco dryforest in NW Argentina (1972e2007). Forest Ecology and Management, 258(6),913e921. https://doi.org/10.1016/j.foreco.2009.02.024

Hardy, F., Bariteau, L., Lorrain, S., Th�eriault, I., Gagnon, G., Messier, D., & Rougerie, J.-F. (2010). Geochemical tracing and spatial evolution of the sediment bed load ofthe Romaine River, Qu�ebec, Canada. Catena, 81(1), 66e76. https://doi.org/10.1016/j.catena.2010.01.005

Harris, D., Horw�ath, W. R., & Van Kessel, C. (2001). Acid fumigation of soils toremove carbonates prior to total organic carbon or carbon-13 isotopic analysis.Soil Science Society of America Journal, 65(6), 1853e1856.

IAEA. (2019). Sample preparation of soil and plant material for isotope ratio massspectrometry. IAEA TECDOC 1870. IAEA Publication). Joint FAO/IAEA Division ofNuclear Techniques in Food and Agriculture.

Little, D., & Bolger, T. (1995). The effects of contrasting land uses on soil propertiesand animal communities in Brown Earth soils. Biology and environment: Pro-ceedings of the Royal Irish Academy, 95B(3), 183e193.

Morosini, A. F., Su�arez, A. E. O., Otamendi, J. E., Pagano, D. S., & Ramos, G. A. (2017).La Escalerilla pluton, San Luis Argentina: The orogenic and post-orogenicmagmatic evolution of the famatinian cycle at Sierras de San Luis. Journal ofSouth American Earth Sciences, 73, 100e118.

Penalba, O. C., & Vargas, W. M. (2004). Interdecadal and interannual variations ofannual and extreme precipitation over central-northeastern Argentina. Inter-national Journal of Climatology, 24(12), 1565e1580.

Phillips, D. L., & Gregg, J. W. (2003). Source partitioning using stable isotopes:Coping with too many sources. Oecologia, 136(2), 261e269. https://doi.org/10.1007/s00442-003-1218-3

le Polain de Waroux, Y., Garrett, R. D., Heilmayr, R., & Lambin, E. F. (2016). Land-usepolicies and corporate investments in agriculture in the Gran Chaco and Chi-quitano. Proceedings of the National Academy of Sciences of the United States ofAmerica, 113(15), 4021e4026. https://doi.org/10.1073/pnas.1602646113

Roger, A., Libohova, Z., Rossier, N., Joost, S., Maltas, A., Frossard, E., & Sinaj, S. (2014).Spatial variability of soil phosphorus in the Fribourg canton, Switzerland.Geoderma, 217e218, 26e36. https://doi.org/10.1016/j.geoderma.2013.11.001

de los Santos-Villalobos, S., Bravo-Linares, C., Meigikos, dos A. R., Cardoso, R.,Gibbs, M., Swales, A., Mabit, L., & Dercon, G. (2017). The CSSIAR v.1.00 software:A new tool based on SIAR to assess soil redistribution using compound specificstable isotopes. SoftwareX, 6, 13e18. https://doi.org/10.1016/j.softx.2016.12.005

Teruggi, M. E. (1957). The nature and origin of Argentine loess. Journal of Sedi-mentary Research, 27(3). http://archives.datapages.com/data/sepm/journals/v01-32/data/027/027003/0322.htm.

Torres Astorga, R., de los Santos-Villalobos, S., Velasco, H., Domínguez-Quinteros, O.,Pereira Cardoso, R., Meigikos dos Anjos, R., Diawara, Y., Dercon, G., & Mabit, L.(2018). Exploring innovative techniques for identifying geochemical elementsas fingerprints of sediment sources in an agricultural catchment of Argentinaaffected by soil erosion. Environmental Science and Pollution Research, 25(21),20868e20879. https://doi.org/10.1007/s11356-018-2154-4

Viglizzo, E. F., Frank, F. C., Carre~no, L. V., Jobb�agy, E. G., Pereyra, H., Clatt, J., Pinc�en, D.,& Ricard, M. F. (2011). Ecological and environmental footprint of 50 years ofagricultural expansion in Argentina. Global Change Biology, 17(2), 959e973.https://doi.org/10.1111/j.1365-2486.2010.02293.x

Whalen, J. K., Chang, C., Clayton, G. W., & Carefoot, J. P. (2000). Cattle manureamendments can increase the pH of acid soils. Soil Science Society of AmericaJournal, 64(3), 962e966. https://doi.org/10.2136/sssaj2000.643962x

Zak, M. R., Cabido, M., & Hodgson, J. G. (2004). Do subtropical seasonal forests in theGran Chaco, Argentina, have a future? Biological Conservation, 120(4), 589e598.https://doi.org/10.1016/j.biocon.2004.03.034


417


Soil erosion: An important indicator for the assessment of landdegradation neutrality in Russia

Petr Tsymbarovich a, b, German Kust a, *, Mikhail Kumani c, Valentin Golosov a, b, **,Olga Andreeva a

a Institute of Geography RAS, Russian Academy of Sciences, Moscow, Russiab Lomonosov Moscow State University, Moscow, Russiac Kursk State University, Kursk, Russia


Article history:Received 5 February 2020Received in revised form22 May 2020Accepted 8 June 2020Available online 9 July 2020

Keywords:Russian FederationSoil erosionLand degradation neutrality

a b s t r a c t

This review of soil erosion (SE) studies in Russia focuses on two main tasks: (i) ensuring thecompleteness and reliability of SE data in Russia, a large country (17.1 million km2) with a variety ofnatural and socio-economic causes of land degradation, (ii) assessing the possibility of including a SEindicator among the indicators of land degradation neutrality (LDN). A wide range of statistical, remotesensing, mathematical modeling data, the results of scientific and field studies obtained at differentlevels were analyzed. It is asserted that in Russia the total area of eroded lands and those under erosionrisk occupy more than 50% of all agricultural lands, whereas soil fertility of croplands decreased in Soviettime (from 1950s to 1980s) by 30e60% only due to water erosion. However, recent scientific studiesindicate a decrease in erosion rate and in the area of eroded land during the last 30e40 years as a resultof abandonment of arable land and subsequent overgrown with natural vegetation. The climate changeresulting in decrease of the depth of soil freezing, flow of spring runoff also adds to the decrease of soilerosion. The SE indicator was suggested as an important complement to three global LDN indicators. Atnational and subnational level, it can be interpreted through such indices as “Rate of soil loss” (ton ha-1yr-1) and “Total soil loss” (1000 tons, in certain area during selected time period). At local level the set ofindices can be wider and site-specific, including those obtained through remote sensing data by usingthe classifier of thematic applications of remote sensing technologies; the example was tested at the localsite.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction

Soil erosion is the most common problem of land degradation inRussia, covering not only agricultural, but also pasture and forestlands. The issue of correct assessment of soil erosion dynamics is ofparticular importance in connection with the recent developmentof the new concept of land degradation neutrality (LDN). UNCCDdetermined LDN as “a statewhereby the amount and quality of landresources, necessary to support ecosystem functions and services

and enhance food security, remains stable or increases withinspecified temporal and spatial scales and ecosystems” (UNCCD,2015). At global level the state of LDN is evaluated through usingthe SDG indicator 15.3 “Proportion of land that is degraded overtotal land area” and its three subindicators (changes in land cover,land productivity and soil organic carbon) considering the “one outall out” principle: (Akhtar-Schuster et al., 2017). Studies of theapplicability of global LDN indicators for Russia, taking into accountthe wide climate, landscape, and land use diversity, have shown(Kust, 2019) that the indicator of SOC changes is the most prob-lematic for application of the LDN concept. Detailed monitoring ofSOC indicator stopped in Russia with the collapse of the Sovietsystem in the early 90s, and the new methods based mainly onremote sensing data (ISRIC, 2018) are still far from sufficient ac-curacy. At the level of regional (national and subnational) assess-ments the trends in changes of SOC coverage and dynamics

* Corresponding author.** Corresponding author. Institute of Geography RAS, Russian Academy of Sci-ences, Moscow, Russia.

E-mail addresses: [email protected] (G. Kust), [email protected], [email protected](V. Golosov).






ascertained by national institutions in general coincide with thetrends obtained by the analysis of global databases (Hengl et al.,2017) although with some difference in absolute values of SOCcontent and stocks. However, at the local level the reliability of theanalysis and the accuracy of the estimates are still at unacceptablevalues and currently cannot serve as a reliable source to indicatechanges in SOC. According to our expert estimates, the globaldatasets have reported SOC differences with Russian national dataof 1.5e3 times (Kust et al., 2018).

In this respect, it is important to search for alternative adequateindicators of land degradation, the dynamics of which at differentlevels (from national to local) can be determined using modernmethods and reliable mathematical and cartographic modeling. Assuchan indicator, in thispaperweconsider thepossibilityofusing soilerosion data obtained using various methods and data sources,including remote sensing. The importance of taking soil erosion intoaccount as an independent or additional subindicator for assessingLDN, aswell as an alternative to the SOC indicator, was pointed out bymany authors (Cowie et al., 2019; Gilbey et al., 2019; Solomun et al.,2018; Speranza et al., 2019). Some authors have noted a close rela-tionship between SOC and soil erosion (Chappel et al., 2019; Chaseket al., 2019). Keesstra et al. (2018) indicated that in most (northern)European countries, water erosion along with pollution is the mostimportant process of land degradation. The fact that the impact of soilerosion processes are more important than that of salinization,compaction, anddecreaseof the SOCcontentwere indicated for othercountries (Wunder & Bodle, 2019; Liniger et al., 2019; ).

2. Methodology

This work is of an overview nature, therefore, for the collectionand generalization of information, we used numerous scientific,archival, analytical and statistical materials and the results ofmathematical modeling that describe various generalizations: fromlocal to national.

At the national (federal) level in Russia, the assessment of thearea prone to soil erosion, its rate and degree is carried out mainlywith the help of the state institutes of the so-called agrochemicalservice, which conducts observations on agricultural lands andprovides extension service. These lands occupy less than 25% of thetotal area of the country, including arable land, meadows andpastures, including deer pastures (Rosreestr, 2017), therefore theagrochemical service does not represent a complete picture for allthe country’s land resources. This makes it difficult to assess thedynamics of soil erosion at the national level, especially regardingindustrial disturbances beside agricultural land (due to the con-struction of linear infrastructure (roads, pipelines)), mining, pro-specting, drilling and other anthropogenic impacts. Significantviolations of this kind are also noted in the subarctic and arcticregions (Grigoriev et al., 2006; Kapitonova et al., 2016), whereagriculture is not carried out. Therefore, to assess soil erosion at thenational level only approximate estimates based on expert syn-thesis of different data sources can be obtained by collecting andsummarizing disparate data from state official statistics, scientificresearches carried out at different scales, and various materials ofcartographic and mathematical modeling.

At the local level, the accuracy of the soil erosion assessment ismore detailed, because a complex of methods is used for this pur-pose: from direct observations that record runoff rates at field plots(Medvedev et al., 2016) to mathematical modeling methods thatevaluate the risks, also rate and degree of soil erosion development(Gobin et al., 2002; Ketema & Dwarakish, 2019; Kuznetsov &Glazunov, 1996; Lisetsky et al., 2012).

Recently remote sensing methods for assessment and mappingof soil erosion are increasingly used (Garcia et al., 2019; von Maltitz

et al., 2019; Crossland et al., 2018; Medvedeva et al., 2018). With theaccumulation of large arrays of dynamic remote sensing data andthe development of correct methods for their interpretation, it ispossible to receive reliable, comparable and feasible results for thewhole country with minimal need for confirmation by ground data.This is of particular importance for remote inaccessible areas whereland degradation occur as a result of trigger reactions caused byrandom short-term impact (for example, prospecting works withheavy geophysical equipment) and the subsequent long-termdevelopment of a negative loop of after-effect reactions(Dorozhukova & Yanin, 2004).

3. Soil erosion assessment and dynamics in the RussianFederation

3.1. Assessment at national level and “soil erosion neutrality”

Soil erosion is the most common process of land degradation inRussia. Sheet, rill and gully erosion during snowmelt (MarcheApril)and rainstorm (MayeSeptember) seasons are the main factors forsoil degradation of agricultural lands and other human-affectedsurfaces.

The analysis of the state monitoring of agricultural land (Statereport, 2016, 2017; EMISS, 2018) shows that the total area oferoded, deflated and lands potentially prone to deflation and watererosion is over 50%. The processes of erosion especially dramati-cally affect Chernozems (known as highly fertile black soils), whichconstitute more than 40% of the total arable land of the country. Forexample, in the central part of Russia, 34% of arable land and 51% ofpasture area are vulnerable to water erosion, and respectively 18%and 15% are vulnerable to wind erosion. In the southern regions,two thirds of the total agricultural land area are at risk of deflation,where actual wind erosion is manifested in 11.3% of the agriculturalland area. In the boreal regions of Siberia, the Far East, and the northof the European territory of Russia, a high water and wind erosionhazard and risk persists where the natural balance of forest eco-systems is disturbed, for example, during forestry, fire, oil andtransport infrastructural development.

Among the total area of the land prone to water erosion, slightlyeroded soils occupy 87%, moderately and strongly eroded soilsoccupy 13%. The strongest development of gully erosion (gullydensity is more than 5 km km�2, gully length density is more than1.3 km km�2) is observed in forest-steppe and steppe zones, inareas of long-term and old agricultural development (State report,2017). These include deeply dissected and intersected parts of theuplands (south of Central Russian upland, separate parts of theVolga upland), composed of silty loess-like sediments.

Averaging and summarizing sample estimates conducted by theagrochemical service over the past 15 years (data taken fromEMISS,2018) show that among those agricultural lands prone to winderosion, 76% are characterized by a low degree, and 24% by mod-erate and strong deflation. Significant areas (about 50% of the total)of agricultural lands having certain risk of wind erosion are locatedin southern Siberia, although the southern European territory ofRussia and the North Caucasus are characterized by the strongestdeflation. Here, in open plain areas, the intensity of deflation rea-ches 50e100 ton ha-1 yr-1 or even more, in some areas not pro-tected by forest belts, a decrease in soil thickness during last 30e40years can reach 30e35 cm (State report, 2017). On light texturedsoils (sandy and light loam) in the south of Western Siberia in somespots the deflation also may be higher than 50 ton ha-1 yr-1.However, due to the widespread use of “flat-cut tillage” (cultivationof the soil with keeping stubble on the surface) by farmers, thecommon development of deflation is unlikely here (Kulik et al.,2007, p. 86).

P. Tsymbarovich et al. / International Soil and Water Conservation Research 8 (2020) 418e429 419

At the same time, these estimates are approximate, since theyare based on sample and irregular observations, updated in recentyears only in certain regions of the country. The dynamic data thatdemonstrate changes in the state of eroded lands are required toobtain information on achieving/not achieving LDN. Although suchassessments of the degree of soil erosion of arable landwere carriedout every 15 years throughout the country during the Soviet years,regular field surveys in modern Russia have been discontinuedsince early 1990s. Therefore, generalized expert estimates carriedout and presented in the scientific literature and official statisticalreports are using different approaches (State soil erosion map ofRussia, 1999, 2004; Kulik et al., 2007, p. 86; Kuznetzov &Kashtanov, 2011; Kust et al., 2011; Krasilnikov et al., 2016; Statereport, 2016, 2017). Hence, they differ by data and provide a bigchallenge for the interpretation and comparing them for use in theLDN methodology.

Most official sources, such as State Reports, demonstrate theaverage annual water erosion of about 80 million ha of the culti-vated land is estimated at 0.56 billion tons from the late 80s andearly 90s. Up to 80e90% of phosphorus, nitrogen and pesticides inrivers and water bodies originates from agricultural soils. Expertsestimate that soil fertility decreased in Soviet time (from early1950s to late 1980s) by 30e60% only due to water erosion of arableland. The area of gullies reached more than 1 billion hectares, andthe rate of gullying was assessed as up to 10e15 thousand ha yr-1.From 80 to 90% of gullies in the agricultural area of Russia have theanthropogenic origin due to improper cultivation of arable lands.There are more than 2 million of individual gullies, with a totallength of about 300 thousand km and an area occupied of over 6million ha (State report, 2017). Almost all sources indicated an in-crease in the rate of water and wind erosion throughout thecountry until the 1990s. In this respect, soil erosion is traditionallyconsidered the main scourge of all agricultural lands.

However, since official sources mainly repeat the results ob-tained 30 years ago, they do not express the trends in the devel-opment of erosion processes nationwide, although recent scientificstudies indicate that the dynamics of erosion processes are multi-directional with a clear predominance of trends for their reductionin most regions of Russia.

In current time in Russia, especially in the boreal regions, agri-cultural lands abandoned in the late 80s and 90s of the 20th cen-tury occupy a vast area. Many of these lands were overgrown notonly with grassy vegetation, but also with shrubs and even forests(Lyuri, 2015). We also described this phenomenon in a survey of thedynamics of the land cover based on the long-term series of sat-ellite images using the Trends.Earth GIS system (2018) as a tool fordetermining LDN (Kust et al., 2018; National report, 2019). Erosionprocesses on abandoned arable lands slowed down significantly upto their stop and reverse development of progradation phenomena(Lyuri et al., 2010).

Significant socio-economic reasons caused these trans-formations and drastic changes of the main trends of soil erosion inthe last 30e40 years. The economic crisis of the late 80s led to aconsiderable transformation of land. In boreal regions the crop-lands, which covered a relatively small proportion of land cover(<20%) were abandoned due to drastically decreased productivityThe lack of resources to support irrigation systems was the mainreason for the reduction of cultivated area in the dry steppe zonewith a higher proportion of cropland (40e80%) (Nefedova, 2013).For example, the modelling of the total annual soil loss for theRussian Plain showed that it reduced from 436 Mt in 1960e1980 to245 Mt in 1991e2012, mainly due to the decrease in cropland area(Golosov et al., 2018b). The highest reduction was identified for theforest zone, where soil losses reduced up to 75% and the meanannual erosion rate decreased from 7.3 to 4.1 ton ha-1 yr-1 (Golosov

et al., 2018), although some experts consider that these data areoverestimated by 5e15% (Barabanov et al., 2018). The increasingfrequency of heavy rainstorms in the southern part of steppe zone,on the contrary, led to the minor intensification of soil erosion rate(Golosov et al., 2017). The density of active gullies decreasedconsiderably during last decades (Golosov et al., 2018c). The trendof reduction of gully head retreat is confirmed by the results oflong-termmonitoring in the south of the forest zone of the RussianPlain (Medvedeva et al., 2018). The main reason for this is likely thedecrease in soil freezing depths, which results in a significantreduction in surface runoff during snowmelt (Petelko et al., 2007).

These trends are confirmed by the processing of numerous ev-idences obtained from various sources, as well as by the results ofmathematical modeling (Chalov et al., 2017), Table 1. A modifiedversion of the Universal Soil Loss Equation (USLE) were used forcalculation soil losses because of rainfall-induced sheet erosion. Amodel developed by the Russian State Hydrological Institute (SHI)was also modified to estimate sheet erosion from snowmelt runoff(Larionov, 1993). The detail description of both models and resultsof their verification can be found elsewhere (Sidorchuk et al., 2006;Golosov et al., 2018). The contemporary erosion rates for the twotime windows (1980 and 2010) were calculated using a modifiedversion of the USLE and the SHI on a 1:500,000 scale for thecultivated slopes. The data provided in Table 1 describe how theindicator of soil erosion (SE) can assess the achievement of LDN.“Soil erosion neutrality” at the level of large regions can bemeasured by two main indices: “Rate of soil loss” (ton ha-1 yr-1)and “Total soil loss in the area” (103 ton yr-1). Following the prin-ciple “one out all out” used for LDN assessment by global indicators(Orr et al., 2017, p. 129), it is easy to see that the dynamics of erosionprocesses in different regions of Russia remain multidirectionaldespite the clear and widespread tendency of overgrowing ofagricultural lands. This multidirectional dynamic does not alsoallow to achieve “Soil erosion neutrality” (herewith LDN) in manyregions. The assumption is that the result in the rate of theachievement of LDN could be even less impressive for assessing soilerosion at a more detailed level with an increase in the number ofindices.

Thereby, in many lands, particularly in the boreal and forest-steppe regions of Russia (for the latter, especially in the case ofthe Siberian region), the rate of erosion has been sharply reduced orstopped over the past 30e40 years as a result of land abandonment(Litvin et al., 2017). The productivity of shrubbery and woodyvegetation on abandoned lands is increasing (Land degradation,2019), which indicates the restoration of natural landscapes and,therefore, may indicate the achievement of LDN evaluated by theindicator of SE dynamic in such areas. However, the multidirec-tional trends of changes of the individual indices that make up theSE indicator are noted in different regions. This indicates that eventhe general trend of overgrowing of abandoned lands at nationaland subnational levels does not allow considering lands as thoseachieved LDN within individual territories (Table 1). For a moredetailed assessment, especially at the local level, studies of addi-tional parameters (measures or indices) of SE dynamic are required.These parameters, unfortunately, cannot be provided by officialstatistics, but can be obtained using modern methods.

3.2. Possibilities for soil erosion assessment for LDN assessmentpurposes at local level

3.2.1. Classifier of thematic applications of remote sensing dataRemote sensing has a great potential for evaluation of soil

erosion. However, its application can be complicated due to highdissimilarity of remote sensing data and approaches of processingand interpretation of the data. The multipurpose classifier of

P. Tsymbarovich et al. / International Soil and Water Conservation Research 8 (2020) 418e429420

Table 1Changes in area of croplands and soil erosion in Russian economic regions from 1980 to 2010 (Chalov et al., 2017; adapted and added by authors).


thematic applications of remote sensing data and its processingalgorithms STRATO RS (Kushnyr et al., 2018) could serve as one ofthe possible solutions of the issue under discussion. This classifierwas developed to simplify automatic assessment of different pro-cesses and phenomena indicated by remote sensing methods,including evaluation of soil erosion in different spatial scalesregardless of land cover or land use types. STRATO RS is a webapplication with client-server architecture. It includes only thosethematic applications and algorithms, which are validated byparticular study cases and practices. The more successful algo-rithms are known for any particular task, the more reliable thisapplication is rated. The more cases are known used for any algo-rithm, the more confident its results are considered. Several the-matic applications of remote sensing data for assessing soil erosionwere formalized and placed in STRATO RS according to thisapproach (Table 2).

3.2.2. Testing universal classifier for assessing soil erosion (casestudy and examples)

A plot for testing classifier STRATO RS for soil erosion assess-ment is located near Panino Village in the Kursk region, Fig. 1. A fewexamples of testing results are provided below.

Two images of Landsat7 and Landsat8 for 2001 and 2018 wereused. The images were obtained from earthexplorer.usgs.gov andregistry. opendata.aws and were processed with open sourcesoftware: GDAL/OGR (gdal.org) was applied for batch processingsuch as converting between different formats, clipping, projectionand raster calculations; SAGA GIS (www.saga-gis.org) was used forprocessing DEM for analysis of topography, classification of imagesand pansharpening; QGIS (qgis.org) was used for digitization andrendering of final maps. Images in natural colors were synthesized(channels 321 for Landsat7 and 432 for Landsat8) and pan-sharpened to enhance spatial resolution. These synthesizes were

Table 2Thematic applications of remote sensing data formalized in multipurpose classifier STRATO RS for assessing soil erosion in forest, forest-steppe and steppe farmlands of humidplains of the temperate zone.

Thematic applications References

Mapping of the state of gully erosion Yermolaev et al. (2017)Analysis of dynamics of gully erosion Yermolaev et al. (2017)Mapping of land cover Singh and Dubey (2012)Indicating changes in land cover Ilmidin Kyzy et al. (2017)Indicating hollow forms of relief Тrofimetz et al. (2017)Soil erosion risk assessment in agricultural landscapes Buryak, 2014; Rozhkov, 2007Algorithms for applying remote sensing dataAssessment of state and changes in gully erosion based on gully density index in time series of space imagery Gaifutdinova et al., 2016; Yermolaev et al., 2017Assessment of state and changes in gully erosion based on river basin approach Yermolaev et al. (2017)Large-scale mapping of erosional forms based on visual interpretation of remote sensing images Тrofimetz et al. (2017)Assessment of soil erosion risk based on the topography calculated by remote sensing data (LS-factor) Buryak, 2014; Lisetsky & Polovinko, 2012Indicating changes in land cover based on NDVI time series Ilmidin Kyzy (2017)

Fig. 1. Location of the testing plot “Panino”.


used to interpret the state of erosion network by visual and auto-matic methods which showed high similarity (Fig. 2). The densityindex for erosional pattern (LS-factor, km km�2) was calculated fora regular net of cells of 500 � 500 m size. .

The results of testing of the algorithm of automatic mapping oferosion network and density of erosional pattern based on riverbasin approach and digital evaluation model (DEM) are shown inFig. 3. Boundary of river basins were calculated by the SAGA GISmodule “Basic Terrain Analysis”.

3.2.3. Mapping of the risk of soil erosionThe classifier considers an approach of mapping of the risk of

soil erosion in case if traditional soil loss models like USLE cannotbe applied due to the lack of data. In similar conditions of limitedterritories relief becomes the most critical factor of soil erosion. Inthis case its mesurable representation such as LS-factor can be usedto distinguish areas of risk of erosion. Buryak (2014) offered inter-preting maps of LS-factor based on the criteria developed throughvisual indication of soil loss on space imagery by the difference in

Fig. 2. Changes of the erosion networkUpper left: The erosion network drawn by the method of visual interpretation of images; Upper right: The increased erosion network in 2001e2018 calculated by the changes in density index; Down: The density index of the erosion network: left e 2001, based on Landsat-7 (ETMþ, 2001-05-04); right e 2018, based onLandsat-8 (OLI, 2018-08-15).

Fig. 3. Analysis of the density of erosional pattern. Left: the visualized SRTM DEM; Right: the map of erosional density drawn with the help of the river basin approach.


hue and color. This approach was tested using SRTM data and theSentinel-2A image (2019-06-06): visual interpretation detected thesurface pattern with a lighter hue, which corresponds to erodedsoils. The median value of the LS-factor was calculated for thispattern and its value 0.45 was considered as critical when erosioncan be harmful to soil fertility. It was used as a threshold for clas-sification of soil erosion risk and on this base the map of LS-factorwas developed using the SAGAGISmodule « Basic Terrain Analysis»(Fig. 4).

The analysis of the map of the testing plot evaluated it as aterritory where the critical risk of soil erosion is less than 50% of thetotal area. Using the same algorithm for time series of images, thedynamics of the risk of the erosion can be also described.

Based on these examples, it can be concluded that for local as-sessments the multipurpose classifier can help to evaluate changesfor different indices of soil erosion and to monitor more accuratelythe achievement of LDN in erosion-risky areas. The peculiarity ofusing such a classifier for assessing soil erosion is that it is appli-cable for a limited number of the most common applications. Onthe other hand, it makes it possible to apply the similar algorithmfor big and different territories that provides a possibility for esti-mates comparative in time and space.

3.3. Assessment of soil erosion trends related to climate changes

We noted earlier (Kust et al., 2018) that for the LDN assessment,it is important to take into account not only anthropogenic impactsand trends, but also climatic long-term changes, which imprints onthe nature of the processes of land degradation and restoration,taking into account local and regional characteristics. In this paper,using the example of the Kursk region of Russia we consider somefeatures that affect the long-term dynamics of soil erosion and haveto be taken into account when modeling erosion and LDN forecastsbased on the SE indicator.

The erosion runoff and, therefore, the total soil loss on arableslopes in most of Russia is determined, on the one hand, by theconditions of the soil surface and meteorological characteristicsduring spring snowmelt, and on the other hand, by the frequency ofheavy rainfalls (Chalov et al., 2017; Demidov, 2016, p. 62). Anotherimportant factor in the formation of surface runoff is the type of thesoil treatment and the soil cover pattern, which in turn are deter-mined by the set of crops and soil conservation measures used.

A very significant feature of the conditions of the formation ofsnowmelt and flood is the degree of soil freezing by the beginning

of the formation of surface runoff in spring.We analyzed the data ofthe depth of soil freezing obtained since 1938 at weather stationsPetrinki and Ushakovo in Kursk region. The maximum freezingdepth is observed, as a rule, at the end of the period with negativeair temperatures, in late March - early April, coinciding in mostcases with the beginning of spring snowmelt and the formation offloods. The degree of soil freezing affects soil water permeabilityand the formation of a “locking screen”, which determines thecoefficient of surface runoff from the slopes during the spring floodperiod and its extent.

The long-term values of the spring flood features for the TuskarRiver basin (A ¼ 2380 км2) under consideration form coincidinglinear trends in the analyzed period (Fig. 5 a, b): the maximum flowof the flood runoff and thickness of the layers of runoff decreasealong with the decrease of the maximum depth of the soil freezing.The trend of freezing depth decrease is statistically significant andcoincides with the trends in spring flood characteristics. Similartrends were found for the other regions of the Central ChernozemnyFederal okrug (Golosov et al., 2017, 2018a; Golosov et al., 2017).

The opposite trend is typical for the minimum low-water flowsof summer and winter (Fig. 6b). Long-term values of the consideredfeatures of low-water runoff both in winter and in summer time(“open channel period”) have a significant trend to increase. Thisindicates an increase in the share of underground runoff as a resultof a decrease in surface runoff on arable land. At the same time, thelong-term dynamics of the annual runoff layer is not wellexpressed, and annual precipitation is also relatively stable (Fig. 6a).

A same analysis of the flood characteristics was carried out for38 hydrological stations of the Central Chernozem region(Apukhtin, Kumani, 2015). They demonstrate the similar trend forthe all rivers including in the entire region: with a relatively stableannual flow and annual precipitation, a decrease in spring floodrunoff is observed, especially in the second half of the snowmeltperiod, and in parallel an increase in low-water runoff is takingplace during the summer and winter.

The described long-term trend of redistribution of river flows byseasons of the year (spring water discharge decrease, low-waterwater discharges increase, annual runoff only slightly reduce)clearly indicates serious transformation of the proportional con-tributions of surface and underground runoff in this region. Inrecent decades the surface runoff from arable slopes during theperiod of spring snowmelt has practically ceased, which led to thecessation of meltwater erosion previously amounted to 1e1.5 tonha-1 yr-1 per year (Chernyshev, 1980). This is also evidenced bydecrease in sedimentation rates in the upper parts of the fluvialnetwork (the bottoms of dry valleys) closest to arable land (Golosovet al., 2014, 2017; 2018a; Gusarov, Golosov,& Sharifullin, 2018). Thisconclusion is also confirmed by information on changing overbanksedimentation rates within small and medium sized river basinswithin the Central Russian Upland (Golosov et al., 2010). Sedi-mentation rate decreased up to 2.5e3 times in 1986e2009comparing to the period 1964e1986 (Markelov et al., 2012).

The frequency of runoff-forming showers gradually increases incertain regions of the European part of Russia (EPR) during lastdecades. This trend is statistically significant only for the westernregions of the EPR and Ciscaucasia within the southern megaslopeof the EPR (Golosov et al., 2018). In the rest of the EPR the trends aremultidirectional. Although in Russia the field observations of long-term average rainfall runoff from cultivated slopes are extremelylimited, it can be stated that in contrast to meltwater runoff, therainfall runoff occurs locally, but soil loss significantly exceeds thatwith melt runoff. The consequences of rain-storms were studied ontwo sites located to the south of Kursk City (Belyaev et al., 2008).Relatively moderate magnitude rainstorm events caused soilerosion evaluated as from 36 to 44 ton ha-1 yr-1. However, such a

Fig. 4. The risk of the soil erosion.


tremendous soil loss is observed on individual fields only, wherebad soil conditions contribute to the surface runoff (over-compaction of topsoil after harvesting, fallow land technology, etc.).

The estimates of the average annual soil washout based onevaluation of sediment accumulation in ponds and reservoirs witha known lifetime reliably indicate the long-term rates of soil lossfrom arable land in the Kursk region over the past decades. Mostreservoirs and ponds were constructed between late 1960s andmid-1980s. That is the estimates obtained characterize the annualrunoff values over a period of 35e50 years. This is especially true

for ponds close to arable lands, since in this case there is practicallyno redeposition of sediment washed out from the arable land andaccumulated in the reservoir (Table 3). The bigger the pondcatchment area is, the smaller is the long-term average runoffwithin these catchments, and vary in the region from 0.69 to 0.1 tonha-1 yr-1. Such a phenomenon occurs in consequence of theredeposition of about 70e80% of sediments washed away fromarable land in the bottoms of dry valleys (Golosov et al., 2017;2018a). Nevertheless, the modern long-term annual soil losses,evaluated based on data received for the near-field ponds (0.69 ton

Fig. 5. Changes of long-term values of the maximum depth of soil freezing (a) and the maximumwater discharge of spring flood (Q) and thickness of the layer of flood runoff (H) (b)for Tuskar River (data collected from gauging station Kursk, belong to Roshydromet State system).


ha-1 yr-1) is almost twice lower than the long-term annual soillosses on slope catchments during snowmelt in the Kursk region in1960e1980s (Chernyshov, 1976; Gerasimenko, 1995). This is addi-tional confirmation a significant decrease in soil erosion in the lastthree decades.

4. Conclusion

Soil erosion is considered the main process of land degradationin Russia that is reflected in different sources of data includingofficial. It is asserted that the total area of eroded and deflated lands

Fig. 6. Changes of long-term values and trends of the minimum water discharges (Qmin) in summer and winter (a) and thickness of the layers of annual runoff (H) and annualprecipitation (h) for Tuskar River (b); data collected from gauging station Kursk, belong to Roshydromet State system.


and lands potentially prone to wind and water erosion is over 50%of agricultural lands. On average the annual water erosion from thecultivated land (about 80 million ha) is estimated as 0.56 billiontons. The flow of water and sediments from the slopes in theagricultural area supplies up to 80e90% of phosphorus, nitrogenand pesticides to rivers and water bodies. Experts estimated soilfertility of arable land decreased from early 1950s to late 1980s by30e60% only due to water erosion. However, there is no more ac-curate data on the current state of soil erosion in Russia, sinceofficial information characterizes mainly agricultural land and hasnot been updated since the early 1990s. Relatively new scientificdata are scattered and describe various regions that differ in naturaland socio-economic conditions. Despite this, many recent scientificstudies show that in the last 30e40 years the opposite processeshave been actively developing and leading to a decrease in erosionrate and degree and in the total area of eroded land. The reason isthat a significant area of arable land has been abandoned and isovergrownwith shrubs and forests due to the economic crisis of thelate 80se90s.

The soil erosion dynamics indicator (SE) with such indices as“Rate of soil loss” (ton ha-1 yr-1) and “Total soil loss” (1000 tons, incertain area during selected time period) can serve as an importantcomplement to three global LDN indicators (dynamics of landcover, land productivity and soil organic carbon) for assessing landdegradation at the national and sub-national level in Russia. It wasfound by using these indices that despite the general trend ofreduction of the area of arable land, the dynamics of erosion pro-cesses (and these two indices) are multidirectional. In at least halfof Russia’s economic regions, a LDN state evaluated according to theSE indicator has not been achieved over the past 30 years. Thereason for this is the deterioration of at least one of these indices orboth at once.

It is important to create capacities to analyze the dynamics ofsoil erosion on a unified methodological basis for different types oflands cover and land use for Russia, a country with a large territory.For this purpose, a universal classifier of thematic applications ofremote sensing data and its processing algorithms can serve as auseful solution, which allows to obtain in a detailed scale data ongully dynamics, changes in erosion fragmentation and erosionhazard based on processing of satellite images. This classifier wastested at the local site in the Kursk region.

Simulation of soil washout taking into account the volumes ofmelt and rainfall runoff and using maps and other data obtainedwith the universal classifier in the future will make it possible toanalyze the dynamics of “Rate of soil loss” and “Total soil loss”, aswell as other indices of soil erosion at a local level. Taking intoaccount the principle of “one out all out”, this can provide morefully application of the entire set of soil erosion indices to establishand diagnose the achieving/non-achieving of LDN targets atdifferent scale levels, including detailed assessment.

The decrease in volume and rate of soil erosion in Russia overthe past 30e40 years is connected not only with abandonment andovergrowing of arable land, but also with certain climatic trends,such as a decrease in the maximum depth of the soil freezing,maximum flow of spring runoff and thickness of the layers of floodrunoff. The revealed long-term trend of redistribution of river flowsby seasons of the year (flood runoff decreases, low-water runoff

increases) clearly indicates a sharp reduction in surface runoff fromarable slopes. Surface runoff from cultivated lands on slopes duringthe period of spring snowmelt has practically stopped in recentdecades. This resulted in the cessation of meltwater erosion.Therefore, these climatic trends, along with land abandonment andanti-erosion SLM measures on arable land, jointly contribute to theachievement and maintenance of LDN.


I declare that there is no financial/personal interest or belief thatcould affect our objectivity, and/or personal relationships thatcould have appeared to influence the work reported in this paper.

Acknowledgements

The work was prepared in the framework of State scientific taskof the Institute of Geography of the Russian Academy of Sciences(RAS), and of Lomonosov Moscow State University, and supportedby Russian Science Foundation project # 18-17-00178 and byRussian Foundation for Basic Research project # 19-29-05025mk.

References

Akhtar-Schuster, M., Stringer, L. C., Erlewein, A., Metternicht, G., Minelli, S.,Safriel, U., & Sommer, S. (2017). Unpacking the concept of land degradationneutrality and addressing its operation through the Rio Conventions. Journal ofEnvironmental Management, 195(Pt 1), 4e15. https://doi.org/10.1016/j.jenvman.2016.09.044

Apukhtin, A. V., & Kumani, M. V. (2015). Mnogoletnjaja dinamika osnovnyh jele-mentov vesennego stoka malyh i srednih rek Central’nogo Chernozem’ja[Longstanding dynamics of basic elements of the spring drain of the small andmiddle rivers of the Central Chernozem region], Nauchnye vedomosti Belgor-odskogo gosudarstvennogo universiteta. no 21 Estestvennye nauki, V., 33(218),114e120 (in Russian).

Barabanov, A. T., Dolgov, S. V., Koronkevich, N. I., Panov, V. I., & Petel’ko, A. I. (2018).Surface runoff and snowmelt infiltration into the soil on plowlands in theforest-steppe and steppe zones of the east European plain. Eurasian Soil Science,51(1), 66e72. https://doi.org/10.1134/S1064229318010039

Belyaev, V. R., Golosov, V. N., Kislenko, K. S., Kuznetsova, J. S., & Markelov, M. V.(2008). Combining direct observations, modelling and 137Cs tracer for evalu-ating individual event contribution to long-term sediment budgets. InJ. Schmidt, et al. (Eds.), Vol. 325. Sediment dynamics in changing environments(pp. 114e122). Wallingford, UK: IAHS Publ. IAHS Press https://iahs.info/uploads/dms/14502.18-114-122-13-325-Belyaev.pdf.

Buryak, Z. A. (2014). Improvement of approaches to assessing the risk of erosion inagricultural landscapes using GIS technology. Belgorod State University ScientificBulletin. Natural Sciences, 23(194), 140e146 [In Russian: Sovershenstvovaniepodkhodov k otsenke ehrozionnoj opasnosti agrolandshaftov s ispol’zovaniemGIS-tekhnologij].

Chalov, R. S., Sidorchuk, A. Y., & Golosov, V. N. (Eds.). (2017). Erosion and fluvialsystems (Vol. 2017, p. 698). Moscow: INFRA-M [In Russian: Erozionno-ruslovyyesistemy: monografiya/pod nauch. red. R.S. Chalova, A.Yu. Sidorchuka, V.N.Golosova].

Chappell, A., Webb, N. P., Leys, J. F., Waters, C. M., Orgill, S., & Eyres, M. J. (2019).Minimising soil organic carbon erosion by wind is critical for land degradationneutrality. Environmental Science & Policy, 93, 43e52. https://doi.org/10.1016/j.envsci.2018.12.020, 2019.

Chasek, P., Akhtar-Schuster, M., Orr, B. J., Cowie, A. L., Ratsimba, H. R., & Safriel, U.(2019). Land degradation neutrality: The science-policy interface from theUNCCD to national implementation. Environmental Science & Policy, 92,182e190. https://doi.org/10.1016/j.envsci.2018.11.017, 2019.

Chernyshov, E. P. (1976). Tendencies in erosion processes in the southern Russianplain. In Problems in anthropogenic changes of water reserves. Moscow, 1976 (pp.47e63) ([in Russian]).

Cowie, A. L., Waters, C. M., Garland, F., Orgill, S. E., Baumber, A., Cross, R.,O’Connell, D., & Metternicht, G. (2019). Assessing resilience to underpin

Table 3Rate of annual soil erosion evaluated by the volume of sediments in ponds and reservoirs in the Kursk region (Yermolaev et al., 2014, pp. 333e339 with modifications).

Size of reservoir Catchment area (km2) Number of studied reservoirs Mean catchment area (km2) Mean annual soil erosion rate from catchment area, ton ha-1 yr-1Small in-field ponds <0.25 10 0.12 0.69Small reservoirs 0.25e10 11 4.5 0.13Medium reservoirs 10e50 28 32 0.1


implementation of land degradation neutrality: A case study in the rangelandsof western new south wales, Australia. Environmental Science & Policy, 100,37e46. https://doi.org/10.1016/j.envsci.2019.06.002, 2019.

Crossland, M., Winowiecki, L., Pagella, T., Hadgu, K., & Sinclair, F. (2018). Implica-tions of variation in local perception of degradation and restoration processesfor implementing land degradation neutrality. Environ. Dev., 28, 1e13. https://doi.org/10.1016/j.envdev.2018.09.005, 2018.

Demidov, V. V. (2016). Patterns of formation of erosion processes during snowmelt inthe forest-steppe zone of central Russia: Theory and experimental studies. Novo-sibirsk: Center for the Development of Scientific Cooperation ([in Russian]).

Dorozhukova, S. L., & Yanin, E. P. (2004) [In Russian: Ekologicheskiye problemyneftegazodobyvayushchikh territoriy Tyumenskoy oblasti]. Ecological problemsof oil and gas producing territories of the Tyumen region (Vol. 2004, p. 56).Moscow: IMGRE ([in Russian]).

EMISS. (2018). Unified interdepartmental information and statistical system. Availableat https://www.fedstat.ru. (Accessed 4 March 2019) ([in Russian]).

Gaifutdinova, R. A., & Yermolaev, O. P. (2016). Factors and space-time characteristicsof gullies in Tatarstan [In Russian: Faktory i prostranstvenno-vremennye oso-bennosti razvitiya ovragov na territorii respubliki Tatarstan] Bulletin of UdmurtUniversity. Series Biology. Earth Sciences, 26(2), 132e141 ([in Russian]).

García, C. L., Teich, I., Gonzalez-Roglich, M., Kindgard, A. F., & Liniger, H. (2019). Landdegradation assessment in the Argentinean Puna: Comparing expert knowl-edge with satellite-derived information. Environmental Science & Policy, 91,70e80. https://doi.org/10.1016/j.envsci.2018.10.018, 2019.

Gerasimenko, V. P. (1995). Average long-term soil erosion on arable land in variousnatural and agricultural conditions. Soil Science, 1995(5), 608e616 [In Russian].

Gilbey, B., Davies, J., Metternicht, G., & Magero, C. (2019). Taking Land DegradationNeutrality from concept to practice: Early reflections on LDN target setting andplanning. Environmental Science & Policy, 100, 230e237. https://doi.org/10.1016/j.envsci.2019.04.007, 2019.

Gobin, A., Govers, G., Jones, R., Kirkby, M., & Kosmas, C. (2002). Assessment andreporting on soil erosion. Background and workshop report. European Environ-ment Agency. https://esdac.jrc.ec.europa.eu/ESDB_Archive/eusoils_docs/Pub/bob/EEATechRepNo84.pdf.

Golosov, V. N., Belyaev, V. R., Markelov, M. V., & Kislenko, K. S. (2010). Overbanksedimentation rates on the flood plains of small rivers in Central EuropeanRussia. Sediment Dynamics for a Changing Future. In K. Banasik, et al. (Eds.),Proceedings of the ICCE symposium held at the warsaw university of life Sciences -SGGW, Poland, 14e18 june 2010. IAHS publ (pp. 129e136). Wallingford, UK: IAHSPress. no. 337 https://iahs.info/uploads/dms/14994.22-129-136-337-15-Warsaw_Golosov_sm.pdf.

Golosov, V. N., Ivanova, N. N., Gusarov, A. V., & Sharifullin, A. G. (2017). Assessmentof the trend of degradation of arable soils on the basis of data on the rate ofstratozem development obtained with the use of 137CS as a Сhronomarker.Eurasian Soil Science, 50(10), 1195e1208. https://doi.org/10.1134/S1064229317100039, 2017.

Golosov, V., Koiter, A., Ivanov, M., Maltsev, K., Gusarov, A., Sharifullin, A., &Radchenko, I. (2018a). Assessment of soil erosion rate trends in two agriculturalregions of European Russia for the last 60 years. Journal of Soils and Sediments,18, 3388e3403. https://doi.org/10.1007/s11368-018-2032-1, 2018.

Golosov, V., Yermolaev, O., Litvin, L., Chizhikova, N., Kiryukhina, Z., & Safina, G.(2018b). Influence of climate and land use changes on recent trends of soil erosionrates within the Russian Plain//Land Degradation and Development, 29 pp.2658e2667). https://doi.org/10.1002/ldr.3061, 8.

Golosov, V., Yermolaev, O., Rysin, I., Vanmaercke, M., & Zaytseva, M. (2018c).Mapping and spatial-temporal assessment of gully density in the Middle Volgaregion, Russia. Earth Surface Processes and Landforms, 43, 2818e2834. https://doi.org/10.1002/esp.4435

Golosov, V. N., Zhang, X., Qiang, T., Zhou, P., & He, X. (2014). Quantitative assessmentof sediment redistribution in the Sichuan hilly basin and the Central Russianupland during the past 60 years. Geogr., Environ., Sustainability, 7(3), 39e64.https://doi.org/10.24057/2071-9388-2014-7-3-35-39, 2014.

Grigoriev, M. N., Razumov, S. O., Kunitzkiy, V. V., & Spektor, V. B. (2006). Dynamics ofthe Russian east arctic sea coast: Major factors, regularities and tendencies.Kriosfera zemli, X(4), 74e94, 2006 http://www.izdatgeo.ru/journal.php?action¼output_more&id¼2&lang_num¼2&year¼2006&issue¼4&page¼74.

Gusarov, A. V., Golosov, V. N., & Sharifullin, A. G. (2018). Contribution of climate andland cover changes to reduction in soil erosion rates within small cultivatedcatchments in the eastern part of the Russian Plain during the last 60 years.Environmental Research, 167, 21e33. https://doi.org/10.1016/j.envres.2018.06.046

Hengl, T., Mendes de Jesus, J., Heuvelink, G., Gonzalez, M. R., Kilibarda, M.,Blagoti�c, A., … Kempen, B. (2017). SoilGrids250m: Global gridded soil infor-mation based on machine learning. PloS One, 12, Article e0169748. https://doi.org/10.1371/journal.pone.0169748. http://www.geokniga.org/bookfiles/geokniga-ekologicheskie-problemy-neftegazodobyvayushchih-territoriy-tyumenskoy-oblasti.pdf

Ilmidin Kyz, G., & Urmambetova, T. K. (2017). Identification of changes in land usebased on remote sensing data (on example of Kadamzhay district). [In Russian:Vyyavleniya izmenenij v zemlepol’zovanii s primeneniem dannykh distant-sionnogo zondirovaniya zemli (na primere Kadamzhajskogo rajona)]. VestnikKSUCTA, 2(56), 62e67 [in Russian].

ISRIC. (2018). International soil reference and information centre. http://www.isric.org/. (Accessed 29 May 2018).

Kapitonova, O. A., Selivanov, A. E., & Kapitonov, V. I. (2016). Vascular plants on sandyoutcrops in West Siberian arctic and subarctic. Vestnik of Perm State Humani-tarian Pedagogical University. ser. 2, 2016 https://cyberleninka.ru/article/n/sosudistye-rasteniya-na-peschanyh-obnazheniyah-v-zapadnosibirskoy-arktike-i-subarktike/viewer ([in Russian]).

Keesstra, S., Mol, G., de Leeuw, J., Okx, J., Molenaar, C., de Cleen, M., & Visser, S.(2018). Soil-related sustainable development goals: Four concepts to make landdegradation neutrality and restoration work. Land, 7, 133. https://doi.org/10.3390/land7040133, 2018.

Ketema, A., & Dwarakish, G. S. (2019). Water erosion assessment methods: A review.ISH Journal of Hydraulic Engineering. https://doi.org/10.1080/09715010.2019.1567398, 2019.

Krasilnikov, P., Makarov, O., Alyabina, I., & Nachtergaele, F. (2016). Assessing soildegradation in northern eurasia. Geoderma Regional, 7(1), 1e10. https://doi.org/10.1016/j.geodrs.2015.11.002, 2016.

Kulik, K. N., Gabunshchina, E. B., Kruzhilin, I. P., Kust, G. S., Manaenkov, A. S.,Pavlovsky, E. S., … Tulokhonov, A. (2007). Desertification and complex amelio-ration of dryland agrolandscapes. Volgograd. VNIALMI. ([in Russian]).

Kushnyr, O., Sizov, O., Tsymbarovich, P., & Slavko, V. (2018). Principles of develop-ment and structure of a multi-purpose classifier for thematic applications ofremote sensing data [in Russian: Sovremennye problemy distantsionnogozondirovaniya Zemli iz kosmosa]. 15(3), 39e52. https://doi.org/10.21046/2070-7401-2018-15-3-39-52 [in Russian].

[in Russian: Degradatsiya zemel’ i opustynivaniye v Rossii: novyye podkhody kanalizu problem i poiskov putey resheniya]. In Kust, G. S. (Ed.). (2019), Landdegradation and desertification in Russia: New approaches to the analysis ofproblems and the search for solutions (Vol. 2019, p. 235). Moscow: IzdatelstvoPero ([in Russian]).

Kust, G. S., Andreeva, O. V., & Dobrynin, D. V. (2011). Systematic study of arid ter-ritories. Desertification assessment and mapping in Russian Federation. AridEcosystems, 1(N1), 14e28. https://doi.org/10.1134/S2079096111010057, 2011.

Kust, G., Andreeva, O., Lobkovskiy, V., & Telnova, N. (2018). Uncertainties and policychallenges in implementing land degradation neutrality in Russia. Environ-mental Science & Policy, 89, 348e356. https://doi.org/10.1016/j.envsci.2018.08.010, 2018.

Kuznetsov, M. S., & Glazunov, G. P. (1996) [in Russian: Eroziya i okhrana pochv].Erosion and soil conservation (Vol. 1996, p. 335). Moscow State University http://www.pochva.com/?content¼3&book_id¼0018.

Kuznetzov, M., & Kashtanov, A. (2011). Soil erosion. National soil atlas. Available athttps://soilatlas.ru/rasprostranenie-erozii-pochv. (Accessed 4 March 2019).

Larionov, G. A. (1993). Water and wind erosion: Main features and quantitativeassessment. Moscow: Izdatelstvo Moskovskogo Universiteta (in Russian).

Liniger, H., Harari, N., Lynden, G., Fleiner, R., Leeuw, J., Bai, Z., & Critchley, W. (2019).Achieving land degradation neutrality: The role of SLM knowledge in evidence-based decision-making. Environmental Science & Policy, 94, 123e134. https://doi.org/10.1016/j.envsci.2019.01.001, 2019.

Lisetsky, F. N., & Polovinko, V. V. (2012). Erosion catenas on earthen fortifications [InRussian: Erozionnye kateny na zemlyanykh fortifikatsionnykh sooruzheniyakh]Geomorphologiya, 2, 65e77 [in Russian].

Lisetsky, F. N., Svetlichny, A. A., & Cherny, S. G. (2012) [in Russian: Sovremennyyeproblemy erozipovedeniya]. Current problems of erosion behavior Belgorod (Vol.2012, p. 456). Belgorod. Konstanta ([in Russian]).

Litvin, L. F., Kiryukhina, Z. P., Krasnov, S. F., & Dobrovol’skaya, N. G. (2017). Dynamicsof agricultural soil erosion in European Russia. Eurasian Soil Science, 50(11),1344e1353. https://doi.org/10.1134/S1064229317110084, 2017.

Lyuri, D. (2015). Dynamics of agricultural land in Russia and in the World in the XXcentury//Potential of idle agricultural lands of the Post-Soviet area to mitigatethe climate changes and improve an environment. Interdisciplinary Workshop:Book of extended abstracts, 2015, 19e22.

Lyuri, D. I., Goryachkin, S. V., Karavaeva, N. A., Denisenko, E. A., & Nefedova, T. G.(2010) [in Russian: Dinamika sel’skokhozyaystvennykh zemel’ Rossii v XX vekei postagrogennoye vosstanovleniye rastitel’nosti i pochv]. Dynamics of agricul-tural lands in Russia in the XX century and post-agrogenic restoration of vegetationand soil (Vol. 2010, p. 416). Moscow GEOS ([in Russian]).

Maltitz, G. P., Gambiza, J., Kellner, K., Rambau, T., Lindeque, L., & Kgope, B. (2019).Experiences from the South African land degradation neutrality target settingprocess. Environmental Science & Policy, 101, 54e62. https://doi.org/10.1016/j.envsci.2019.07.003, 2019.

Markelov, M. V., Golosov, V. N., & Belyaev, V. R. (2012). Changes in the sedimen-tation rates on the floodplains of small rivers in the central Russian plain.Vestnik Moskovskogo Universiteta, Seriya, 5. Geografiya 5, 70e76 [in Russian].

Medvedeva, R. A., Golosov, V. N., & Ermolaev, O. P. (2018). Spatio-temporalassessment of gully erosion in the zone of intensive agriculture in the Europeanpart of Russia. Geography and Natural Resources, 39(3), 204e211. https://doi.org/10.1134/S1875372818030034

Medvedev, I. F., Levitskaya, N. G., Makarov, V. Z., & Nazarov, V. A. (2016). The resultsof monitoring of erosion processes on chernozems of the Volga region [InRussian: Aktivnost’ Erozionnykh Protsessov Na Chernozemakh Povolzh’ya]Agrarian Scientific Journal [In Russian: Agrarny nauchny zhurnal], 8, 26e34, 2016.

National report. (2019) [in Russian: Natsional’nyy doklad “Global’nyy klimat i poch-vennyy pokrov Rossii: opustynivaniye i degradatsiya zemel’, institutsional’nyye,infrastrukturnyye, tekhnologicheskiye mery adaptatsii (sel’skoye i lesnoye kho-zyaystvo”)]. Global climate and soil cover of Russia: Desertification and land degra-dation, institutional, infrastructure, technological adaptation measures (agricultureand forestry) (Vol. 2019, p. 476). Moscow: Publishing House of MBA.


Nefedova, T. G. (2013) [In Russian: Desyat’ aktual’nykh voprosov o sel’skoy Rossii:otvety geografa]. Ten topical questions about rural Russia: The answers of ageographer (Vol. 2013, p. 456). Moscow: URSS-LENAND.

Orr, B. J., Cowie, A. L., Castillo, V. M. S., Chasek, P., Crossman, N. D., Erlewein, A., …Welton, S. (2017). Scientific conceptual framework for land degradation neutrality.Bonn, Germany: United Nations Convention to Combat Desertification(UNCCD). A Report of the Science-Policy Interface https://www.unccd.int/publications/scientific-conceptual-framework-land-degradation-neutrality-report-science-policy.

Petelko, A. I., Golosov, V. N., & Belyaev, V. R. (2007). Experience of design of systemof counter-erosion measures. In Proceedings of the 10-th international sympo-sium on river sedimentation (Vol. 1, pp. 311e316). Moscow.

Rosreestr. (2017). Information on the availability and distribution of land in the RF[In Russian: Svedenija o nalichii i raspredelenii zemel’ v RF]. Moscow https://rosreestr.ru/site/activity/sostoyanie-zemel-rossii/gosudarstvennyy-natsionalnyy-doklad-o-sostoyanii-i-ispolzovanii-zemel-v-rossiyskoy-federatsii/Accessed November 2019.

Rozhkov, V. A. (2007). Assessment of soil erosion hazard. [In Russian: Otsenkaehrozionnoj opasnosti pochv]. Dokuchaev Soil Bulletin, 59, 77e91 [in Russian].

Sidorchuk, A., Litvin, L., Golosov, V., & Cherhysh, A. (2006). European Russia andbyelorus. In J. Boardman, & J. Poesen (Eds.), Soil erosion in europe. Chapter 1.8.Chichester, England: John Wiley & Sons. https://doi.org/10.1002/0470859202.ch8.

Singh, V., & Dubey, A. (2012). Land use mapping using remote sensing & GIStechniques in naina -gorma basin, part of rewa district, M.P., India. InternationalJournal of Emerging Technology and Advanced Technology, 2, 151e156. https://www.researchgate.net/publication/308353798_Land_use_mapping_using_remote_sensing_GIS_techniques_in_Naina-Gorma_basin_part_of_Rewa_district_MP_India.

Solomun, M. K., Barger, N., Cerda, A., Keesstra, S., & Markovi�c, M. (2018). Assessingland condition as a first step to achieving land degradation neutrality: A casestudy of the republic of srpska. Environmental Science & Policy, 90, 19e27.https://doi.org/10.1016/j.envsci.2018.09.014, 2018.

Speranza, C. I., Adenle, A., & Boillat, S. (2019). Land degradation neutrality - po-tentials for its operationalisation at multi-levels in Nigeria. Environmental Sci-ence & Policy, 94, 63e71. https://doi.org/10.1016/j.envsci.2018.12.018, 2019.

State report. (2016). Moscow. On the state and use of land in the Russian Federation in2015 (Vol. 2016, p. 207) ([in Russian]).

State report. (2017). Moscow. On the state and environmental protection of theRussian Federation in 2016 (Vol. 2017, p. 760) ([in Russian]).

State soil erosion map of Russia (Asian part). (2004). 1: 2 500 000 (Vol. 2004).Dockuchaev Soil. Institute. Moscow.

State soil erosion map of Russia and neighboring countries (European part). (1999).1: 2 500 000 (Vol. 1999). Ekor. Moscow.

Trends.Earth. (2018). Conservation international. Available online at: http://trends.earth.2018.

UNCCD. (2015). ICCD/COP (12)/4. Report of the IWG on LDN. Integration of thesustainable development goals and targets into the implementation of theUNCCD. In Conference of the parties. Twelfth session. Ankara, Turkey. UNCCD.

Wunder, S., & Bodle, R. (2019). Achieving land degradation neutrality in Germany:Implementation process and design of a land use change based indicator.Environmental Science & Policy, 92, 46e55. https://doi.org/10.1016/j.envsci.2018.09.022, 2019.

Yermolaev, O. P., Golosov, V. N., Dvinskikh, A. P., Litvin, L. F., Kumani, M. V., &Rysin, I. I. (2014). REcent changes in sediment redistribution in the upper parts ofthe fluvial system of European Russia: regional aspects//IAHSeAISH proceedingsand reports “Sediment dynamics from the summit to the sea. Wallingford: IAHSPress.

Yermolayev, O. P., Rysin, I. I., & Golosov, V. N. (2017). Mapping assessment of gullyerosion in the east of the Russian Plain. [In Russian: Kartografirovanieovrazhnoj ehrozii na vostoke Russkoj ravniny]. Geomorphologiya, 2, 38e51 [inRussian].

Тrofimetz, L. N., Panidi, E. A., Ivanekha, T. L., & Petelko, A. I. (2017). Estimation oferosion activity in the gullie complex of arable slopes. [In Russian: Otsenkaehrozionnoj aktivnosti lozhbinnogo kompleksa raspakhivaemykh sklonov].Intercarto. Intergis., 23(1), 323e332 [in Russian].



The assessment of soil loss by water erosion in China

Baoyuan Liu a, Yun Xie a, *, Zhiguang Li b, Yin Liang c, Wenbo Zhang a, Suhua Fu a,Shuiqing Yin a, Xin Wei a, Keli Zhang a, Zhiqiang Wang a, Yingna Liu a, Ying Zhao d,Qiankun Guo d

a Faculty of Geographical Science, Beijing Normal University, Beijing, 100875, Chinab Soil Conservation Monitoring Centre, Ministry of Water Resources of People’s Republic of China, Beijing, 100053, Chinac Institute of Soil Science, Chinese Academy of Sciences, Nanjing, 210008, Chinad China Institute of Water Resources and Hydropower Research, Ministry of Water Resources of People’s Republic of China, Beijing, 100038, China


Article history:Received 17 March 2020Received in revised form26 June 2020Accepted 7 July 2020Available online 11 July 2020

Keywords:National soil erosion surveyCSLESample unitsChinese soil loss mapSoil erosion rateRatio of soil erosion area

a b s t r a c t

Soil erosion is a major environmental problem in China. Planning for soil erosion control requires ac-curate soil erosion rate and spatial distribution information. The aim of this article is to present themethods and results of the national soil erosion survey of China completed in 2011. A multi-stage, un-equal probability, systematic area sampling method was employed. A total of 32,948 sample units, whichwere either 0.2e3 km2 small catchments or 1 km2 grids, were investigated on site. Soil erosion rateswere calculated with the Chinese Soil Loss Equation in 10 m by 10 m grids for each sample unit, alongwith the area of soil loss exceeding the soil loss tolerance and the proportion of area in excess of soil losstolerance relative to the total land area of the sample units. Maps were created by using a spatialinterpolation method at national, river basin, and provincial scales. Results showed that the calculatedaverage soil erosion rate was 5 t ha�1 yr�1 in China, and was 18.2 t ha�1 yr�1 for sloped, cultivatedcropland. Intensive soil erosion occurred on cropland, overgrazing grassland, and sparsely forested land.The proportions of soil loss tolerance exceedance areas of sample units were interpolated through thecountry in 250 m grids. The national average ratio was 13.5%, which represents the area of land in Chinathat requires the implementation of soil conservation practices. These survey results and the mapsprovide the basic information for national conservation planning and policymaking.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction

Sustainable development, and especially sustainable land use, isvery important worldwide. In 2015, the United Nations (UN)adopted Sustainable Development Goals (SDGs), in which landdegradation caused by soil erosion is one of the main issues(Keesstra et al., 2018). In China, soil erosion is also a major envi-ronmental and ecological problem. Soil and water conservation isthe key to soil erosion control and development of an ecologicallystable environment and society. The Chinese government has madegreat effort toward soil and water conservation, with projects suchas Grain for Green (retiring cropland to forest and grass land) andthe bench-terrace construction project, which have reduced soil

erosion greatly. Sediment load in the Yellow River decreased from1.6 billion t yr�1 to approximately 0.2 billion t yr�1 (Liu et al., 2019).However, the status of soil loss as a function of geographic regionand land use should be well understood if soil and water conser-vation planning and relevant policies are to be efficiently made.

Formal scientific surveys in China began in the 1930s(Lowdermilk, 1943), mainly at local scales under the guidance of W.C. Lowdermilk (US soil conservationist). A large-scale survey star-ted in the 1950s in the Loess Plateau, known also as the YellowRiver basin. Three major national surveys were conducted in 1985(Zhou & Wu, 2005), 1999 and 2001 (Zeng & Li, 2000), which werepredominantly qualitative in nature. The fourth was a quantitativesurvey conducted in 2011 (Liu et al., 2013), which aimed at deter-mining the rate of soil loss, soil loss acreage, and their spatial dis-tributions; providing basic data for national soil and waterconservation planning, project resource allocation and policyformulation for construction of an ecologically stable society.* Corresponding author.

E-mail address: [email protected] (Y. Xie).






Many other countries and regions in the world have also con-ducted national and regional surveys. In the United States, soilerosion surveys began as early as 1935 (Goebel, 1998; Nusser &Goebel, 1997). After 1977, assessments of soil erosion rate andamount were carried out every five years (Harlow, 1994); and after2000 (USDA, 2009) they have been conducted as annual routinework. A Europe-wide assessment of soil erosion risk began in 2000and has been updated several times (Van der Knijff et al., 2000;Grimm et al., 2002; Knikby et al., 2004; Panagos et al., 2015). Othercountries have conducted national soil erosion surveys as well,such as India (Singh, Babu, Narain, Bhushan, & Abrol, 1992), Italy(Van der Knijff et al., 1999), Iceland (Arnalds, 2000), France(Bissonnais et al., 2001), Australia (Lu et al., 2001), and the UnitedKingdom (Evans et al., 2015), etc. This article will present the resultsof the Chinese soil erosion survey completed in 2011.

2. Methodology

2.1. Overview

A multi-stage, unequal probability, systematic area sampling,method was employed for the national soil erosion survey of 2011.One small catchment of 0.2e3 km2 or a square of 1 km2 wassampled within each square of 100 km2. Each of these wasconsidered to be a sample unit (SU), and a total 32,948 SUs weretaken. A field survey was made for each SU to collect data based on1:10,000 topographic maps and remote sensing images. The datacollected was land use and cover, vegetation canopy cover, groundcover, and soil conservation practices. The field survey data wastransformed into grids of 10 m by 10 m within each SU. Dailyrainfall and soil survey map information were identified forcalculating rainfall erosivity and soil erodibility at the nationalscale, which was then intersected with SU boundaries and resam-pled as 10m grids for use as erosion factors for each SU. Soil erosionrate for each 10 m by 10 m grid was calculated by multiplying the 7erosion factors of the Chinese Soil Loss Equation (CSLE, Liu, Zhang,& Xie, 2002) within each SU using the geographical informationsystem (GIS) software of ArcMap. The soil erosion rate of each 10 mgrid exceeding soil loss tolerancewas considered as “Soil Loss Area”(SLA). The SLA was summed up and then divided by the SU area,which determined the ratio, or proportion, of the SLA for each SU.The SLA-ratio is considered to be the proportion of the arearequiring attention for soil conservation measures. Finally, the soilerosion rate maps and the SLA-ratio maps were made in national,river basin, and provincial scales by spatially interpolating with theSU average values.

2.2. Sample design

A total of 32,948 SUs were sampled from the land area ofapproximately 6.32 million square kilometers. There were nosamples made on land of snow or ice cover, desert, and the Gobi,together constituting an area of approximately 3.28 million squarekilometers. The SUs were selected according to the 1:10,000topographic map divisions and numbering system developed bythe China Bureau of Surveying and Mapping (Fig. 1). First, eachsheet of the 1:10,000 topographic map was used as control area(CA), which was 5 by 5 km2 with a 1 by 1 km2 grid embedded in it,and one SU was selected at the center grid from each CA. The SUcould be either a catchment (0.2e3 km2) or a square (1 by 1 km2). Inmost hilly andmountainous areas, the SUwas the catchment. In flatareas where it would be a great challenge to locate a small catch-ment, the center 1 by 1 km2 square was selected as the SU instead.When a small catchment was used, it was selected to intersect withthe center grid. In this case the sampling density was 1/25, i.e., 4%.

This was the basic sampling density for the nation. If it wasnecessary to decrease the density for whatever reason, one SUwould be taken from the township level, transfer area, and evencounty-level, accordingly (Fig. 1).

2.3. Data collection

Two groups of data were collected. The first was field surveydata including land use and cover, soil conservation practices,canopy cover, and ground cover. The other was basic data preparedin the office, including rainfall, soil, DEM, and vegetation coverinterpreted from satellite images.

2.3.1. Field survey dataA base map and a satellite image of each SU were prepared for

the field survey. The base map was a digital topographic mapincluding serial numbers, contour lines, road, boundaries, latitudeand longitude grids. The satellite image had the same latitude andlongitude grids as did the base map. During field investigation, theinformation of the soil erosion polygons were recorded on a form(Fig. 2a), and the polygon boundaries were drawn on the base map(Fig. 2b). Pictures were taken to display features of each soil erosionpolygon, GPS coordinates (Fig. 2c), any typical or special conser-vation practices (Fig. 2d), and an overall view of the landscape ofthe SU (Fig. 2e). The soil erosion parcel (or soil erosion polygon onthe map) was defined as the uniform piece of land parcel with thesame type of land use and cover, same conservation measure, andsimilar canopy cover and ground cover (�10% difference).

2.3.2. Other dataDaily rainfall of 1981 through 2010 from 2678 stations were

collected, and only erosive daily rainfall, greater than 10 mm, wasused for calculating average annual rainfall erosivity and 24 ratiosof average half-month erosivity to annual erosivity for estimatingvegetation cover factors. Soil series maps and related features ofmore than 7764 soil profiles were used for calculating soil erod-ibility. The soil maps were digitized at a provincial scale of1:500,000 mainly, including 60 soil groups, 203 soil subgroups, and402 soil genera. The locations of the 7764 soil profiles wereassigned to a soil genus map with data of soil organic matter, clay(less than 0.002 mm), silt (0.002e0.02 mm), and sand(0.02e2 mm), soil structure, and permeability of the surface layers.NDVI (Normal Difference Vegetation Index) from HJ-1 satelliteMultispectral reflectance and from MODIS (Moderate-resolutionImaging Spectroradiometer), and national land use map of the year2010 in a scale of 1:100,000 were collected for calculating 24 half-

Fig. 1. A spatial representation of the sample unit selected.

B. Liu et al. / International Soil and Water Conservation Research 8 (2020) 430e439 431

month fractions of vegetation cover (FVC) from January toDecember. HJ-1 satellite NDVI data was for three periods ofdifferent seasons of the year for 2010 in a resolution of 30 m, andMODIS NDVI data (MCD43B4) was for continuous periods of every16 days every year from the year 2005 to the year 2010 at a reso-lution of 250 m grids. The fusion method was used to produce timesequence NDVI of 24 half-months for the year 2010 by using thesetwo datasets. First, the average NDVI seasonal variation curves wereextracted from MODIS data in 16-day temporal resolutions. Thenthe actual NDVI values of the year 2010 were obtained for threetime points within 24 half-months from HJ-1 satellite data in 30 mspatial resolutions. Finally, the fusion equation was used to esti-mate actual seasonal variation curves for every 30 m by 30 m pixel.Both of the MODIS land cover classification data (MOD12Q1) andland use map of the year 2010 covering China were used to identifyvegetation types through this processes.

2.4. Soil loss calculation for sample units

2.4.1. Model descriptionThe Chinese Soil Loss Equation (CSLE) was used to calculate soil

loss for each SU in a grid of 10 m by 10 mwith the ArcMap softwareof Geographical Information System (GIS).

The equation for the CSLE is (Liu et al., 2002):

A¼R,K,L,S,B,E,T (1)

Where A is soil loss in t ha�1 yr�1. R is rainfall erosivity in MJ mmha�1 h�1 yr�1. K is soil erodibility in t h MJ�1 mm�1. L and S aredimensionless topographic factors of the slope length and the slopesteepness. B is the dimensionless vegetation cover factor of bio-logical practices for trees, shrubs, and grasslands. E is the dimen-sionless factor of engineering practices such as terraces and fish-

scale pits. T is the dimensionless factor of tillage practices such ascrop rotation, contour tillage, residue cover, inter-cropping strips.Each of these factors are discussed individually below. Differentfrom the USLE (Wischmeier & Smith, 1978), factors of B, E, and T ofthe CSLE replace factors of C and P of the USLE where C is the coverand management factor and P is the support practice factor. Foreasy application, the vegetation cover factor was divided into non-cultivated land cover represented by the B factor and cultivatedland crop cover was represented by the T factor. The supportpractice factor P was divided into either manual labor or mecha-nized long-term practices represented by the E factor and tillagemanagement practices such as the contour tillage represented by Tfactor. Thus, a part of the C factor of the USLEwas grouped into the Tfactor of the CSLE, such as crop rotations, no till, residue manage-ment, and the P factor of the USLE was grouped into the E factor forpractices such as terraces and into the T factor for contour tillageand inter-cropping strips. The T factor of the CSLE includes bothcrop coverage effects due to crop rotations and conservation tillagepractices effects due to tillage management. All the factors of theCSLE have been calibrated using more than 2000 plot-years runoffdata observed in China.

2.4.2. Rainfall erosivity factor, RThe equations of average annual rainfall erosivity and 24 ratios

of half month erosivity to annual erosivity calculation using dailyerosive rainfall data were represented as Eqs. (2)e(4) (Xie et al.,2016):

Ryear ¼X24

j¼1

Rj (2)

Rj ¼1N

XN

i¼1

Xm

k¼0

�a , P1:7265i;j;k

�(3)

RRj ¼Rj�Ryear (4)

where Ryear is average annual rainfall erosivity in MJ,mm,hm�2,h�1,a�1, j represents the half month sequence of 1, 2,…, 24 ineach year, Rj is half month rainfall erosivity inMJ,mm,hm�2,h�1, i istime series of daily rainfall from1981 to2010, k is days of dailyerosiverainfall (rainfall equal or greater than 10mm)within each halfmonthperiod, Pi,j,k is 0whenno erosive rainfall occurs in the halfmonth, anda represents calibrated parameter values of 0.3957 for warmmonthsfrom May to September and 0.3101 for cool months from October toApril. RRj is the ratio of the average erosivity of half month j to theaverage annual erosivity. After calculations of R for 2678 stations,these were interpolated by kriging in 30 m by 30 m grids over all ofChina.

2.4.3. Soil erodibility factor, KSoil erodibility was estimated with the USLE equation shown by

Eq. (5) (Wichmeier and Smith, 1978), and calculated results weremodified by using observed data from runoff plot data.

K ¼Cm,�2:1�10�4ð12�OMÞM1:14 þ3:25ðs�2Þþ2:5ðp� 3Þ

�

�.759

(5)

where K is soil erodibility in t hm2 h hm�2 MJ�1 mm�1, Cm is acoefficient varying from 0.09 to 0.87 from the southeast to thenortheast of China. OM is the organic matter content in percent, M

Fig. 2. Field survey information collection for a sample unit: (a) Survey form; (b) TheGPS coordinates of one point within the unit; (c) The unit survey map; (d) The con-servation practice picture; (e) The unit landscape picture.

B. Liu et al. / International Soil and Water Conservation Research 8 (2020) 430e439432

is the textural factor with N1,(100 e N2) in percent, and N1 ispercent content of soil particles from 0.002 to 0.1 mm, and N2 isclay content percentage (less than 0.002 mm). S is the soil structureclass from 1 to 4 determined by soil textures, and p is the perme-ability class from 1 to 4 determined by soil structures and texture.After calculations of K for 7764 soil profiles with the polygons of soilmaps, they were resampled to 30 m by 30 m grids over all of China.

2.4.4. Slope length and steepness factors, LSThe slope length and steepness factors were calculated only

within the SUs in 10 m spatial resolutions after obtaining the DEM(Digital Elevation Model) by using digital contour maps at 1:10000scale.

The slope length factor was calculated by using a slope segmentlength equation, Eq. (6), (Foster & Wischmeier, 1974):

Li ¼lmþ1i � lmþ1

i�1ðli � li�1Þ,ð22:13Þm

(6)

where Li is segment slope length factor, li and li-1 are slope lengthfor segment i and segment i-1 in m, respectively, and m is theparameter varying with slope degree q, calculated by:

m¼

8>><>>:

0:2 q � 1�0:3 1� < q � 3�0:4 3� < q � 5�0:5 q>5�

(7)

The slope steepness factor was calculated by using the USLEequations, Eq. (8) (Wichmeier and Smith, 1978), and the equationdeveloped by Liu et al. (1994):

S¼8<:

10:8 sin qþ 0:03 q<5�16:8 sin q� 0:5 5� � q<10�21:9 sin q� 0:96 q � 10�

(8)

2.4.5. Vegetation and biological practice factor, BThe calculations of vegetation and biological practice factor B

were made at two spatial scales.One was at the country scale for calculating the fraction of

vegetation cover (FVC) of 24 half months from NDVI in 30 m by30 m grids using Eq. (9).

FVCj ¼NDVIj � NDVIsoil;j

NDVIveg;j � NDVIsoil;j(9)

where NDVIveg is NDVI values of pure vegetation pixels. NDVIsoil, isNDVI values of pure soil pixels, j is the sequence number of the 24half months.

Another was at the SU scale for calculating B values using Eq.(10) based on Polygon land use information of surveyed maps andforms:

B¼X24

j¼1

SLRj,RRj (10)

where SLRi is dimensionless soil loss ratio of soil loss from thevegetated land under the FVC and ground cover (GC) conditions inhalf month j to the calculated soil loss from bare land with all otherconditions the same, and RRj is the ratio of half month erosivity tothe annual erosivity. Thus, SLR values varied with vegetated landconditions.

For trees the equation used was:

SLRj ¼0:44468� exp��3:20096�GCj

��0:04099pt� exp

�FVCj � FVCj �GCj

�þ 0:025 (11)

For shrubs:

SLRj ¼1

1:17647þ 0:86242� 1:05905100�FVCj(12)

For grassland:

SLRj ¼1

1:25þ 0:78845� 1:05968100�FVCj(13)

For other land uses and covers, B values are given in Table 1.In the above equations, FVC is the fraction vegetation cover for

canopy calculated from NDVI in percent. GC is the ground surfacecover including green vegetation such as lichens, mosses, and lit-ters. Before the national soil erosion survey, more than one hun-dred GC seasonal variation curves in half-month intervals fortypical trees were observed for two years, such as natural andplanted forests and various fruit trees. If a polygon in a SU was fruittree land, its FVC would be obtained from corresponding FVC maplayers of 24 months for its canopy cover, and its GC would be ob-tained from the nearest GC seasonal curves and modified bymultiplying the ratio of observed GC to the value of the GC curve atthe survey time.

The procedures of estimating B factor with the GIS software aresummarized as: (1) developing FVC maps of 24 half-months in thecountry scale in 30 m grids; (2) Intersecting 24 half-months FVCgrid maps by digital SU survey maps with polygon boundaries andresampled in 10 m grids; (3) Calculating B factors in 10 m grids byidentifying polygon land uses, selecting equations, choosing andmodifying GC seasonal curves if the polygon represented tree land.

2.4.6. Engineering practice factor, EBoth runoff plot data and published literature were collected to

obtain the E values involving eleven engineering practices (Table 2).All the polygons within each SU were given E values from Table 2with engineering practices recorded or from Table 1 without en-gineering practice recorded. Then the vector E of the digital surveymap was transformed into grid maps in 10 m resolution in GISsoftware for all the SUs.

2.4.7. Tillage practice factor, TTillage practices influence soil loss from the two aspects of

management and the crop rotation, so the T factor is the product ofthe management sub-factor and rotation sub-factor. After review-ing the published literature, sub-management factor values weredetermined (Table 3). Based on cropland runoff plot observations,single crop factors of main crop types were obtained for soil loss offive crop growing periods, i. e., fallow, seedbed, establishment,development, residue and stubble. According to the Chinese systemof crop rotations, crop distributions, and crop growing periods, croprotation factors for eighteen regions were estimated (Guo et al.,2015), which were average annual rotated single crop factorsweighted by ratio of crop growth period rainfall erosivity to annualerosivity (Table 4). All the polygons of cultivated cropland withineach SU were given management sub-factor values from Table 3with tillage practices recorded, and given crop rotation sub-factorvalues from Table 4 according to the location. Then the vector T,product of two sub-factors, of the digital survey map was trans-formed into grid maps with 10 m resolution using GIS software forall the SUs.

2.4.8. Soil loss calculationsFor each SU, seven grid maps in 10 m resolutionwere multiplied


to obtain the soil erosion rate using GIS software. Each grid had asoil erosion rate in t ha�1 yr�1, while the deposited area that hadzero degrees of slope steepness or stone, water, and high coverageareas had little or no soil loss. The soil erosion rate of the SU wasaveraged of all 10 m grids for each SU. Soil erosion rate of every gridwas comparedwith soil loss tolerance, and then all of the grids withsoil erosion rate greater than tolerancewere summarized as the soilloss area for each SU. Furthermore, the soil loss areawas divided bythe SU area to obtain the proportion of soil erosion area, whichrepresents the area requiring soil erosion control.

2.5. Intensity assessment for province, river basin and nationalscales

The regional and national soil erosion rate was interpolatedusing an inverse distance weighting method in 250 m spatial res-olution using the average soil erosion rate of the SUs. Soil erosionrate was then classified into six grades to create the national soilerosion ratemap. The average proportions of excessive soil loss areaof each SU was used to make the national ratio of soil loss area mapby the same interpolation method. Then, this map was intersectedwith county and provincial boundaries. Computed total areas ofexcessive soil loss were the products of the excessive soil loss arearatios and land area for both provinces and counties, and the totalnational soil loss area was the sum of provincial soil loss areas.


3.1. Map of soil erosion rates in China

China is divided into three major climate-topographic region-s:eastern monsoon climate, northwestern dry climate, and Tibetplateau high cold region. Rainfall erosivity varies from the south-eastern coast with values greater than 10,000MJ mm ha�1 h�1 yr�1

to the northwestern inland of less than 100 MJ mm ha�1 h�1 yr�1.Within the eastern monsoon climate region, R value increases fromnorth to south. It was 1000e4000 MJ,mm,ha�1,h�1,yr�1 in thenortheast mainly north of the Liao River, between 1000 and6000 MJ mm ha�1 h�1 yr�1 in north China, and greater than6000 MJ mm ha�1 h�1 yr�1 in southern and southeastern Chinasouth of the Yangtze River. Resulting from monsoon and conti-nental climates, rainfall erosivity had a high percentage of annualvalues during the warm season from May to September, atapproximately 70e80%. Soil erodibility decreased from north tosouth due to clay contents decreasing and other influencing factors,varyingmainly from 0.035 to 0.007 t hMJ�1 mm�1. B values in mostmountainous regions covered by forest varied from 0.003 to 0.02,and from 0.003 to 0.11 for inland grass regions. Eighty-eight croprotation groups were classified, and their T values were deter-mined, which varied from 0.147 to 0.669 (Guo et al., 2015).

The average soil loss of all the SUs was 5.02 t ha�1 yr�1, and thetotal soil loss was 2.58 billion tons. The soil loss from cultivatedcropland, orchard, forestland, shrub land, grassland, and other landwas 1.64, 0.15, 0.13, 0.21, 0.26, 0.19 billion tons, respectively. How-ever, it varied greatly for different regions. To provide informationfor making soil conservation policies and plans, the spatial distri-bution map of the soil erosion rate was made by inverse distanceweighted interpolation, with the average soil erosion rate of each32,948 SU. The interpolated results were grouped into six grades(Fig. 3). The first grade was soil erosion rate less than 5 t ha�1 yr�1,and then 5e10, 10e15, 15e20, 20e25 for the second to the fifthgrades, and finally greater than 25 t ha�1 yr�1 for grade six. Theareal percentage of the six grades was 76.2%, 12.6%, 5.9%, 3.0%, 1.6%,and 0.7% of national land, respectively. The higher grades werelargely distributed in three, general regions (Fig. 3). One area ofhigh erosion was located in northeastern China, in the black soiland rolling sloped cropland area, which is one of the most impor-tant grain bases for maize and soybean in the nation. Another was

Table 1Values of the B factor for non-vegetated lands.

Land uses or covers B values Explanations

Urban developed area 0.006 Equivalent to land with 20% of FVC and 90% of SCRural developed area with annual rainfall �800 mm 0.016 Equivalent to land with 30% of FVC and 80% of SCRural developed area annual rainfall <800 mm 0.1 Equivalent to land with 30% of FVC and 40% of SCConstruction sites 1 Equivalent to bare landOther industry sites & Transportation land 0.018 Equivalent to land without canopy cover and 80% of SCField roads 0.073 Equivalent to land without canopy cover and 50% of SCBare land 1Water body 0

Table 2Values of E factor for engineering practices.

Bench terrace for paddy Bench terrace Inward slopping terrace Terrace with slope Segregated bench terrace Broad base terrace

0.01 0.102 0.151 0.414 0.347 0.414

Contour ditch for tree crops Big hole for tree crops Fish scale hole for tree planting Vegetation hedge Bamboo joint like ditch

0.335 0.16 0.249 0.347 0.8

Table 3T Values of the management sub-factor.

Contour farming Contour furrow ridge Contour strip cropping Contour ditch seeding Reduced tillage No till Mulch film Cover crop

0.431 0.425 0.225 0.213 0.212 0.136 0.5 0.225

Gravel cover field Grain crop grass rotation Checkboard- liked hole Pot seeding Ditch in the fallow field Intertill ridge rebuilt Cross slope ridge

0.2 0.225 0.152 0.499 0.425 0.499 0.7


the Loess Plateau, the noted serious soil loss area due to the heavystorms in summer, loose soils, the steep slopes, and long historicalintensive cultivation. The third was the Yungui Plateau and sur-rounding hills of Sichuan Basinwith very steep farmland, located insouthwestern China.

Soil erosionrate alsovariedgreatlyamongdifferent landuses andcovers. The cultivated cropland had the greatest soil erosion rate,varying from 0 t ha�1 yr�1 in the plain region to over 240 t ha�1 yr�1

in very steep slope land, with an average of 8.6 t,ha�1,yr�1. Orchardfollowed as the second, varying from 0 t,ha�1,yr�1 to over201 t,ha�1,yr�1, with an average of 6.1 t ha�1 yr�1. The soil erosionrate fromforestlandhad the least variation, from0 tha�1 yr�1 to over111 t ha�1 yr�1 at an average of 0.6 t ha�1 yr�1. The shrub land andgrassland had a similar rates, varying from 0 t ha�1 yr�1 to approx-imately 127 t ha�1 yr�1, with the average of 3.4 and 3.5 t ha�1 yr�1.The average rate from other lands was 2.6 t ha�1 yr�1.

3.2. Soil loss area and its spatial distribution map

Soil loss area was defined as the area where the soil erosion ratewas greater than soil loss tolerance (Tv) at a 10 m grid. Soil loss areafor each soil erosion polygon (land parcel with same land use andcover, same soil conservation practice and similar canopy andground cover) in each SU was summarized based on the 10 m griderosion rate. Tv was 2 t ha�1 yr�1 for cultivated cropland in bothnortheastern and northern China, and 5 t ha�1 yr�1 for cultivatedcropland in other regions and all other land uses and covers nation-wide. Soil loss intensity was classified into six grades for differentland uses and covers (Fig. 4). These were �Tv, Tv-2Tv, 2Tv-3Tv, 3Tv-4Tv, 4Tv-5Tv, >5Tv. Total soil loss area in China was 1,293,200 km2,13.5% of the national land. Of national erosion areas, 37.1% was fromcropland, 27.7% from forest, 20.0% from grassland,11.6% from shrub,and 3.6% from the others. When soil loss rate less than Tv it wasconsidered to be a non-soil loss area. The cultivated cropland hadthe lowest ratio of non-soil loss area, and the greatest ratio of highintensity classes (Fig. 4).

The soil loss area percentage was also estimated for seven largeriver basins, the greatest value was for the Yellow River Basin with28.2% of total basin land area, followed by Liaohe, Haihe, and Yangzibasins, with a similar percent of 23.3%, 22.9%, and 20.2%. Zhujiangriver basin had 19.9%, and Songhuajiang was 14.8%. These sevenriver basins are mainly in the eastern monsoon climate region, anddominated by water erosion. The northwestern dry climate areahad little water erosion (Fig. 5).

At the province scale, 10 provinces had soil loss areas percent-ages of the total province land greater than 20% (Table 5), 10provinces were between 10 and 20%, and 11 provinces had per-centages of less than 10%. Ten provinces had greater than 20%erosion area, and were distributed in north China, the LoessPlateau, and southwestern China (Fig. 5).

From the view of land uses and covers, although 28.3% ofcultivated cropland had terraces at the national scale, 32.7% of thenational erosion area was from cultivated cropland, especiallydistributed in northeastern and southwestern provinces. InGuangdong and Hainan provinces of southern China, the erosionarea came from orchard, because 48.8% and 14.8% of the totalprovince land was orchard in these two provinces, respectively. InXinjiang, Xizang, and Ningxia provinces in western China, the soilloss area was mainly in grassland.

3.3. Changes of soil loss areas

Compared with the results of the remote sensing survey for theyear 1999, the estimated water erosion area decreased by354,000 km2 in 2011, which is 29,500 km a�1, on average. Theerosion areas of most provinces decreased except the four prov-inces of Jilin, Guangdong, Hainan, and Guangxi (Fig. 6).

There were two reasons for this great reduction, one was adifferent method used for two surveys, and the other was soilconservation policies. An artificial identification method was usedfor the last survey, using only one factor of slope steepness forcropland and two factors of slope steepness and FVC for vegetatedlands. In the previous survey the slope degree was obtained by1:50,000 DEM, and FVC was estimated by TM images. The crop-lands with a slope degree greater than 5 were considered erosionareas. The vegetation lands with FVC greater than 75% wereconsidered non-eroding areas. If FVC was less than 75%, the vege-tation lands were or were not erosion areas depending on thecombination of slope steepness and vegetation cover. Difference inmethod accounts for the increases in the four provinces, which hadmany orchard lands for Hainan, Guangdong, and Guangxi insouthern China, and many long slope lengths but gentle slope de-grees for cropland for Jilin in northeastern China. In addition,conservation practices, especially terraces, were considered in thissurvey but not for the last survey. During 1999 through 2011, theChinese government implemented many conservation programs,which no doubt decreased soil erosion, such as Grain for Green inthe Loess Plateau.

Table 4T Values of crop rotation sub-factor.

Zones T Zones T Zones T Zones T Zones T Zones T

I1-1 0.41 I2-10 0.38 II1-1 0.34 II3-1 0.35 III1-1 0.38 III3-2 0.48I1-2 0.44 I2-11 0.44 II1-2 0.42 II3-2 0.42 III1-2 0.38 III3-3 0.47I1-3 0.35 I2-12 0.24 II1-3 0.44 II3-3 0.36 III1-3 0.24 III3-4 0.15I1-4 0.38 I3-1 0.38 II1-4 0.42 II3-4 0.43 III1-4 0.27 III3-5 0.15I1-5 0.36 I3-2 0.36 II1-5 0.48 II3-5 0.15 III1-5 0.36 III3-6 0.16I1-6 0.26 I3-3 0.51 II1-6 0.27 II3-6 0.16 III1-6 0.15 III3-7 0.27I1-7 0.3 I3-4 0.63 II1-7 0.32 II3-7 0.48 III1-7 0.16I1-8 0.15 I3-5 0.56 II1-8 0.44 II3-8 0.46 III1-8 0.21I2-1 0.49 I4-1 0.18 II1-9 0.35 II4-1 0.37 III2-1 0.39I2-2 0.49 I4-2 0.36 II2-1 0.31 II4-2 0.44 III2-2 0.31I2-3 0.46 I5-1 0.3 II2-2 0.42 II4-3 0.36 III2-3 0.38I2-4 0.5 I5-2 0.27 II2-3 0.33 II4-4 0.44 III2-4 0.15I2-5 0.44 I5-3 0.3 II2-4 0.39 II4-5 0.15 III2-5 0.16I2-6 0.51 I5-4 0.31 II2-5 0.15 II4-6 0.16 III2-6 0.17I2-7 0.33 I5-5 0.32 II2-6 0.16 II4-7 0.47 III2-7 0.15I2-8 0.36 I5-6 0.19 II2-7 0.4 III2-8 0.16I2-9 0.37 II2-8 0.41 III3-1 0.44


4. Conclusions

There are two general methods of conducting soil erosion sur-veys. One is the sampling method introduced in this article and

used in the national soil erosion survey of the Natural ResourcesInventory by the Department of Agriculture of the United States(USDA, Goebel, 1998; Nusser & Goebel, 1997). The other is agridding method used in assessments of European soil erosion

Fig. 3. Soil erosion rates in China.

Fig. 4. Soil Loss Area in different intensities of different land uses and covers. Tv is soil loss tolerance, which was 2 t ha�1 yr�1 for cultivated cropland in both northeast and northChina, and 5 t ha�1 yr�1 for cultivated cropland in other regions and all other land uses and covers.


Fig. 5. Ratio of Soil Loss Area (proportion of area exceeding tolerable) in China.

Table 5Erosion areas and percentages.

Province Erosion area km2 Percent of erosion area to total erosion area

Cultivated Cropland Orchard Forest Shrub Grass Others

Nation 1293246 32.7 4.4 27.7 11.6 20.0 3.5

Anhui 13899 21.9 7.3 63.0 4.1 0.5 3.2Beijing 3202 7.7 8.7 33.9 46.5 1.0 2.2Chongqing 31363 51.8 2.6 30.0 9.8 4.3 1.5Fujian 12181 12.2 8.9 66.6 4.2 0.9 7.2Gansu 76112 36.0 1.2 10.1 7.7 43.0 1.9Guangdong 21305 10.8 30.5 36.1 15.2 2.2 5.1Guangxi 50537 25.4 9.0 36.9 21.1 6.0 1.5Guizhou 55269 51.9 3.0 23.6 14.6 5.7 1.3Hainan 2116 32.9 61.2 1.8 1.6 0.4 2.0Hebei 42135 24.7 4.8 9.4 41.5 16.4 3.2Heilongjiang 73251 78.3 0.3 16.6 1.2 2.6 1.1Henan 23464 34.7 2.5 45.8 10.6 4.6 1.8Hubei 36903 24.9 3.3 50.5 17.8 2.5 1.0Hunan 32288 15.6 5.1 61.6 10.8 3.4 3.6Jiangsu 3177 60.9 4.6 20.1 0.5 6.0 7.9Jiangxi 26497 6.3 5.1 79.5 4.4 2.0 2.6Jilin 34744 67.9 0.8 26.5 1.0 2.3 1.5Liaoning 43988 48.5 10.5 22.3 10.0 6.1 2.7Neimenggu 102398 38.8 0.1 19.8 1.1 36.7 3.5Ningxia 13891 24.1 0.1 3.3 7.3 63.6 1.5Qinghai 42805 7.8 0.0 11.0 18.1 54.1 9.0Shandong 27253 59.3 11.9 15.7 2.2 8.2 2.6Shanghai 4 96.8 0.0 0.0 0.0 0.0 3.2Shannxi 70807 18.1 3.1 29.1 21.9 27.1 0.7Shanxi 70283 23.1 3.3 14.9 19.3 37.7 1.7Sichuan 114420 31.3 2.2 43.8 10.3 7.9 4.5Tianjin 236 14.1 11.3 33.3 41.1 0.0 0.1Xinjiang 87621 1.3 0.1 1.5 3.6 81.0 12.5Xizang 61602 10.1 0.0 10.6 22.8 43.1 13.3Yunnan 109588 34.4 10.1 33.8 11.0 8.7 2.1Zhejiang 9907 14.6 14.0 62.5 1.8 0.3 6.8


(Panagos et al., 2015) or for assessing global impact of land usechange on soil erosion (Borrelli et al., 2017), which requires inves-tigation of entire areas. Both the China and US surveys used thesampling method, but the distributions and sizes of the sampleunits were different between them. A systematic (uniform) sam-pling method was used in Chinese survey, while a random sam-pling method was used in the US. A small watershed was surveyedas a sample unit and soil erosion rate for thewhole sample unit wasestimated in China, while only a slope profile was surveyed and soilloss from the slope was estimated in the US.

The average rate of soil erosion across the country was5.02 t ha�1 yr�1, and was 8.61 t ha�1 yr�1for cultivated croplandsincluding plain fields, rice paddies, and terraces. If only the slopedcultivated cropland was considered, the estimated soil erosion ratewas 18.17 t ha�1 yr�1. The most serious soil erosion problem camefrom the sloped cultivated cropland. The other soil erosion problemareas were orchard, overgrazed grassland, and some sparselyplanted tree fields.

From the view of spatial distribution, three regions were foundwith more soil loss area ratios than other regions, i.e., northeastChina, the Loess Plateau, and southwest China. In north-easternChina, the serious soil loss came from the long sloped, cultivatedcroplands. In the Loess Plateau, the soil erosion occurred in thesteep slope croplands including fruit trees such as apple tree fields.In the southwest China, more soil erosion was from steep slopecroplands.

The sustainability of human societies depends on thewise use ofnatural resources. Soils contribute to basic human needs (Keesstraet al., 2016). To make national policies and plans for soil conser-vation, it is necessary to identify where the soil erosion problemsare, which means knowing where soil erosion rate is exceeding thesoil loss tolerance. Since this investigation, the Chinese governmenthas paid much more attention to soil conservation in slopedcultivated cropland and orchards, etc., especially in the northeastblack soil region of China, because most government officialsbelieved previously that soil erosion there was slight due to the flatlandform. In addition, national soil conservation planning for yearsof 2015 through 2030 was issued in 2015 according to the surveyresults.

A sampling method was recommended for the regional soilerosion survey which could collect the detailed information on asampling unit, and accurately estimate its soil erosion rate. A smallwatershed, which can represent the local soil erosion better than aslope profile, will be recommended as a sample unit.


None.

Acknowledgements

This research is supported by the National Key R&D Program2018YFC0507000.

References

Arnalds, O. (2000). The Icelandic “rofabard” soil erosion features. Earth SurfaceProcesses and Landforms, (25), 17e28.

Bissonnais, Y. L., Montier, C., Jamagne, M., Daroussin, J., & King, D. (2001). Mappingerosion risk for cultivated soil in France. Catena, (46), 207e220.

Borrelli, P., Robinson, D. A., Fleischer, L. R., Lugato, E., Ballabio, C., Alewell, C., et al.(2017). An assessment of the global impact of 21st century land use change onsoil erosion. Nature Communications. https://doi.org/10.1038/s41467-017-02142-7

Evans, R., Collins, A. L., Foster, I. D. L., Rickson, R. J., Anthony, S. G., Brewer, T., et al.(2015). Extent, frequency and rate of water erosion of arable land in Britain ebenefits and challenges for modelling. Soil Use & Management, 1e13.

Foster, G. R., & Wischmeier, W. H. (1974). Evaluating irregular slopes for soil lossprediction. Transactions of the American Society of Agricultural Engineers, 17,305e309.

Goebel, J. J. (1998). The national resources inventory and its role in US agriculture//Conference on agricultural statistics. In Proceedings of Agricultural Statistics 2000(pp. 181e192). Washington, DC: USDA.

Grimm, M., Jones, R., & Montanarella, L. (2002). Soil erosion risk in Europe (p. 40).European Soil Bureau, Institute for Environment & Sustainability, JRC Ispra.

Guo, Q., Liu, B., Xie, Y., Liu, Y., & Yin, S. (2015). Estimation of USLE crop and man-agement factor values for crop rotation systems in China. Journal of IntegrativeAgriculture, 14(9), 1877e1888.

Harlow, J. T. (1994). History of natural resources conservation service national re-sources inventories. Washington, DC: USDA.

Keesstra, S. D., Bouma, J., Wallinga, J., Tittonell, P., Smith, P., & Bardgett, R. D. (2016).The significance of soils and soil science towards realization of the UnitedNations Sustainable Development Goals. Soil, 2, 111e128.

Keesstra, S., Mol, G., de Leeuw, J., Okx, J., de Cleen, M., & Visser, S. (2018). Soil-relatedsustainable development goals: Four concepts to make land degradation

Fig. 6. Changes from remote sensing survey in 1999 to the sampling survey in 2011.


neutrality and restoration work. Land, 7(4), 133.Kirkby, M. J., Jones, R. J. A., Irvine, B., Gobin, A., Govers, G., Cerdan, O., … Huting, J.

(2004). Pan-European soil erosion risk assessment: The PESERA map, version 1October 2003. Explanation of special publication Ispra 2004 No.73 (S.P.I.04.73).European soil Bureau research report No.16, EUR 21176, 18pp. And 1 map in ISO B1format. Luxembourg: Office for Official Publications of the EuropeanCommunities.

Liu, B., Guo, S., Li, Z., et al. (2013). Sampling method of water erosion survey inChina. Soil and Water Conservation in China, (10), 26e34 (In Chinese).

Liu, B., Nearing, M. A., & Risse, L. M. (1994). Slope gradient effects on soil loss forsteep slopes. Transaction of American Society of Agriculture Engineers, 37(6),1835e1840.

Liu, B., Tang, K., Jiao, J., Ma, X., Zhang, X., Cao, Q., et al. (2019). Temporal and spatialatlas of runoff and sediment in the Yellow River (2nd ed.). Beijing: Science Press(In Chinese).

Liu, B., Zhang, K., & Xie, Y. (2002). An empirical soil loss equation//Proceedings-Processof soil erosion and its environment effect, 12th international soil conservation or-ganization conference (pp. 21e25). Beijing: Tsinghua University Press.

Lowdermilk, W. C. (1943). Preliminary report on findings of a survey of a portion of thenorthwest for a program of soil, water and forest conservation (p. 58).

Lu, H., Gallant, J., Prosser, I. P., et al. (2001). Prediction of sheet and rill erosion over theAustralian continent, incorporating monthly soil loss distribution. CSIRO Land andWater Technical Report.

Nusser, S. M., & Goebel, J. J. (1997). The national resources inventory: A long-termmulti-resource monitoring programme. Environmental and Ecological Statistics,4(3), 181e204.

Panagos, P., Borrelli, P., Poesen, J., Ballabio, C., Lugato, E., Meusburger, K., et al.

(2015). The new assessment of soil loss by water erosion in Europe. Environ-mental Science & Policy, 54, 438e447.

Singh, G., Babu, R., Narain, P., Bhushan, L. S., & Abrol, I. P. (1992). Soil erosion rates inIndia. Journal of Soil and Water Conservation, 47(1), 97e99.

U.S. Department of Agriculture. (2009). Summary report: 2007 national resourcesinventory (p. 123). Ames, Iowa: Natural Resources Conservation Service,Washington, DC, and Center for Survey Statistics and Methodology, Iowa StateUniversity. http://www.nrcs.usda.gov/technical/NRI/2007/2007_NRI_Summary.pdf.

Van der Knijff, J. M., Jones, R. J. A., & Montanarella, L. (1999). Soil erosion riskassessment in Italy (p. 54). European Soil Bureau, Joint Research Center of theEuropean Commission.

Van der Knijff, J. M., Jones, R. J. A., & Montanarella, L. (2000). Soil erosion riskassessment in Europe (p. 34). European Soil Bureau, Joint Research Center of theEuropean Commission.

Wischmeier, W. H., & Smith, D. D. (1978). Predicting rainfall erosion losses: A guideto conservation planning. In U.S. Dep. Agric., Agric. Handbook No.537. Washing-ton, D.C.

Xie, Y., Yin, S., Liu, B., Nearing, M. A., & Zhao, Y. (2016). Models for estimating dailyrainfall erosivity in China. Journal of Hydrology, 535, 547e558.

Zeng, D., & Li, Z. (2000). Method and Thinking of the second national remotesensing soil erosion survey. Soil and Water Conservation in China, (1), 28e31 (InChinese).

Zhou, W., & Wu, B. (2005). Overview of remote sensing approaches to soil erosionmonitoring. Remote Sensing Technology and Application, 20(5), 537e542 (InChinese).



Study on a soil erosion sampling survey in the Pan-Third Pole regionbased on higher-resolution images

Qinke Yang a, *, Mengyang Zhu a, Chunmei Wang a, Xiaoping Zhang b, Baoyuan Liu b, c,Xin Wei c, Guowei Pang a, Chaozhen Du a, Lihua Yang b

a College of Urban and Environment, Northwest University, Xian, Shaanxi, 710126, Chinab Institute of Soil and Water Conservation, Chinese Academic Sciences and Ministry of Water Resources, Yangling, Shaanxi, 712100, Chinac College of Geography, Beijing Normal University, Beijing, 100875, China


Article history:Received 8 March 2020Received in revised form19 July 2020Accepted 20 July 2020Available online 24 July 2020

Keywords:Pan -third pole areaLand useSoil conservation measuresRemote sensingVariable probability sampling

a b s t r a c t

Soil erosion is one of the most severe global environmental problems, and soil erosion surveys are thescientific basis for planning soil conservation and ecological development. To improve soil erosionsampling survey methods and accurately and rapidly estimate the actual rates of soil erosion, a Pan-ThirdPole region was taken as an example to study a methodology of soil erosion sampling survey based onhigh-spatial-resolution remote sensing images. The sampling units were designed using a stratifiedvariable probability systematic sampling method. The spatiotemporal characteristics of soil erosion andconservation were taken into account, and finer-resolution freely available and accessible images inGoogle Earth were used. Through the visual interpretation of the free high-resolution remote sensingimages, detailed information on land use and soil conservation measures was obtained. Then, combinedwith the regional soil erosion factor data products, such as rainfall-runoff erosivity factor (R), soilerodibility factor (K), and slope length and steepness factor (LS), the soil loss rates of some sampling unitswere calculated. The results show that, based on these high-resolution remote sensing images, the landuse and soil conservation measures of the sampling units can be quickly and accurately extracted. Theinterpretation accuracy in 4 typical cross sections was more than 80%, and sampling accuracy, describedby histogram similarity in 11 large sampling sites, show that the landuse of sampling uints can representthe structural characteristics of regional land use. Based on the interpretation of data from the samplesurvey and the regional soil erosion factor data products, the calculation of the soil erosion rate can becompleted quickly. The calculation results can reflect the actual conditions of soil erosion better than thepotential soil erosion rates calculated by using the coarse-resolution remote sensing method.© 2020 International Research and Training Center on Erosion and Sedimentation and China Water andPower Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-


1. Introduction

Soil is a major environmental element and a nonrenewablenatural resource. The provision of ecosystem services on whichhumans depend, including food provision, require healthy soil tosupport and regulate them (Amundson et al., 2015; Montanarella,2015; Robinson et al., 2017; Morgan, 2005). However, soils inmost parts of the world are threatened by erosion (Oldeman et al.,1991; FAO, 2019a).

As one of the global environmental problems (Crosson et al.,

1995; Montanarella, 2015; FAO, 2019a), soil erosion not only canlead to land degradation, reduced ecological service levels andbiodiversity loss (on-site impact) but also can cause river andreservoir/lake siltation downstream, thus endangering regionalsustainable development (off-site impact, FAO, 2019a). At the sametime, soil erosion possibly accelerates soil carbon loss, increasingatmospheric CO2, thereby affecting global change (global impacts,Lal, 2019). The International Symposium on Soil Erosion (GSER19)and the Sixth International Soil Day (Dec 5, 2019) having thecommon topic of “stop soil erosion, save our future” aimed toimprove soil health by encouraging farmers, social groups andgovernments to actively participate in soil erosion control (FAO,2019b).

To scientifically and effectively control soil erosion, the location* Corresponding author. st Xuefu Street, Xi’an, Shaanxi, 710127, PR China.

E-mail address: [email protected] (Q. Yang).






and extent of soil erosion occurrence, the current situation andbenefits of soil conservation measures, and the dominant factorsaffecting soil erosion must be investigated and understood.Therefore, a quantitative assessment of regional soil erosion mustbe conducted (Oldeman et al., 1991; Crosson et al., 1995; FAO,2019b). Since the 1990s, a series of studies on assessing and map-ping soil erosion have been carried out. Based on these studies,some valuable datasets have been collected in countries and re-gions (Table 1). These studies have provided a good knowledge ofthe macro-regional pattern of soil erosion and hotspots, and themain drivers of soil erosion, and therefore are of great significancein supporting soil conservation policies (Oldeman et al.,1991, 1994;Yang et al., 2006; Borrelli et al., 2017). However, the studiesmentioned above were mostly based on low- and medium-resolution remote sensing imagery and small cartographic scalegeographic information system (GIS) data products, so they havemainly provided potential soil erosion to a certain degree (Borrelliet al., 2017; Teng et al., 2018) but cannot fully meet the re-quirements of a comprehensive understanding of soil erosion andrelated decision making.

From the perspective of soil erosion processes and mechanics,soil erosion by water occurs at a hillslope (Morgan, 2005; Weil &Brady, 2017; Liu et al., 2018). Therefore, soil erosion assessmentshould be based on field investigation, high-resolution remotesensing and/or detailed cartographic data, and the soil erosion in-formation finally obtained at the hillslope scale. Therefore, soilsurvey sampling methods have been developed in the previousstudies.

The “Natural Resources Conservation Service, United StatesDepartment of Agriculture” (USDA-NRCS) uses a stratified two-stage spatial sampling scheme to design sampling units andrandom samples (a hillslope) in the National Resources Inventory(“NRI”). Based on field surveys and remote sensing interpretationmethods, datawere collected for all of sampling units and samplingpoints, and Universal Soil Loss Equation (USLE) or Revised UniversalSoil Loss Equation (RUSLE) was used to calculate the soil loss rate ofeach sample point to reflect the soil loss on the nation’s nonfederallands in the United States (Nusser et al., 1988; Nusser & Goebel,1997; USDA, 2018). After some adjustment of the NRI methodsaccording to the soil erosion and conservation characteristics ofChina, in the China Water Resources Inventory (“WRI”), the surveyunits (the unit is a small watershed with an area of 0.5e2 km2)were designated using a stratified unequal probability systematicsampling method, and soil loss rate was calculated using the Chi-nese Soil Loss Equation (CSLE) model (Liu et al., 2002, pp. 143e149).The equation for the CSLE is:

A¼RKLSBET (1)

where A is soil loss in t km�2 a�1. R is rainfall erosivity in MJ mmha�1 h�1 yr�1. K is soil erodibility in t h MJ�1 mm�1. L and S aredimensionless topographic factors of the slope length and the slopesteepness. B is the dimensionless vegetation cover factor of bio-logical practices for trees, shrubs, and grasslands. E is the dimen-sionless factor of engineering practices such as terraces and fish-scale pits. T is the dimensionless factor of tillage practices such ascrop rotation, contour tillage, residue cover, inter-cropping strips.

WRI uses field survey-based and remote sensing-assisted tech-nologies to collect data on land use and soil conservation measures,which are then combined with regional soil erosion factor dataproducts (RKLS and biological factor, B). The land use, soil conser-vation measures, and calculated soil erosion rates obtained by thesampling survey method are closer to the actual soil loss rate.Therefore this method can provide stronger support for the plan-ning of soil conservation, dynamic analysis of human factor impactsand soil erosion (Liu et al., 2013; Yin et al., 2018).

There are two problems to be urgently solved in soil erosionsurveys and mapping studies. One is how to collect the informationabout dominant factors of soil erosion quickly and accurately at thelocal scale (hillslope or small watershed) and then estimate the soilerosion rate. The second is how to make a map of soil erosion at theregional scale, so that the macro differentiation and local variationsin soil erosion can be well represented and analyzed simulta-neously. The unit for assessing the soil erosion rate in the NRI is ahillslope; however, because of its small area, it is not an idealapproach in steep areas and the field work currently adopted by theWRI method covers a relatively large area. In addition, neither theNRI norWRI methods are used outside the USA or China. Therefore,how to obtain soil loss rate data at the sampling level quickly andaccurately across continents and even at a global scale is still amajor issue that needs further study.

This research focuses on the first of the two issues mentionedabove, i.e. how to collect the information about dominant factors onerosion and then estimate the soil erosion rate for the samplingunit. The paper stems from our previous work, mainly focused onthe method of interpretating landuse and soil conservation mea-sures for sampling units (Yang et al., 2019; Zhu et al., 2019). Thispaper first briefly introduces the sampling unit designing andinterpretation of landuse and soil conservation measures, thendiscusses the calculation methods of sampling unit interpretationaccuracy, regional representativeness, and finally discusses theapplicability of the interpretation results in calculation of soil lossrate of selected units. The purpose of this study is to develop a rapid

Table 1Main researches on regional soil erosion mapping.

author date Method resolutionor map scale Geographical extent data accessibility

Zhu, 1965; Zhu et al., 1999 19565, 1999 artificial compilation 1:15000 million China available (1)

Oldeman et al., 1991; Oldeman et al., 1994 1991, 1994 artificial compilation 1:10000 million Global available (2)

Batjes, 1996 1996 RS, GIS, simple model 0.5 deg Global not availableLu & Yu, 2002 2002 RS、GIS、RUSLE 1000 m Australia not availableYang et al. (2003) 2003 RS、GIS、RUSLE 1000 m Global not availableLiu et al. (2013) 2013 Soil erosion map of China not availableLi et al. (2014) 2014 CSLE based soil erosion mapping 1000 m China Atlas (available)Panagos et al. (2015) 2015 RS、GIS、RUSLE 1000 m Pan-EU not availableBosco et al. (2015) 2015 RS、GIS、RUSLE 1000 m Pan-EU not availableTeng et al. (2016) 2016 RS、GIS、RUSLE 250 m Australia not availableBorrelli et al. (2017) 2017 RS、GIS、RUSLE 250 m Global available (3)

USDA, 2018 2018 sampling þ USLE/RUSLE ? US not available

(1) Apply through the Loess Plateau Scientific Data Center (http://loess.geodata.cn/), or apply to the author ([email protected])(2) https://www.isric.org/projects/global-assessment-human-induced-soil-degradation-glasod.(3) https://esdac.jrc.ec.europa.eu/content/global-soil-erosion.

Q. Yang et al. / International Soil and Water Conservation Research 8 (2020) 440e451 441

and practical sampling survey method of soil erosion based onhigh-resolution remote sensing interpretation and provide supportfor large-scale and even global-scale soil erosion surveying andmapping. The highlights of this method are its high accuracy, goodapplicability, and high efficiency.

2. Data and methods

2.1. Study area

The Pan-Third Pole (PTP) region, with a total area of approxi-mately 5.14 � 107 km,2 covers the Qinghai-Tibet Plateau and itssurrounding areas, geographically including Eurasia, a small part ofCentral and Eastern Europe andNorth Africa (Fig.1; Yao et al., 2017).This region is an ideal study area for soil erosion and conservationdue to its large area (37% of the global land area) and the variety oftypes of soil erosion, such as strong and extensive water erosion onthe Loess Plateau, wind erosion in arid regions in northwest Chinaand central Asia, and freeze-thaw erosion in the Qinghai-TibetPlateau and high-latitude region (Lin et al., 2017; Ma et al., 2018;Teng et al., 2018; Zhang et al., 2019; Chen et al., 2020). The envi-ronment is fragile, and the economy is relatively underdeveloped.This study can also provide data support for soil erosion controlplanning in the PTP region, as well as accumulate data and expe-rience for global soil erosion mapping.

2.2. Data sources

The datasets used in this study can be divided into two groups(Table 2): (1) Basic data: including remote sensing images for theinterpretation of sampling units and regional soil erosion factors forthe calculation of the unit soil erosion rate. (2) Auxiliary data: thesedata indirectly support the remote sensing interpretation of sam-pling survey units and the calculation and analysis of unit soilerosion rate, including the global 1 arc-second resolution Shuttle

Radar Topography Mission (Farr et al., 2007),1 the global 30-m-resolution land cover map (GLC30) (Ban et al., 2015), etc.

2.3. Methods for the sampling survey

The soil erosion sampling survey (PTP sampling) in this studyrefers to the processes of designating sampling units, extractinginformation about soil erosion factors (such as land use and soilconservation measures) on each sampling unit, calculating the soilloss rate of each sampling unit and making the necessary analysis.

2.3.1. Sampling surveying unit designing and informationextraction

The sampling survey unit (sampling unit, SU) for soil erosionsampling survey is a small geographical areawith specified locationand area, basic characteristics of soil erosion and spatial hetero-geneity in soil erosion factors (especially soil and water conserva-tion measures) (Zhu et al., 2019).

In this study, a two-stage, unequal probability, systematic areasampling method developed by the WRI (Liu et al., 2013; Xie et al.,2019) is used to designate the sampling units (SUs). A sample unit(SU) was designed within each of the regions in different latitude,with specified latitude and longitude interval (Table 3), and a total20880 SUs were designed in first stage. Then more sampling unitswere designated on the Qinghai-Tibet Plateau and its surroundingareas, such as Xinjiang, Qinghai, Gansu, Sichuan and Yunnan, China.This designing is performed by inserting an extra SU between twoSUs in the first stage, a total 11,041 SUs were designed. As a result,more than 31,921 sampling units have been designated in the PTParea, exact longitude and latitude coordinates of the centre pointfor each of the SUs were specified.

In soil erosion survey practices, the sampling unit of the soilerosion sampling survey is a small basin just surrounding the SUpoint, which can be clearly identified on the ground or with highresolution images (or DEM), and the suitable unit area can be

Fig. 1. Position of The Pan-Third Pole and distribution of large sampling area. Note: the black boxes on the map are large sampling sites for Representative analysis of theinterpretation results (in section of 3.3); they are: LH – loess hilly (northwestern China), BSUH– block soil undulated hilly (northeastern China), RSH– red soil hilly (southern China),NTP–northern TP area (China, southeastern), STP–southern TP area (China, southern), DM–dry mountain area (northwestern China), HM–humid mountain area (northern Thailand),GL–the Great Lakes region (central Kazakhstan), MEM–the Middle East mountain (central Iran), EU–rolling plain in eastern Europe (central Ukraine), CP–cold plain (central Russia).

2 Center of Soil Conservation Monitoring of MWR, Guide of Soil Erosion Moni-toring (draft), 2018.8.

1 https://www.usgs.gov/centers/eros/science/usgs-eros-archive-digital-elevation-shuttle-radar-topography-mission-srtm-1-arc.

Q. Yang et al. / International Soil and Water Conservation Research 8 (2020) 440e451442

0.2e3 km2 (the area can be slightly larger if high-resolution im-agery is not available). In relatively flat, a rectangular area of 1 km2

(4 km2 if a high-resolution image is not available) can be used assampling unit (Liu et al., 2013; Zhu et al., 2019).

To calculate the soil erosion rate for SUs, this study extractedinformation on land use and soil conservation measures based onhigh-resolution images for the sampling units. Among them, soilconservation measures include biological measures (e.g., manmadegrassland and forest), engineering measures (e.g., sand sand-checkdams and terraced cropland) and tillage measures (e.g., contourtillage, and crop rotation). Considering the limitation of remotesensing interpretation, tillage measures have not been consideredat this stage, but records are taken in the field check. The classifi-cation systems for land use and soil conservation measures pro-posed by theWRI (MWR, 2010.11; Liu et al., 2013) are applied. In theinformation extraction process, a template file prepared by theauthor’s team is first loaded, then visual interpretation is per-formed based on the downloaded images in ArcMap (Law& Collins,2018), and the attributes are directly assigned to the polygon andline features (Tables 4 and 5). Mainly based on the practices of

China’s WRI and after some testing, the minimum area (MA) wasdecided for different categories in the interpretation. For con-struction land, water bodies, and transportation land, the MA is400 m2; for cropland, orchard the MA is 600 m2; for forestland andgrassland, the MA is 2500 m2. The minimum width of linear fea-tures (roads, rivers) is 20 m (Zhu et al., 2019).

2.3.2. Field investigationIn southeast Qinghai, China; southeastern Tibet, China; eastern

Xinjiang, China; northern Thailand; northern Pakistan; and centralKazakhstan, cross sections are designated along the direction of themaximum gradient of the dominant soil erosion factor. The domi-nant factor is elevation which equivalent vegetation in Tibet, Xin-jiang and Thailand, and precipitation in Pakistan (Fig. 2). Then,sampling units have been defined and sampled along the crosssections approximately every 10 km and interpreted in the labo-ratory. Additionally, observations on the characteristics of soilerosion factors were performed in the field to help understand theprocess of soil erosion, establish interpretation marks of typicalground features, and check the previous interpreted results.Because the cross sections in Qinghai and Tibet are all located onthe Qinghai-Tibet Plateau, so just selected Tibet section to analyze;

Table 2Summary list of the datasets in this study.

datasets resolution Source Simple description

remote sensing images 0.5 m, 16 m Downloaded from Google Earth high resolution images are mainly quickbird, partly are IKNOS; medium resolution images aremainly landsat Thematic Mapper (TM) or Operational Land Imager (OLI)

rainfall erosivity (R) 0.5�

(55 km)Provided by Chinese Academy ofSciences (CAS) project team

Based on 1986e2015, 0.5� resolution daily rainfall of Climate Prediction Center (CPC)(1)

Soil erodibility(K) 30 m Based on soil survey datasets and USLE algorithmTerrain factor (LS) 1 arc (30 m) Calculated based on shuttle radar

topography mission (SRTM)Based on 1 arc-second SRTM, and LS_Tool(2)

Fractal vegetationcoverage (FVC)

30 arc(1000 m)

Provided by CAS project team Based on 2014e2016 MODIS NDVI, and pixel dichotomy model, used to calculate B factor

non-photosyntheticvegetation data(NPV)

3 arcminute(5.5 km)

CSIRO Land and Water(Guerschman et al., 2009)

Based on MODIS images and field observations used to calculate B factor

SRTM 1 arc (30 m) https://lta.cr.usgs.gov/SRTM1Arc NASA Shuttle Radar Topography Mission conducted in 2000 to obtain elevation data for most ofthe world. Used to calculate slope for sampling unit during interpretation.

GLC30 30 m National Geomatics Center ofChina, (Ban et al., 2015)

Based on Landsat TM/OLI, and visual interpretation http://www.globeland30.org/GLC30Download/index.aspxused to make a rough understanding about the background of land use for sampling unit

Note: (1) https://climatedataguide.ucar.edu/climate-data/cpc-unified-gauge-based-analysis-global-daily-precipitation.(2)LS_Tool is a LS calculator developed by Prof Hongmin Zhang and author, it can be applied by send the email ([email protected], [email protected])

Table 3Regions for designing sampling units.

region Interval of latitude (degree) Interval of longitude (degree) between SUs

90 Ne75 N No sampling uints (1)75 Ne60 N 0.5 160 Ne40 N 0.5 0.7540 Ne10 S 0.5 0.5

Note: (1)because of the relatively low level of human activity and very slight soil erosion in the Arctic region, the sampling does not include thisarea.

Table 4Structure of polygon attribute table.

items explanation Type of items

LU_Code Code of landuse textBM_Code Code of bio-measures textEM_Code Code of engineering-measures textresolution Resolution of Google floatDate Date of image datequality Quality of image text

Polygon attribute table (PAT) is a necessary table for coverage model in ArcGISsoftware of Environmental Systems Research Institute, Inc (ESRI), used to record thedescriptive information of areal geo-features.

Table 5Structure of line attribute table.

items explanation Type of items

LU_Code Code of landuse textWIDTH width (m) floatCLASS type text

Note: (1) railway, express highway, highway, field road, sandcheck dam (dam body).(2) line attribute table, or arc attribute table (AAT) is a necessary table for coveragemodel in ArcGIS software of Environmental Systems Research Institute, Inc (ESRI),used to record the descriptive information of line geo-features.


the field data for the cross section in Kazakhstan have not finishedyet, so this article analyzes only the four cross sections shown inFig. 2 (see Fig. 3).

2.4. Calculation of the soil loss rates for sampling units

Based on the interpretation results, the soil loss rates (t km�2

a�1) for the selected sampling units were calculated based on theCSLE (Liu et al., 2002, pp. 143e149) and regional soil erosion factors(R, K, LS) and other data products, such as fractal vegetation

coverage (FVC) and nonphotosynthetic vegetation (NPV), as listedin Table 2, according to the methods of soil erosion assessment inthe WRI (Liu et al., 2013). Then the applicability of the samplinginterpretation was analyzed.

2.5. Accuracy and applicability analysis methods

Based on themodified interpretation results after field work, thekappa coefficients and accuracy values (formula 2 and 3; Lillesandet al., 2015) for each of the four sections were calculated and used to

Fig. 2. Sections and sampling distribution for field works. (a) Southeastern Tibet (China): field work has been conducted in August of 2018, the team consists of 2 scientists andfive postgraduate or doctoral candidates; (b) Northern Thailand: field work has been conducted in November of 2018, the team consists of 6 scientists and 3 postgraduate or doctoralcandidates; (c) Northern Pakistan: April of 2019, the team consists of 3 scientists and 4 postgraduate or doctoral candidates; (d) Eastern Xinjiang (China): June of 2019, the teamconsists of 6 scientists/technicians and 5 postgraduate or doctoral candidates.


make an assessment of interpretation accuracy. Then, by checkingother units that have not been field-verified and making furtheredits for those units if the problems found in the field work exist inthe interpretation results, the accuracy of the 31,921 sampling unitswas improved to some extent.

U¼ xiixiþ

(2)

K ¼NPn

i¼1xii �Pn

i¼1xiþxþi

N2 �Pni¼1xiþxþi

(3)

where r is number of rows in the error matrix, xii is number ofobservations in row i and column i (on the major diagonal), xiþ istotal of observations in row i (shown as marginal total to right ofthe matrix), xþi is total of observations in column i (shown asmarginal total at bottom of the matrix), N is total number of ob-servations included in matrix.

Eleven large sampling sites were selected in the PTP to represent

Fig. 3. Flow diagram of high-resolution image based soil erosion sampling survey. (Note: SRTM– Shuttle Radar Topography Mission; SU e sampling unit; plug-in tool isdeveloped by the authors for load).


the main landscapes of the PTP area. They included black soil un-dulated hilly (northeastern China, BSUH), loess hilly (northwesternChina; LH), red soil hilly (southern China; RSH), northern TP area(China, southeastern NTP), southern TP area (China, southern TP;STP), dry mountain area (northwestern China; DM), humidmountain area (northern Thailand; HM), the Great Lakes region(central Kazakhstan; GL), the Middle East mountain (central Iran;MEM), rolling plain in eastern Europe (central Ukraine; EU), andcold plain (central Russia; CP). Each of the above samples wascovered by 100 sampling units. For each sampling area, the histo-gram similarity index (HS) between the SU and GLC30 data (Banet al., 2015) is calculated for the first category of land use, andused to make an assessment of sampling accuracy. Here, theintersection (HS) of histograms is used to represent the similarity ofthe two datasets, and its value range is 0e1. The larger the HS valueis, the higher the similarity between the two datasets (Swain &Ballard, 1991).

To illustrate the applicability of interpretation results, 42 sam-pling survey units were designatedaround the world and inter-preted with the same method describe in 2.2, and water erosionrate was calculated with CSLE (Liu et al., 2002, pp. 143e149) basedon sampling units and soil erosion factors (RKLS, Table 2), theerosion rate calculation results were compared with the literaturereports (Du et al., 2020).

The technical flow diagram of this study is as follows, in whichquality check of the interpretation mainly includes three aspects:(1) whether the recognizable features (land use and soil conser-vationmeasures) on the images have been interpreted; (2) whetherthe boundaries of the interpreted features are consistent with theactual situation, in images or in the field; (3) whether each polygonin the attribute table has complete attribute items. Therefore,monitors were appointed to be responsible for the inspection. Ifthere is any problem, it must be returned to the interpreter formodification until it meets the requirements.

3. Results

3.1. Results of the interpretation for SUs

The result of the interpretation for each of the units is a digitalmaps in ESRI shapefile format (Fig. 4), therefore in total, 31,921digital maps have been completed after interpretation, by the ef-forts of 10 scientists and more than 40 postgraduate or doctoral

candidates from April of 2018 to Oct of 2019. Each polygon layerrecords six items, including land use, biological measures, engi-neering measures, image quality, image resolution, and image date(Fig. 4). The map scale of the interpretation result is larger than1:5000 in the case of high resolution images; otherwise, the scale isapproximately 1:10,000 (in the case of no high-resolution images).The sampling survey results can meet the needs of the calculationof soil conservation measure factors (biological practice factor, B;engineering practice factor E; tillage practice factor, T) at the levelof each sampling unit and then be used to obtain the soil loss ratedata at the unit level (Liu et al., 2013; Du et al., 2020).

The interpretation results are managed using a ESRI geo-database (Law & Collins, 2018), and the data volume is approxi-mately 359 gigabytes (GB), including 319 GB of basic data (imagesand DEMswithin a range of 5� 5 kmnortheast of the sampling unitpoint) and 50 GB of interpretation unit maps in vector format.

Comparing to the field work based survey, the technique of thisstudy is more efficient and accurate, and t he former strategy moretime/labour consuming. The large volume resultant data, especiallythe soil conservation measures within watershed for each of theSUs, will be a valid base for calculation of soil erosion rate at theregional even global scale.

3.2. Accuracy analysis of interpretation

Based on the modified interpretation maps and problems foundduring the field work, the accuracy of interpretationwas calculatedand analyzed for the four field sections.

3.2.1. Section of Southern TibetThe kappa coefficient and the accuracy values for the interpre-

tation results were calculated based on the interpretation of veryhigh-resolution unmanned aerial vehicle (UAV) images in the Ti-betan area, China. The results (Fig. 4a) show that the mean kappacoefficient of 9 units (No.1, No. 2, No. 3, No. 4, No. 5, No. 6, No. 7, No.8 and No. 9 villages) is 0.7, and the average accuracy is 80.04%,which means that the interpretation results are basically consistentwith the actual situation. The accuracy of No. 2 and No. 4 reached93.76% and 91.84%, respectively, and the kappa coefficients werebetween 0.7 and 0.9, which is fully consistent with that found in thefield. The accuracy in No. 8 is 89.37%, but the kappa value is 0.456, itis because the unit has a confusion with grassland, shrubs and bareland in interpretation.

Fig. 4. Example for result of interpretation of sampling unit. (Note: ‘LU_Code’ is landuse code, ‘BM_Code’ is biological measure code, ‘EM_Code’ is engineering measure code,‘quality’ is quality of images (G is good), ‘resolution’ is image resolution (m), ‘date’ is date of image).


3.2.2. Section of Northern ThailandAccuracy assessment in Thailand is based on the results of field

investigation and modified interpretation. The accuracy calcula-tions of 18 units (Fig. 4b) show that, except for Unit 2, the kappacoefficients of the other small watersheds are all greater than 0.5with an average accuracy of 85.78%, and the interpretation resultshave high consistency with the field observations. The mainproblem for interpretation in this area is that it is difficult toidentify some crops/fruit such as lychee and coffee, when in theseedling stage.

3.2.3. Section of Northern PakistanIn the northern Pakistan Section, field surveys and accuracy

evaluations were performed on 15 survey units. As shown in Fig. 4c,the average accuracy of the units reached 88.3%, except for units 12,13, 15, and 17. The accuracy of the other units is greater than 80%,and kappa is between 0.75 and 0.95. The sampling interpretationresults in Pakistan are consistent with the actual land use types andhave high accuracy. In the sampling units with lower accuracy, themain problem is the difficulty in separating the shrubs and forest. Inaddition, there are many unused construction sites and brick kilnsin Pakistan, which are difficult to distinguish based on the texturefeatures of the images, so land use types are easily classified asfarmland or grassland.

3.2.4. Section of Southeast XinjiangThe accuracy assessment in the Xinjiang section is based on the

land use interpretation results after a field trip. The relief in thisarea is quite high. In total, 42 small watersheds were interpreted,and considering accessibility, only 16 units were finally checked inthe field.

In general, the accuracy in the Xinjiang section is relatively high.The average accuracy of 16 interpretation units is 81.9%, but theaverage kappa coefficient is only 0.44. The variation in the kappacoefficient among units is quite large, with aminimumof 0.05 and amaximum of 0.99 (Fig. 4d). According to the field work, the mainproblem for remote sensing interpretation in Xinjiang in the lab-oratory stage is that the grassland in this area is the dominant landuse type. The texture of the image varies with the season; therefore,misclassifications of bare soil/rock or shrubland are inevitable.

3.3. Representative analysis of the interpretation results

The histogram similarity for the top-level category of land usefor 11 large sampling sites (Figs. 1 and 5) shows that the HS valuesof the two datasets (SU vs GLC30 data) in the larger sampling areasof LH, BSUH, NTP, DM, GL, MEM, EU are more than 0.8, which in-dicates that the results of interpretation in these areas are verysimilar to that of GLC30 data. The HS values for the RSH and HMsampling areas are 0.78 and 0.75, respectively, and the interpre-tation results in these areas are consistent with the land usestructure of the GLC30 data to some extent. However, the HS valuefor the STP is only 0.31, and the interpretation in this area isdifferent from that of the GLC30 data mainly because of thegrassland (refer to 2.2.1).

The absolute values of the differences between the two datasetsand the calculation results (Fig. 6) indicate that the differences inland use types are mainly in grassland, forestland, water body, andbare land. Cropland, shrubland, and construction land have highconsistency. For the LH, NTP, DM, and GL sampling areas, the mainproblems are how to effectively cognitive grassland and bare landand then make a separation of grassland and bare land. The inter-pretation process needs to be based on all available images, someliterature-based understanding, and requisite field checks to makean accurate interpretation for these types.

The differences between the two datasets in the STP area aremainly in the shrub, grassland, water and unused land categories,especially in the water body category, where the difference is 51.7%.The water bodies in GLC30 can include glacial water and snow, butmost of them were found to not truly be permanent snow coverafter analyzing the historical images. They are more likely to begrassland according to the field observations.

For the RSH and HM areas, the difference mainly occurs incropland and forest, partly because there is no orchard category inthe GLC30 classification system, and because the resolution of theimages is coarse, separation of cropland and forest is difficult. Forthe EU and CP areas, the problems are in shrubland, and grassland.In the processes of interpretation, these types are interpreted ac-cording to the image texture and the canopy density, so the resultsare more accurate than that of the GLC30 data.

Overall, the SU data have a high similarity with the GLC30 dataand therefore have a higher representative accuracy in land usestructure in the PTP area. From the perspective of statisticalmathematics, representative accuracy is truly an analysis of sam-pling error or precision.

3.4. Applicability of selected SUs for calculating the soil erosion rate

The calculation results for the water erosion loss rate in theselected units using the CSLE (Fig. 7) show that: (1) the soil erosionrate for each of the polygons (or cells) can be calculated, and a set ofsoil erosion maps for the sampling units (small watershed orrectangular small area) can be generated. The local spatial variationin soil erosion is well represented. It is expected that the macro-scopic differentiation of the soil erosion rates can be represented ordescribed for a large region if the soil erosion rates for all thesampling units have been calculated and simple statistics candetermine the regional rate (e. g., mean). (2) There is an obviousdifference between the potential soil erosion (or soil erosion risk,i.e. RKLSB; Ap in Fig. 7) and the actual soil erosion rate (RKLSBET; Ain Fig. 7)., and this difference is consistent with (Yin et al., 2018).Otherwise, the soil erosion map based on the low- and medium-resolution datasets only represent soil erosion vulnerability orpotential (Yang et al., 2003; Borrelli et al., 2017; Teng, 2017). (3) Theresults shown in Fig. 7 are basically consistent with the results offield surveys (Ma et al., 2018; Zhang et al., 2019); calculations on 42survey units around the world (Fig. 8, Du et al., 2020) show thatthere is good consistency with reports in the literature (Panagoset al., 2015; Teng et al., 2016; USDA, 2018; Du et al., 2020)(seeFig. 9).

4. Discussion

4.1. Interpretation methods and possible improvements

(1) Comparing the PTP sampling with related methods: Asurvey of sampling point (SP) in the NRI has evolved from afield work based to aerial interpretation with necessary fieldsurveys. This evolution has significantly reduced labor andtime consumption (Nusser et al., 1988; USDA-SCS, 1994;Nusser & Goebel, 1997; USDA, 2000; Schnepf & Flanagan,2016; USDA, 2018). It can be concluded that remote sensinginterpretation is an effective way to undertake samplingsurveys, but it cannot completely replace field surveys. TheWRI in China has mainly used a field survey approach, atleast in the first time of WRI, which is a labor and timeconsuming approach (Liu et al., 2013). The soil loss ratecalculated by the NRI is only for random sampling points(each point is actually a hillslope), whereas the WRI and the“PTP sampling” are for a small watershed (flat areas are small


rectangular areas) so are more informative for soil erosionsurveys. Bosco et al. (2015) interpreted 725 small samples(1 km � 1 km) and 85 large samplings (3 km � 3 km) in theEuropean soil erosion survey and obtained the soil erosionrate (Bosco et al., 2015). The results have been used toperform a plausibility check for 1-km-resolution mappingresults based on RUSLE (Bosco et al., 2014). The “PTP sam-pling” has provided a more efficient soil erosion samplingsurvey method by making full use of modern remote sensingand geographic information technology and has also notignored field work. In follow-up research and practice, moreinformation in addition to land use and soil conservationmeasures will be collected based on the integrated ap-proaches of RS interpretation, data sharing, and field work, asin the NRI (Nusser & Goebel, 1997; Schnepf & Flanagan,2016), Such as WEQ factors, land ownership, soil data (soilname, texture, slope class, flooding class, etc), land capability

class, cropping history, irrigation, ephemeral gully erosion,etc, and an attempt will be made to collect information aboutthe impact of human activities on soil erosion (such as con-struction projects), learning from the ‘Project of China Soiland Water Loss Dynamic Monitoring (MWR, 2010.11; Liuet al., 2013), ,2 to form a more complete information collec-tion system.

(2) Improve the automation of the sampling survey: Thesource data of the “PTP sampling” are 0.5-m-resolution im-agery freely downloaded from Google Earth, most of whichare Quickbird images with clear texture structure and canmeet the demand of visual interpretation, especially forextracting soil conservation measures. However, the imagesdownloaded fromGoogle Earth are only digital imageswith asingle channel, with some known and unknown distortionsand no complete metadata. It is not easy for us to undertake

Fig. 5. Kappa coefficient and Accuracy for 4 sections. (a) Kappa coefficient and Accuracy in Southern Tibet section; (b) Kappa coefficient and Accuracy in Northern Thailandsection; (c) Kappa coefficient and Accuracy in Northern Bakinstan section; (d) Kappa coefficient and Accuracy in Southeastern Xinjiang section.

Fig. 6. Histogram similarity of landuse between sampling survey and GLC30 for 11larger sampling areas.

Fig. 7. Histogram similarity of landuse between sampling survey and GLC30 forlanduse types.


image processing and rapid (even automatic) classification.The next development direction should try to look for anduse digital high-resolution remote sensing images (withmultiple channels, systematic distortion correction, and

complete metadata provided by the data sponsor) and try todevelop a semiautomatic classification method to achievethe rapid extraction of land use and soil conservation mea-sures in large areas. At the same time, more attention will bepaid to big data methods (e. g., Ramcharan et al., 2018) in thefuture, such as methods for crowdsourced data collection, tofurther reduce costs and obtain datasets with much highertemporal accuracies (or more current) that are moreinformative.

4.2. Accuracy improvement and uncertainty analysis

(1) Interpretation accuracy and quality control: The accuracyassessment shows that the accuracy of the “PTP sampling” inthe four sections (Tibet, Thailand, Pakistan, and the easternpart of Xinjiang) are 80.04%, 85.78%, 88.3%, and 81.9%respectively. This accuracy is slightly higher than the remotesensing interpretation accuracy of the China Soil and WaterLoss Dynamic Monitoring Project (64%e91%, with an areaweighted average of 76.86%; Li et al., 2018). The interpreta-tion accuracy for orchards and shrubland is relatively lowboth in the China Soil Erosion Monitoring Project and in this

Fig. 8. Map of water erosion loss rate for selected units. (a) Water erosion loss rate in southern Tibet (the sampling unit id is N08132); (b) Water erosion loss rate in NorthernThailand (the sampling unit id is THA02).

Fig. 9. Comparison of calculated soil loss rate and related value from references forselected 42 sampling units.


study. Therefore, further improving the interpretation accu-racy will become the key issue in the future work forapplying “PTP sampling” at a larger geo-extent.

(2) Uncertainty analysis: Data uncertainty is in the nature ofgeo-data (Longley et al., 2015) and is also an inevitableproblem in any data. The NRI and some of the recent globalsoil erosion assessment methods have included compre-hensive uncertainty analyses (Borrelli et al., 2017 USDA,2018). This study has primarily analyzed the accuracy forsampling units, the sampling representativeness for allsampling units, and the applicability of “PTP sampling”.Complete data quality assessment (including data accuracy,precision, consistency, and completeness) and uncertaintyassociated with the geodistribution of sampling will befurther researched for future surveys.

4.3. Implications of this study

With the development of remote sensing, the availability ofhigh-resolution remote sensing data and the application of vol-unteered geographic information (VGI) and geo-big data, GI sciencehas developed to the stage in which we have access to the neededinformation everywhere at any time (Yuan, 2015). Virtual Earthbased on high-resolution remote sensing images (such as Googleearth) and freely accessed high-resolution data have already beencovered worldwide and are widely used in geoscience research (Yu& Gong, 2012). The method of geographic big data is also nowwidely used in the study of earth surface processes (Bui, 2016; Guo,2018; Boulton, 2018). These factors make it possible to extract soilerosion factors and conservation measures at finer resolution (nearmeter scale) and to accurately calculate the soil erosion rate at thesampling level based on high-speed computing. The implication ofthis and previous studies (Zhu et al., 2019; Yang et al., 2019; Duet al., 2020) is there is a possible for us do implement rapid andaccurate sampling survey of soil erosion on a global scale based onthe high-resolution images, even if the visual interpretationmethod to be improved is used.

5. Conclusions

(1) The dominant factors of soil erosion (in this case, land useand soil conservationmeasures) for sampling units can beextracted rapidly and accurately using the PTP samplingapproach. PTP sampling, based on the spatial and temporalvariations in soil erosion and the dominant factors, is sup-ported by the scientific conclusion that we have access toneeded information everywhere at any time and makes fulluse of the virtual earth and its freely accessible high-resolution remote sensing images, and detailed informationof land-use and soil conservation measures can be collectedrapidly and accurately for a larger region. The main technicalprocedure of the sampling survey includes the sampling unitdesigning, the interpretation of land use and soil conserva-tion measures, field checks, accuracy assessment, andapplicability analysis.

(2) The sampling data are accurate with regional represen-tativeness. The accuracy of data collection at the samplingunit scale can reach more than 80.0%, and the land use in-formation obtained in the sampling survey can be used toanalyze the land use characteristics in regional soil erosionmapping. Additionally, this method is mainly based onremote sensing interpretation and emphasizes field work forestablishing interpretation marks and checking the inter-pretation results. The PTP sampling approach is based on and

improves the WRI so that the advantages of the WRI,including higher accuracy, flexibility, and consideration thebenefits of soil conservation (Liu et al., 2013; Li et al., 2018),have been accepted, and efficiency has been improved. Theapproach has been applied to a large area outside China tomeet the requirement of soil erosion sampling at the regionalscale.

(3) The water erosion loss rate for the sampling unit can becalculated accurately based on the results of the samplinginterpretation. Preliminary calculations of the water erosionloss rate for selected units show that the soil loss rate foreach polygon of the unit can be calculated accurately basedon the sampling information (land use and soil conservationmeasures), the existing regional soil erosion factor dataproducts (including R, K, LS, etc.) and the CSLE model. Thecalculation result is a soil erosion map in 2.5 ~ 5-m-resolu-tion raster format, in which spatial variation of soil loss rate(local change) in a sampling unit can be represented. In thisprocess, if the soil conservation measures have not beentaken into account, the calculated soil loss rate will benoticeably overestimated. Therefore, the shortcomings oftraditional remote sensing mapping, especially those thatjust represent the potential of soil erosion, can be overcometo a large extent. Therefore the approach can potentially beextended to a larger area or even a “global soil erosionsampling surveymethod” after some necessarymodification.


The authors confirm that there is no conflict of interest with thenetworks, organizations, and data centres referred to in the paper.

Acknowledgements

This research was supported by the Strategic Priority ResearchProgram of Chinese Academy of Sciences, Grant No. XDA20040202.

References

Amundson, R., Berhe, A. A., Hopmans, J. W., Olson, C., Sztein, A. E., & Sparks, D. L.(2015). Soil and human security in the 21st century. Science, 348(6235),1261071-1261071e1261071-1261076.

Ban, Y., Gong, P., & Gini, C. (2015). Global land cover mapping using Earth obser-vation satellite data: Recent progresses and challenges. ISPRS Journal of Photo-grammetry and Remote Sensing, 103(1), 1e6.

Batjes, N. (1996). Global assessment of land vulnerability to water erosion on a onehalf degree by one half degree grid. Land Degradation & Development, 7(4),353e365.

Borrelli, P., Robinson, D. A., Fleischer, L. R., Lugato, E., Ballabio, C., Alewell, C.,Meusburger, K., Modugno, S., Schütt, B., Ferro, V., Bagarello, V., Oost, K. V.,Montanarella, L., & Panagos, P. (2017). An assessment of the global impact of21st century land use change on soil erosion. Nature Communications, 8(1),1e13.

Bosco, C., de Rigo, D., Dewitte, O., Poesen, J., & Panagos, P. (2014). Visual validation ofthe e-RUSLE model applied at the pan-European scale. In Scientific topics focus 1,MRI-11a13. Notes transdiscipl. Model. Env., maieutike research initiative. https://doi.org/10.6084/m9.figshare.844627

Bosco, C., de Rigo, D., Dewitte, O., Poesen, J., & Panagos, P. (2015). Modelling soilerosion at European scale: Towards harmonization and reproducibility. NaturalHazards and Earth System Sciences, 2(4), 2639e2680.

Boulton, G. (2018). The challenges of a big data earth. Big Earth Data, 2(1), 1e7.Bui, E. N. (2016). Data-driven critical zone science: A new paradigm. The Science of

the Total Environment, 568, 587e593.Chen, T. D., Jiao, J. Y., Wang, H. L., Zhao, C. J., & Lin, H. (2020). Research progress on

soil erosion in Qinhai-Tibet Plateau. Acta Pedologica Sinica, 57(3), 547e564.Crosson, P., Pimentel, D., Harvey, C., et al. (1995). Soil erosion estimates and costs.

Science, 269(5223), 461e465.Du, C. Z., Yang, Q. K., Wang, C. M., & Pang, G. W. (2020). Comparative study on

assessment methods of soil erosion in typical sample areas of the TibetanPlateau (In Chinese). Chinese Journal of Soil Science, 51(1), 50e57.

FAO. (2019a). Soil erosion: The greatest challenge to sustainable soil management.Rome: FAO.

FAO. (2019b). Outcome document of the global Symposium on soil erosion. Rome: FAO.


Farr, T. G., Rosen, P. A., Caro, E., Crippen, R., & Duren, R. (2007). The shuttle radartopography mission. Reviews of Geophysics, 45. RG2004/2007.

Guerschman, J. P., Hill, M. J., Renzullo, L. J., Barrett, D. J., Marks, A. S., & Botha, E. J.(2009). Estimating fractional cover of photosynthetic vegetation, non-photosynthetic vegetation and bare soil in the australian tropical savanna re-gion upscaling the eo-1 hyperion and modis sensors. Remote Sensing of Envi-ronment, 113(5), 928e945.

Guo, H. D. (2018). A project on big earth data science engineering. Bulletin of theChinese Academy of Sciences, 33(8), 818e824 (in Chinese).

Lal, R. (2019). Accelerated Soil erosion as a source of atmospheric CO2. Soil andTillage Research, 188. S0167198718300345.

Law, M., & Collins, A. (2018). Getting to know ArcGIS desktop (5th ed.). Redland, CA:ESRI press.

Lillesand, T. M., Kiefer, R. W., & Chipman, J. W. (2015). Remote sensing and imageinterpretation (7th ed.). John Wiley & Sons, Inc.

Lin, H. L., Zheng, S. T., & Wang, X. L. (2017). Soil erosion assessment based on theRUSLE model in the Three-river headwater area, Qing-hai-Tibetan Plateau,China. Acta Prataculture Sinica, 26(7), 11e22 (in Chinese).

Liu, B. Y., Guo, S. Y., Li, Z. G., et al. (2013). Sampling survey of soil erosion in China.Soil and Water Conservation of China, 10, 26e34 (in Chinese).

Liu, B. Y., Yang, Y., & Lu, S. J. (2018). Discriminations on common soil erosion termsand their implications for soil and water conservation. Science of Soil and WaterConservation, 16(1), 9e15 (in Chinese).

Liu, B. Y., Zhang, K. L., & Xie, Y. (2002). A empirical soil loss equation. Proc of 12th ISCO,Beijing.

Li, A., Wei, X., & Li, Z. G. (2018). Accuracy analysis of land use interpretation of highresolution image in soil erosion dynamic monitoring. Science of Soil and WaterConservation, 16(3), 8e17 (in Chinese).

Li, R., Zhao, M. D., & Yang, Q. K. (2014). Atlas of soil erosion of China. Beijing: MapPress of China (in Chinese).

Longley, P. A., Goodchild, M. F., Maguire, D. J., & Rhind, D. W. (2015). Geographicinformation science and systems. John Wiley & Sons Inc.

Lu, H., & Yu, B. (2002). Spatial and seasonal distribution of rainfall erosivity inAustralia. Australian Journal of Soil Research, 40(6), 887e901.

Ma, B., Zhang, J. Q., Shui, J. F., et al. (2018). Report on field survey of soil erosion incentral and eastern Tibet. Bulletin of Soil and Water Conservation, 38(5), 1e8 (inChinese).

Montanarella, L. (2015). Govern our soils. Nature, 287(7580), 32e33.Morgan, R. P. C. (2005). Soil erosion and conservation (3rd ed.). Malden, MA:

Blackwell Science Ltd.MWR. (2010). 11) Training materials part 6 for first national water conservancy census

of China. Beijing: China Water Conservancy and Hydropower Press (in Chinese).Nusser, S. M., Breidt, F. J., & Fuller, W. A. (1988). Design and estimation for inves-

tigating the dynamics of natural Resources. Ecological Applications, 8(2),234e245.

Nusser, S. M., & Goebel, J. J. (1997). The national Resources inventory: A long-termmulti-resource monitoring programme. Environmental and Ecological Statistics,4(3), 181e204.

Oldeman, L. R., Hakkeling, T. T. A., & W, G. S. (1994). The global extent of soildegradation. In D. Green, & J. Szabolcs (Eds.), Soil resilience and sustainable landuse (pp. 99e118). Wallingford: CAB International: Wallingford.

Oldeman, L. R., Hakkeling, R. T. A., & Sombroek, W. G. (1991).World map of the statusof human-induced soil degradation: An explanatory note (2nd revised edn.).

Panagos, P., Borrelli, P., Poesen, J., Ballabio, C., Lugato, E., Meusburger, K.,Montanarella, L., & Alewell, C. (2015). The new assessment of soil loss by watererosion in Europe. Environmental Science & Policy, 54, 438e447.

Ramcharan, A., Hengl, T., & Nauman, T. (2018). Soil property and class maps of theconterminous US at 100 meter spatial resolution based on a compilation ofnational soil point observations and machine learning. Soil Science Society ofAmerica Journal.

Robinson, D. A., Panagos, P., & Borrelli, P. (2017). Soil natural capital in Europe; aframework for state and change assessment. Scientific Reports, 7(1), 6706.

Schnepf, M., & Flanagan, P. (2016). A history of natural resource inventories conductedby the USDA’s soil conservation service and natural Resources conservation service;A special report by the soil and water conservation society.

Swain, M. J., & Ballard, D. H. (1991). Color indexing. International Journal of ComputerVision, 7(1), 11e32.

Teng, H. F. (2017). Assimilating multi-souce data to model and map potential soilloss in China. In Doctoral degree thesis of zhejiang university. Zhejiang University(in Chinese).

Teng, H., Liang, Z., Chen, S., Liu, Y., Raphael, A., Rossel, V., Chappell, A., Wu, Y., &Shi, Z. (2018). Current and future assessments of soil erosion by water on theTibetan Plateau based on RUSLE and CMIP5 climate models. The Science of theTotal Environment, 673e686.

Teng, H., Rossel, R. A. V., Shi, Z., Behrens, T., Chappell, A., & Bui, E. (2016). Assimi-lating satellite imagery and visible-near infrared spectroscopy to model andmap soil loss by water erosion in Australia. Environmental Modelling & Software,77.

USDA. (2000). Summary report: 1997 national Resources inventory, natural Resourcesconservation service, Washington, DC, and center for survey statistics and meth-odology. Ames, Iowa: Iowa State University.

USDA. (2018). Summary report: 2015 national Resources inventory, natural Resourcesconservation service, Washington, DC, and center for survey statistics and meth-odology. Ames, Iowa: Iowa State University. http://www.nrcs.usda.gov/technical/nri/15summary.

USDA-SCS. (1994). Summary report: 1992 national Resources inventory, natural Re-sources conservation service, Washington, DC, and center for survey statistics andmethodology. Ames, Iowa: Iowa State University.

Weil, R. R., & Brady, N. C. (2017). The nature and properties of soils. New York:Pearson Education Limited.

Xie, Y., Liu, B. Y., Zhang, W. B., et al. (2019). A survey of soil erosion in China of 2011(extended abstract). In FAO. 2019. Proceedings of the global Symposium on soilerosion 2019. Rome. p139 e 143.

Yang, D., Kanae, S., Oki, T., Koike, T., & Musiake, K. (2003). Global potential soilerosion with reference to land use and climate changes. Hydrological Processes,17(14), 2913e2928.

Yang, Q. K., Li, R., & Cao, M. M. (2006). Advances of quantitative assessment onregional soil erosion. Advances in Earth Science, 21(8), 849e856 (In Chinese).

Yang, Q. K., Zhu, M. Y., Wang, C. M., et al. (2019). Interpretation of land use and soilconservation measures in the Pan -third Pole area using Google earth highresolution remotely sensed images (extended abatract)/FAO. In 2019. Pro-ceedings of the global Symposium on soil erosion 2019. Rome. p39 e 42.

Yao, T. D., Chen, F. H., Cui, P., et al. (2017). From Tibetan plateau to third Pole andpan-third Pole. Bulletin of the Chinese Academy of Sciences, 32(9), 924e931 (inChinese).

Yin, S. Q., Zhu, Z. Y., Wang, L., Liu, B. Y., Xie, Y., Wang, G. N., & Li, Y. S. (2018). Regionalsoil erosion assessment based on a sample survey and geostatistics. Hydrologyand Earth System Sciences, 22(3), 1695e1712.

Yuan, M. (2015). Frontiers of GIScience: Evolution, state-of-art, and future path-ways. In P. S. Thenkabail (Ed.), Remote sensing handbook. CRC Press.

Yu, L., & Gong, P. (2012). Google earth as a virtual globe tool for earth science ap-plications at the global scale: Progress and perspectives. International Journal ofRemote Sensing, 33(12), 3966e3986.

Zhang, J. Q., Ma, B., Shui, J. F., et al. (2019). Investigation report on soil erosion innorthern mountain area of Thailand. Bulletin of Soil and Water Conservation,39(1), 1e8 (in Chinese).

Zhu, X. M. (1965). Soil erosion map of China (1:1,500)/atlas of physical geography ofChina. Beijing: Science press (in Chinese).

Zhu, X. M., Chen, D. Z., & Yang, Q. K. (1999). Soil erosion map of China (1:1,500)/atlasof physical geography of China (2nd ed.). Beijing: Science Press (in Chinese).

Zhu, M. Y., Yang, Q. K., Wang, C. M., et al. (2019). Sampling survey method ofregional soil erosion based on remote sensing image. Journal of Soil and WaterConservatin, 33(5), 64e71 (in Chinese).


INTRODUCTION

Types of paper Contributions falling into the following categories will be considered for publication: Original research papers, reviews, letters to the editor.

Please ensure that you select the appropriate article type from the list of options when making your submission. Authors contributing to special issues should ensure that they select the special issue article type from this list.

BEFORE YOU BEGIN

Ethics in Publishing For information on Ethics in Publishing and Ethical guidelines for journal publication see http://www.elsevier.com/publishingethics and http://www.elsevier.com/ethicalguidelines.

Conf ict of interest All authors are requested to disclose any actual or potential conf ict of interest including any f nancial, personal or other relationships with other people or organizations within three years of beginning the submitted work that could inappropriately inf uence, or be perceived to inf uence, their work. See also http://www.elsevier.com/conf ictsof nterest.

Submission declaration Submission of an article implies that the work described has not been published previously (except in the form of an abstract or as part of a published lecture or academic thesis), that it is not under consideration for publication elsewhere, that its publication is approved by all authors and tacitly or explicitly by the responsible authorities where the work was carried out, and that, if accepted, it will not be published elsewhere including electronically in the same form, in English or in any other language, without the written consent of the copyright-holder.

Changes to authorship This policy concerns the addition, deletion, or rearrangement of author names in the authorship of accepted manuscripts:

Before the accepted manuscript is published in an online issue: Requests to add or remove an author, or to rearrange the author names, must be sent to the Journal Manager from the corresponding author of the accepted manuscript and must include: (a) the reason the name should be added or removed, or the author names rearranged and (b) written conf rmation (e-mail, fax, letter) from all authors that they agree with the addition, removal or rearrangement. In the case of addition or removal of authors, this includes conf rmation from the author being added or removed. Requests that are not sent by the corresponding author will be forwarded by the Journal Manager to the corresponding author, who must follow the procedure as described above. Note that: (1) Journal Managers will inform the Journal Editors of any such requests and (2) publication of the accepted manuscript in an online issue is suspended until authorship has been agreed.

After the accepted manuscript is published in an online issue: Any requests to add, delete, or rearrange author names in an article published in an online issue will follow the same policies as noted above and result in a corrigendum.

Copyright [example: Upon acceptance of an article, authors will be asked to complete a ‘Journal Publishing Agreement’. Acceptance of the agreement will ensure the widest possible dissemination of information. An e-mail will be sent to the corresponding author conf rming receipt of the manuscript together with a ‘Journal Publishing Agreement’ form or a link to the online version of this agreement.

Permission of the International Research and Training Center on Erosion and Sedimentation or the China Water and Power Press is required for resale or distribution outside the institution and for all other derivative works, including compilations and translations (please consult [email protected] or [email protected]). If excerpts from other copyrighted works are included, the author(s) must obtain written permission from the copyright owners and credit the source(s) in the article.]

Role of the funding source You are requested to identify who provided f nancial support for the conduct of the research and/or preparation of the article and to brief y describe the role of the sponsor(s), if any, in study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication. If the funding source(s) had no such involvement then this should be stated. Please see http://www.elsevier.com/funding.

Language and language services Please write your text in good English (American or British usage is accepted, but not a mixture of these). Authors who require information about language editing and copyediting services pre- and post-submission please visit http://webshop.elsevier.com/languageediting or our customer support site at http://support.elsevier.com for more information.

Submission Submission to this journal proceeds totally online. Use the following guidelines to prepare your article. Via the homepage of this journal (http://www.evise.com/evise/faces/pages/navigation/NavController.jspx?JRNL_ACR=ISWCR) you will be guided stepwise through the creation and uploading of the various f les. The system automatically converts source f les to a single Adobe Acrobat PDF version of the article, which is used in the peer-review process. Please note that even though manuscript source f les are converted to PDF at submission for the review process, these source f les are needed for further processing after acceptance. All correspondence, including notif cation of the Editor’s decision and requests for revision, takes place by e-mail and via the author’s homepage, removing the need for a hard-copy paper trail. If you are unable to provide an electronic version, please contact the editorial off ce prior to submission [example: e-mail: [email protected] or [email protected]; telephone: +86 10 68786416; or fax: 86 10 68411174].

Additional Information Tables and f gures may be presented with captions within the main body of the manuscript; if so, f gures should additionally be uploaded as high resolution f les.

PREPARATION

Use of wordprocessing software It is important that the f le be saved in the native format of the wordprocessor used. The text should be in single-column format. Keep the layout of the text as simple as possible. Most formatting codes will be removed and replaced on processing the article. In particular, do not use the wordprocessor’s options to justify text or to hyphenate words. However, do use bold face, italics, subscripts, superscripts etc. When preparing tables, if you are using a table grid, use only one grid for each individual table and not a grid for each row. If no grid is used, use tabs, not spaces, to align columns. The electronic text should be prepared in a way very similar to that of conventional manuscripts (see also the Guide to Publishing with Elsevier: http://www.elsevier.com/guidepublication). Note that source f les of f gures, tables and text graphics will be required whether or not you embed your f gures in the text. See also the section on Electronic illustrations.

GUIDE FOR AUTHORS

To avoid unnecessary errors you are strongly advised to use the “spell-check” and “grammar-check” functions of your wordprocessor.

LaTeX If the LaTeX f le is suitable, proofs will be produced without rekeying the text. The article should preferably be written using Elsevier’s document class “elsarticle”, or alternatively any of the other recognized classes and formats supported in Elsevier’s electronic submissions system, for further information see http://www.elsevier.com/wps/f nd/authorsview.authors/latex-ees-supported.

The Elsevier “elsarticle” LaTeX style f le package (including detailed instructions for LaTeX preparation) can be obtained from the Quickguide: http://www.elsevier.com/latex. It consists of the f le: elsarticle.cls, complete user documentation for the class f le, bibliographic style f les in various styles, and template f les for a quick start.

Article structure Subdivision - numbered sections Divide your article into clearly def ned and numbered sections. Subsections should be numbered 1.1 (then 1.1.1, 1.1.2, ...), 1.2, etc. (the abstract is not included in section numbering). Use this numbering also for internal cross-referencing: do not just refer to “the text”. Any subsection may be given a brief heading. Each heading should appear on its own separate line.

Introduction State the objectives of the work and provide an adequate background, avoiding a detailed literature survey or a summary of the results.

Material and methods Provide suff cient detail to allow the work to be reproduced. Methods already published should be indicated by a reference: only relevant modif cations should be described.

Theory/calculation A Theory section should extend, not repeat, the background to the article already dealt with in the Introduction and lay the foundation for further work. In contrast, a Calculation section represents a practical development from a theoretical basis.

Results Results should be clear and concise.

Discussion This should explore the signif cance of the results of the work, not repeat them. A combined Results and Discussion section is often appropriate. Avoid extensive citations and discussion of published literature.

Conclusions The main conclusions of the study may be presented in a short Conclusions section, which may stand alone or form a subsection of a Discussion or Results and Discussion section.

Appendices If there is more than one appendix, they should be identif ed as A, B, etc. Formulae and equations in appendices should be given separate numbering: Eq. (A.1), Eq. (A.2), etc.; in a subsequent appendix, Eq. (B.1) and so on. Similarly for tables and f gures: Table A.1; Fig. A.1, etc.

Essential title page information • Title. Concise and informative. Titles are often used in information-

retrieval systems. Avoid abbreviations and formulae where possible.• Author names and aff liations. Where the family name may

be ambiguous (e.g., a double name), please indicate this clearly. Present the authors’ aff liation addresses (where the actual work was done) below the names. Indicate all aff liations with a lower-case superscript letter immediately after the author’s name

and in front of the appropriate address. Provide the full postal address of each aff liation, including the country name, and, if available, the e-mail address of each author.

• Corresponding author. Clearly indicate who will handle correspondence at all stages of refereeing and publication, also post-publication. Ensure that telephone and fax numbers (with country and area code) are provided in addition to the e-mail address and the complete postal address. Contact details must be kept up to date by the corresponding author.

• Present/permanent address. If an author has moved since the work described in the article was done, or was visiting at the time, a “Present address” (or “Permanent address”) may be indicated as a footnote to that author’s name. The address at which the author actually did the work must be retained as the main, aff liation address. Superscript Arabic numerals are used for such footnotes.

Abstract A concise and factual abstract is required. The abstract should state brief y the purpose of the research, the principal results and major conclusions. An abstract is often presented separately from the article, so it must be able to stand alone. For this reason, References should be avoided, but if essential, then cite the author(s) and year(s). Also, non-standard or uncommon abbreviations should be avoided, but if essential they must be def ned at their f rst mention in the abstract itself.

Keywords Authors are invited to submit keywords associated with their paper.

Abbreviations Def ne abbreviations that are not standard in this f eld in a footnote to be placed on the f rst page of the article. Such abbreviations that are unavoidable in the abstract must be def ned at their f rst mention there, as well as in the footnote. Ensure consistency of abbreviations throughout the article.

Acknowledgements Collate acknowledgements in a separate section at the end of the article before the references and do not, therefore, include them on the title page, as a footnote to the title or otherwise. List here those individuals who provided help during the research (e.g., providing language help, writing assistance or proof reading the article, etc.).

Nomenclature and units Follow internationally accepted rules and conventions: use the international system of units (SI). If other quantities are mentioned, give their equivalent in SI. Authors wishing to present a table of nomenclature should do so on the second page of their manuscript.

Math formulae Present simple formulae in the line of normal text where possible and use the solidus (/) instead of a horizontal line for small fractional terms, e.g., X/Y. In principle, variables are to be presented in italics. Powers of e are often more conveniently denoted by exp. Number consecutively any equations that have to be displayed separately from the text (if referred to explicitly in the text).

Footnotes Footnotes should be used sparingly. Number them consecutively throughout the article, using superscript Arabic numbers. Many wordprocessors build footnotes into the text, and this feature may be used. Should this not be the case, indicate the position of footnotes in the text and present the footnotes themselves separately at the end of the article. Do not include footnotes in the Reference list.

Table footnotes Indicate each footnote in a table with a superscript lowercase letter.

Artwork Electronic artwork General points • Make sure you use uniform lettering and sizing of your original

artwork. • Save text in illustrations as “graphics” or enclose the font. • Only use the following fonts in your illustrations: Arial, Courier,

Times, Symbol. • Number the illustrations according to their sequence in the text. • Use a logical naming convention for your artwork f les. • Provide captions to illustrations separately. • Produce images near to the desired size of the printed version. • Submit each f gure as a separate f le.

A detailed guide on electronic artwork is available on our website: http://www.elsevier.com/artworkinstructions

You are urged to visit this site; some excerpts from the detailed information are given here.

Formats Regardless of the application used, when your electronic artwork is f nalised, please “save as” or convert the images to one of the following formats (note the resolution requirements for line drawings, halftones, and line/halftone combinations given below):

EPS: Vector drawings. Embed the font or save the text as “graphics”. TIFF: color or grayscale photographs (halftones): always use a minimum of 300 dpi. TIFF: Bitmapped line drawings: use a minimum of 1000 dpi. TIFF: Combinations bitmapped line/half-tone (color or grayscale): a minimum of 500 dpi is required.

If your electronic artwork is created in a Microsoft Off ce application (Word, PowerPoint, Excel) then please supply “as is”.

Please do not: • Supply f les that are optimised for screen use (like GIF, BMP, PICT,

WPG); the resolution is too low; • Supply f les that are too low in resolution; • Submit graphics that are disproportionately large for the content.

Color artwork Please make sure that artwork f les are in an acceptable format (TIFF, EPS or MS Off ce f les) and with the correct resolution. If, together with your accepted article, you submit usable color f gures then Elsevier will ensure, at no additional charge, that these f gures will appear in color on the Web (e.g., ScienceDirect and other sites) regardless of whether or not these illustrations are reproduced in color in the printed version.

Figure captions Ensure that each illustration has a caption. Supply captions separately, not attached to the f gure. A caption should comprise a brief title (not on the f gure itself) and a description of the illustration. Keep text in the illustrations themselves to a minimum but explain all symbols and abbreviations used.

Tables Number tables consecutively in accordance with their appearance in the text. Place footnotes to tables below the table body and indicate them with superscript lowercase letters. Avoid vertical rules. Be sparing in the use of tables and ensure that the data presented in tables do not duplicate results described elsewhere in the article.

References Citation in text Please ensure that every reference cited in the text is also present in the reference list (and vice versa). Any references cited in the abstract must be given in full. Unpublished results and personal

communications are not recommended in the reference list, but may be mentioned in the text. If these references are included in the reference list they should follow the standard reference style of the journal and should include a substitution of the publication date with either “Unpublished results” or “Personal communication” Citation of a reference as “in press” implies that the item has been accepted for publication.

Web references As a minimum, the full URL should be given and the date when the reference was last accessed. Any further information, if known (DOI, author names, dates, reference to a source publication, etc.), should also be given. Web references can be listed separately (e.g., after the reference list) under a different heading if desired, or can be included in the reference list.

References in a special issue Please ensure that the words ‘this issue’ are added to any references in the list (and any citations in the text) to other articles in the same Special Issue.

Reference management software This journal has standard templates available in key reference management packages EndNote (http://www.endnote.com/support/enstyles.asp) and Reference Manager (http://refman.com/support/rmstyles.asp). Using plug-ins to wordprocessing packages, authors only need to select the appropriate journal template when preparing their article and the list of references and citations to these will be formatted according to the journal style which is described below.

Reference style The International Soil and Water Conservation Research (ISWCR) follows the Publication Manual of the American Psychological Association (APA) 6th for reference lists and text citation. The author are referred to the Publication Manual of the American Psychological Association, Fifth Edition, ISBN 1-55798-790-4, copies of which may be ordered from http://www.apa.org/books/4200061.html or APA Order Dept., P.O.B. 2710, Hyattsville, MD 20784, USA. or APA, 3 Henrietta Street, London, WC3E 8LU, UK. Details concerning this referencing style can also be found at http://humanities.byu.edu/linguistics/Henrichsen/APA/APA01.html.

Examples: Text: All citations in the text should refer to:

One author –Smith (2002) found…–(Smith, 2002).

Two Authors:–Smith and Jones (2003) found…–(Smith & Jones, 2003).

Three or More Authors–Smith et al. (2001) found…- (Phelps et al., 2004)–Smith et al. (2002) found…

Groups as Authors:–1st Citation: (American Psychological Association [APA], 2000).–Subsequent Citations:(APA, 2000).

Anonymous or No Author–Use f rst few words of reference list entry (usually title):(—Study Finds, 1995)(TEA, 2007)

Authors with Same Surname–Include initials: S. T. Smith (2000) and J. D. Smith (1999)

Two of more works within the same parentheses–In order alphabetically, as they would appear in references, separated by semi-colons (Jones, 2003; Thomas, 2010)–If by same author, then by date(Jones, 2003, 2007)

References should be arranged f rst alphabetically and then further sorted chronologically if necessary. More than one reference from the same author(s) in the same year must be identif ed by the letters “a”, “b”, “c”, etc., placed after the year of publication

Reference to a journal publication Carlson, L. A. (2003). Existential theory: Helping school counselors attend to youth at risk for violence. Professional School Counseling, 6(5), 10-15.

Sagarin, B. J., & Lawler-Sagarin, K. A. (2005). Critically evaluating competing theories: An exercise based on the Kitty Genovese murder. Teaching of Psychology, 32(3), 167–169.

Hughes, J. C., Brestan, E. V., & Valle, L. A. (2004). Problem-solving interactions between mothers and children. Child and Family Behavior Therapy, 26(1), 1-16.

Journal with more than seven authorsGilbert, D. G., McCleron, J. F., Rabinovich, N. E., Sugai, C., Plath, L. C., Asgaard, G., Botros, N. (2004). Effects of quitting smoking on EEG activation and attendtionlast for more than 31 days and are more severe with stress. Nicotine and Tobacco Research, 6, 249-267.

Herbst-Damm, K.L., & Kulik, J.A. (2005). Volunteer support, marital status, and the survival times of terminally ill patients. Health Psychology, 24, 225-229. doi: 10.1037/0278-6133.24.2.225

Silick, T.J., & Schutte, N.S. (2006). Emotional intelligence and self-esteem mediate between perceived early parental love and adult happiness. E-Journal of Applied Psychology, 2(2), 38-48. Retrieved from http://ojs.lib.swin.edu.au/index.php/ejap.

Reference to a bookBeck, C. A. J., & Sales, B. D. (2001). Family mediation: Facts, myths, and future prospects. Washington, DC: American Psychological Association.

Johnson, R. A. (1989). Retrieval inhibition as an adaptive mechanism in human memory. In H. L. RoedigerIII & F. I. M. Craik(Eds.), Varieties of memory & consciousness (pp. 309-330). Hillsdale, NJ: Erlbaum.

English translation of a book:Lang, P. S. (1951). A philosophical essay on probabilities (F. W. Truscott & F. L. Emory, Trans.). New York, NY: Dover. (Original work published 1814)

*In text, cite original date and translation date: (Lang, 1814/1951).

Dissertations and ThesesCaprette, C. L. (2005). Conquering the cold shudder: The origin and evolution of snake eyes (Doctoral dissertation). Ohio State University, Columbus, OH.

Pecore, J. T. (2004). Sounding the spirit of Cambodia: The living tradition of Khmer music and dance-drama in a Washington, DC community (Doctoral dissertation). Retrieved from Dissertations and Theses database. (UMI No. 3114720)

Caprette, C. L. (2005). Conquering the cold shudder: The origin and evolution of snake eyes (Doctoral dissertation). Retrieved from http://www.ohiolink.edu/etd/send-pdf.cgi?acc_num=osu1111184984

Online resource from group/governmentU.S. Department of Health and Human Services. (2003). Managing asthma: A guide for schools. Retrieved from http://www.nhibi.nih.gov/health/prof/lung/asthma/asth_sch.pdf

Reference in other Language Hughes, J. C., Brestan, E. V., & Valle, L. A. (2004). Problem-solving interactions between mothers and children. Child and Family Behavior Therapy, 26(1), 1-16. (In Chinese)

Journal abbreviations source Journal names should be abbreviated according to Index Medicus journal abbreviations: http://www.nlm.nih.gov/tsd/serials/lji.html; List of title word abbreviations: http://www.issn.org/2-22661-LTWA-online.php; CAS (Chemical Abstracts Service): http://www.cas.org/sent.html.

Submission checklist The following list will be useful during the f nal checking of an article prior to sending it to the journal for review. Please consult this Guide for Authors for further details of any item.

Ensure that the following items are present: One Author designated as corresponding Author: • E-mail address • Full postal address • Telephone and fax numbers

All necessary f les have been uploaded • Keywords • All f gure captions • All tables (including title, description, footnotes)

Further considerations • Manuscript has been “spellchecked” and “grammar-checked” • References are in the correct format for this journal • All references mentioned in the Reference list are cited in the text,

and vice versa • Permission has been obtained for use of copyrighted material from

other sources (including the Web) • Color f gures are clearly marked as being intended for color

reproduction on the Web (free of charge) and in print or to be reproduced in color on the Web (free of charge) and in black-and-white in print

• If only color on the Web is required, black and white versions of the f gures are also supplied for printing purposes

For any further information please visit our customer support site at http://support.elsevier.com.

AFTER ACCEPTANCE

Use of the Digital Object Identif er The Digital Object Identif er (DOI) may be used to cite and link to electronic documents. The DOI consists of a unique alpha-numeric character string which is assigned to a document by the publisher upon the initial electronic publication. The assigned DOI never changes. Therefore, it is an ideal medium for citing a document, particularly ‘Articles in press’ because they have not yet received their full bibliographic information. The correct format for citing a DOI is shown as follows (example taken from a document in the journal Physics Letters B): doi:10.1016/j.physletb.2010.09.059 When you use the DOI to create URL hyperlinks to documents on the web, they are guaranteed never to change.

Proofs One set of page proofs (as PDF f les) will be sent by e-mail to the corresponding author (if we do not have an e-mail address then paper

proofs will be sent by post) or, a link will be provided in the e-mail so that authors can download the f les themselves. Elsevier now provides authors with PDF proofs which can be annotated; for this you will need to download Adobe Reader version 7 (or higher) available free from http://get.adobe.com/reader. Instructions on how to annotate PDF f les will accompany the proofs (also given online). The exact system requirements are given at the Adobe site: http://www.adobe.com/products/reader/systemreqs.

If you do not wish to use the PDF annotations function, you may list the corrections (including replies to the Query Form) and return them to Elsevier in an e-mail. Please list your corrections quoting line number. If, for any reason, this is not possible, then mark the corrections and any other comments (including replies to the Query Form) on a printout of your proof and return by fax, or scan the pages and e-mail, or by post. Please use this proof only for checking the typesetting, editing, completeness and correctness of the text, tables and f gures. Signif cant changes to the article as accepted for publication will only be considered at this stage with permission from the Editor. We will do everything possible to get your article published quickly and accurately – please let us have all your corrections within 72 hours. It is important to ensure that all corrections are sent back to us in one communication: please check carefully before replying, as inclusion of any subsequent corrections cannot be guaranteed. Proofreading is solely your responsibility. Note that Elsevier may proceed with the publication of your article if no response is received.

Offprints The corresponding author, at no cost, will be provided with a PDF f le of the article via e-mail. For an extra charge, paper offprints can be ordered via the offprint order form which is sent once the article is accepted for publication. The PDF f le is a watermarked version of the published article and includes a cover sheet with the journal cover image and a disclaimer outlining the terms and conditions of use.

More information about article offprint is available here: http://webshop.elsevier.com/

AUTHOR INQUIRIES

For inquiries relating to the submission of articles (including electronic submission) please visit this journal’s homepage. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. You can track accepted articles at http://www.elsevier.com/trackarticle. You can also check our Author FAQs (http://www.elsevier.com/authorFAQ) and/or contact Customer Support via http://support.elsevier.com.

Editorial BoardInternational Soil and Water Conservation Research (ISWCR)

AdvisorBlum, Winfried University of Natural Resources and Life Sciences, Austria Critchley, WilliamCIS-Centre for International Cooperation, NetherlandsCui, Peng Institute of Mountain Hazards and Environment, Chinese Academy of SciencesDelgado, Jorge AUSDA-ARS-SMSBRU, Soil Management and Sugar Beet Research Unit, USADumanski, JulianWorld Bank; Gov’t of Canada, CanadaEl-Swaify, Samir University of Hawai‘I, USA

Lal, Rattan Ohio State University, USALi, RuiInstitute of Soil and Water Conservation,Chinese Academy of Sciences, ChinaLiu, BaoyuanBeijing Normal University, ChinaLiu, GuobinInstitute of Soil and Water Conservation,Chinese Academy of Sciences, ChinaNi, Jinren Peking Univeristy, ChinaShao, MinganInstitute of Geographic Science and Natural Resources Research, Chinese Academy of Sciences (CAS), China

Shen, Guofang Beijing Forestry University, ChinaSombatpanit, Samran Department of Land Development Bangkhen, ThailandSun, Honglie Institute of Geographic Science and Natural Resources Research, Chinese Academy of Sciences, ChinaTang, Keli Institute of Soil and Water Conservation, Chinese Academy of Sciences, ChinaYao, WenyiYellow River Institute of Hydraulic Research, China

Editor-In-ChiefNearing, Mark USDA-ARS Southwest Watershed Research Center, [email protected]

Lei, TingwuChina Agricultural University, [email protected]

Associate EditorsAl-Hamdan, OsamaTexas A&M University, Kingsville, USABaffaut, ClaireUSDA-ARS Columbia, MO, USABorrelli, PasqualeUniversity of Basel, SwitzerlandBruggeman, AdrianaEnergy, Environment and Water Research Center of the Cyprus Institute, CyprusChen, HongsongInstitute of Subtropical Agriculture, Chinese Academy of Sciences, China Dazzi, Carmelo University of Palermo, ItalyFang, HongweiTsinghua Univeristy, ChinaFullen, Mike University of Wolverhampton, UKGolabi, Mohammad University of Guam, USAGolosov, Valentin Kazan Federal University, RussiaGomez, Jose AlfonsoInstitute for Sustainable Agriculture, CSIC, Cordob, Spain

Gomez Macpherson, Helena Institute for Sustainable Agriculture (IAS-CSIC), SpainGuzmán, GemaUniversidad of Cordoba, SpainHuang, Chi-hua USDA-ARS National Soil Erosion Research Laboratory, USAKrecek, Josef Czech Technical University, CZ Licciardello, FelicianaUniversity of Catania, ItalyLorite Torres, IgnacioCentro Alameda del Obispo, IFAPA, Junta de Andalucía, SpainMerten, Gustavo University of Minnesota Duluth, USANapier, Ted Ohio State University, USANouwakpo, SayjroUSDA -ARS Northwest Irrigation and Soils Research Laboratory, USANunes, JoãoUniveristy of Lisbon, Portugal

Obando-Moncayo, FrancoGeographic Institute Agustin Codazzi, Soil Survey Department, ColombiaOliveira, Paulo Tarso S.Federal University of Mato Grosso do Sul, BrazilOsorio, JavierTexas A&M University, USAPla Sentís, IldefonsoUniversitat de Lleida, SpainPolyakov, ViktorUSDA-ARS Southwest Watershed Research Center, USAReichert, Jose MiguelFederal University of Santa Maria, Brazil Sadeghi, Seyed HamidrezaTarbiat Modares University, IranSun, Weiling Peking University, ChinaWagner, Larry USDA-ARS, Rangeland Resources & Systems Research Unit, USAWalling, Des E. University of Exeter, UK

Wei, HaiyanSouthwest Watershed Research, USDA-ARS, USA Williams, JasonSouthwest Watershed Research, USDA-ARS, USAYin, ShuiqingBeijing Normal University, China Yu, Xinxiao Beijing Forestry University, China Zhang, GuanghuiBeijing Normal University, China Zhang, QingwenCAAS Institute of Agro-Environment and Sustainable Development, China Zhang, X.C. JohnIUSDA-ARS Grazinglands Research Laboratory, USAZhang, YongguangInternational Institute for Earth System Science, Nanjing University, ChinaZheng, Fenli Institute of Soil and Water Conservation, Chinese Academy of Sciences, ChinaZlatic, MoidragUniversity of Belgrade, Serbia

Editorial Board MembersCai, Qiangguo, ChinaChen, Yongqin David, China (Hong Kong)Dazzi, Carmelo, ItalyDu, Pengfei, ChinaDuan, Xingwu, ChinaFullhart, Andrew, USAHe, Binghui, ChinaHorn, Rainer, GermanyKapur, Selim, TurkeyKunta, Karika, ThailandLi, Li, USALi, Yingkui, USALi, Zhanbin, China

Liang, Yin, ChinaLiu, Benli, ChinaLiu, Xiaoying, ChinaLo, Kwong Fai Andrew, China (Taiwan)McGehee, Ryan, USAMiao, Chiyuan, ChinaMrabet, Rachid, FranceMu, Xingmin, ChinaOliveira, Joanito, BrazilOuyang, Wei, China Paz-Alberto, Annie Melinda, PhilippinesPeiretti, Roberto, ArgentinaRitsema, Coen, Netherlands

Shi, Zhihua, ChinaSukvibool, Chinapat, ThailandSurendran, Udayar, IndiaTahir Anwar, Muhammad, PakistanTang, Qiuhong, China Tarolli, Paolo, ItalyTorri, Dino, ItalyWang, Bin, ChinaWu, Gaolin, ChinaYu, Yang, ChinaZhang, Fan, ChinaZhang, Kebin, China

Editorial Managing CommitteeLiu, Ning Kuang, Shangfu Ying, Youfeng Liu, Jigen Fan, HaomingNing, Duihu Li, Zhongfeng Liu, Guangquan Duan, Xingwu

Managing EditorsChyu, Paige (Executive Editor)International Research and Training Center on Erosion and Sedimentation, China [email protected] Li, KangChina Water Power Press, [email protected]

Zhao, Ying International Research and Training Center on Erosion and Sedimentation, China [email protected]

international soil and water conservation …

Documents