an integrated approach of gis, rusle and ahp to model soil

An integrated approach of GIS, RUSLE and AHPto model soil erosion in West Kameng watershed,Arunachal Pradesh

BISWAJIT DAS1, REETASHREE BORDOLOI

1, LOBSANG TASHI THUNGON1,

ASHISH PAUL1, PANKAJ K PANDEY

2, MADHUSUDHAN MISHRA3 and

OM PRAKASH TRIPATHI1,*

1Department of Forestry, North Eastern Regional Institute of Science and Technology (Deemed-to-be University),Nirjuli, Arunachal Pradesh 791 109, India.2Department of Agriculture Engineering, North Eastern Regional Institute of Science and Technology(Deemed-to-be University), Nirjuli, Arunachal Pradesh 791 109, India.3Department of Electronics and Communication Engineering, North Eastern Regional Institute of Scienceand Technology (Deemed-to-be University), Nirjuli, Arunachal Pradesh 791 109, India.*Corresponding author. e-mail: [email protected]

MS received 10 April 2019; revised 7 December 2019; accepted 24 December 2019

Soil erosion has always been a major environmental problem in many parts of the world including thenortheastern region of India. An increase in the rate of soil erosion has tremendous implications on land degra-dation, biodiversity loss, productivity, etc. Hence, assessment of soil erosion hazard and its spatial distribution isessential to serveasabaselinedata for eAectivecontrolmeasures.Thepresent studyuses reviseduniversal soil lossequation (RUSLE) and analytical hierarchy process (AHP) approach integrated with geospatial technology formodeling soil erosion hazard zone of West Kameng watershed of Arunachal Pradesh, Northeast India. Theassessment showed that the erodibility factor of soil ranged between 0 and 0.38 t/ha/MJ/mm and slope lengthand steepness factor increaseswith increase in slope angle. Lower normalized difference vegetation index (NDVI)values depict vegetation cover and higher values represent the rocky area or barren land. Spatial distribution ofconservation support practice on soil loss indicated the variability (0–1) where lower value represents the higherconservationpractice.Thepredictedaverage soil erosionratewas124.21t/ha/Yr.Normalizedeigenvectorvaluesrangedbetween0.03 and0.20.The areaswithmore slope, relative relief, drainage density, lineament density, andfrequencyhave shown comparatively higher eigenvector values, and it has beennoticed that the strength of theseeigenvectors reduces with a decrease in the values of the parameters. The spatial soil erosion potential map wasdelineated using eight geo-environmental variables (LULC, geomorphology, slope, relative relief, drainagedensity, drainage frequency, lineament density, and lineament frequency). The soil hazardmap showed that themoderate soil erosionhas themaximum(57.71%)areacover followedbyhigherosionclass (26.09%)whichdepictsthat most of the watershed areas are moderate to high vulnerable to soil erosion. The eDciency of the AHP wasvalidated applying area under curve (AUC)methodwhich result 84.90%accuracy in the present study.Based onthe Bndings, it is being recommended that present watershed requires adequate control procedures on a prioritybasis to conserve soil resources and reduce Cood events and siltation of water bodies.

Keywords. AHP; AUC; geospatial; RUSLE; soil erosion.

A Tribute to our Beloved Teacher Late Prof. R S Tripathi.

J. Earth Syst. Sci. (2020) 129:94 � Indian Academy of Scienceshttps://doi.org/10.1007/s12040-020-1356-6 (0123456789().,-volV)(0123456789().,-volV)

http://crossmark.crossref.org/dialog/?doi=10.1007/s12040-020-1356-6&domain=pdf

https://doi.org/10.1007/s12040-020-1356-6

1. Introduction

The soil is one of the vital resources playing asignificant role in sustainable agriculture produc-tion and improved economy of the nation.Increased world population has created anenhanced food supply demand to meet the basicneeds, thereby negatively impacting soil healthdirectly or indirectly. Extensive agriculture andtree felling from the forest for future agriculturalpractices, developmental projects and mining wereseverely aAecting the upper soil layer and simul-taneously soil erosion in the particular area. Theseactivities had drastically altered soil propertieswhich causes a negative impact on the soil pro-ductivity. The significant impacts of soil erosioninclude poor soil productivity, degraded waterquality due to eutrophication in water bodies, sil-tation, and sedimentation of lakes and river beds,enhances Cood risk, etc. (Onyando et al. 2005; Zhouet al. 2008). In recent decades, it is considered assignificant threats for the environment whichseverely impacting the agriculture, naturalresources and the environment globally (Rahmanet al. 2009). It had been reported that the propor-tion percentage of soil erosion to total land degra-ded are 25% for Europe, 18% for Asia, 16% inAfrica, and least 5% is in North America, respec-tively (Oldeman et al. 1992). Hence, it has becomea matter of grave concern for minimizing theimpact of soil erosion through sustainable man-agement practices by considering the magnitude aswell as its spatial variability.The major problem in soil erosion study is the

implication of Beld-based method in determiningthe soil erosion impact that requires a lot ofresources and it should be done in the controlledsystem (Kirkby and Morgan 1980). So researchersaround the world face difBculties in assessment andprediction of soil erosion. Hence, since 1930, theydeveloped various models (Lal 2001). The variousmodels developed so far can be classiBed as aphysical process based, empirical and semi-empir-ical model based. The models used for quantifyingthe soil erosion were USLE – universal soil lossequation (Wischmeier and Smith 1978), RUSLE –

revised universal soil loss equation (Renard et al.1997), Hill slope model (Gronsten and Lundekvam2006), European soil erosion model (EUROSEM)(Morgan et al. 1992), etc. Among the above-mentioned models, USLE and RUSLE are mostcommonly employed (Sinha and Joshi 2012; Joshi2018). These models are more convenient than

others as they rely on parameters like precipita-tion, physiography, soil and land-use practice andconservation sustenance. For sustainable manage-ment of soil eroded area, it is pre-requisite for thedecision maker to locate soil erosion prone area inspite of knowing total soil quantity lost from thearea (Eltner et al. 2013; Cerd�a et al. 2017).Several studies were conducted around the world

on these aspects using USLE and RUSLE approachcombined with GIS (Bera 2017; Das et al. 2018).Most of the studies have done quantitativeresearch on soil eroded from a particular region.However, decision and policy makers are givingmore emphasis on the spatial extent in comparisonto the quantity of soil eroded (Eltner et al. 2013;Cerd�a et al. 2017). Hence, data on spatial distri-bution pattern along with its magnitude of thesternness of soil erosion-prone areas will aid inmaking of strategic decision to minimize the soilerosion risk with incorporating political and socio-economic matter alongside (Prasuhn et al. 2013).Thus, it requires integration of some expertknowledge-based methods to existing soil erosionmodels to assess the soil erosion susceptibilityassessment.Based on various environmental factors,

researchers had used various methods of deter-ministic and probabilistic to fuzzy logic or artiBcialneural network (ANN) approaches such as multi-criteria evaluation, weighting and ranking basedmethod, analytical hierarchical process (AHP) andmodelling to review the degree of erosion risk (Ohet al. 2009). However, it has shown difBculty inquantifying the soil erosion interaction with itsenvironmental factors due to complexity in soilerosion process (Kheir et al. 2006). Therefore, itnecessitates an integrated approach which sys-tematically analyzes various environmental factorsinCuencing soil erosion process and quantifying itsseverity. From the above, the important and mostwidely used technique is the AHP method reportedby Saaty (1980). AHP technology coupled withgeospatial technology has acquired significantattention among the various researchers of recenttimes (Rozos et al. 2011; Li et al. 2015; Das2018, 2019a, b). Several researchers have appliedthe integrated approach of AHP, GIS and RUSLEtechniques to determine soil erosion-prone areas(Nasiri 2013; Abbasi et al. 2017; Thomas et al.2018). The present study has been carried out inthe West Kameng watershed of Arunachal Pra-desh, as the study region falls in high precipitouszone along with unstable geological formations

94 of 18 J. Earth Syst. Sci. (2020) 129:94

which caused frequent soil erosion from hillymountains in monsoon season. Hence, it is neces-sary to map the extent of soil erosion and probablesoil erosion prone areas in the study region forthe futuristic developmental activity in the region.The present study aims to elucidate two differentnatures of soil erosion processes such as quantita-tive analysis of soil erosion by RUSLE model andspatial distribution and extent of soil erosion haz-ard risk through AHP. Both of these methods wereincorporated with geospatial technology to Bnd outthe soil erosion-prone areas of the study regions.

2. Study site and methodology

The study was conducted in West Kameng water-shed (26�540–28�010N; 91�300–92�400E) of ArunachalPradesh having geographical coverage of 7422 km2,i.e., 8.86% of the total state geographical area(Bgure 1). Study area represents part of the EasternHimalayan range sharing an international boundarywith Bhutan in the west, the state of Assam in thesouth, Tawang district in north and East Kamengdistrict in the east. Physiography is characterized byundulating terrain and valleys with an altitudinalrange of 115–5780 masl. Geologically, study area isclassiBed into nine geological rock formations withmaximum coverage of Ziro granite gneiss formation(1515 km2) followed by Dirang, Tenga, high-gradeSchist and Gneiss formations of Sela group, Tour-maline granite, Chillipam formation, Bharali, lowerSiwalik formation and minimum in Bichom forma-tion. Geomorphology of the watershed has beendivided into three groups, e.g., Denudo structuralhills, structural hills and valley Bll. Soil texture wascoarse loam to gravelly sandy loam. The forest coverof the district is 86.2% and the major vegetationtypes include subtropical broad-leaved and pineforests. Certain areas of the district such asBomdila,Dirang and Shergaon exhibit temperate broad-leaved and temperate coniferous forests which ismostly dominated by oaks and members of Magno-liaceae and Ericaceae, particularly, the Rhododen-drons. The average rainfall is about 1607 mm withhot and humid climate up to 1200 masl. The coldclimate prevails in the northern part and averagetemperature varies between 0.1o and 31�C.

2.1 Data acquisition

The foremost task of the proposed research is toevaluate soil loss and erosion process and to

prepare a spatial map of soil erosion hazard-proneareas. Hence, it is essential to study the suit-able physical parameter such as precipitation,digital elevation model (DEM), land-use patternand soil type which largely aAects the soil erosion.The major variables used in current soil erosionstudy are (a) annual precipitation (15 yrs) datawhich was collected from climate hazard groupinfrared precipitation with station data (CHIRPS)database (ftp://chg.ucsb.edu/pub/org/chg/products/CHIRP/monthly/india/) for the period2001–2015 from which rainfall runoA erosivityfactor was determined, (b) length of slope andsteepness factor were derived from high-resolutionPALSAR DEM having 12.5 m resolution collectedfrom vertex (https://vertex.daac.asf.alaska.edu/),(c) land-use and normalized difference vegetationindex (NDVI) was determined from Sentinel-2(13th December 2016) data downloaded from(https://scihub.copernicus.eu/dhus), and (d) soiltype and its attributes were collected from mapNBSS & LUP, Kolkata. Flowchart for methodol-ogy adopted in the present study is given inBgure 2.

2.2 Soil, precipitation and land-use

Soil texture of the study area was classiBed intonine soil textural class through digitizing districtsoil map acquired from NBSS & LUP. Studyrevealed that maximum area coverage was byloamy soil (347,010 ha) followed by sandy loam(60,326 ha), coarse loam (26,359 ha), Bne loam(25,389 ha), rocky mountain (17,459 ha), sandyclay loam (10,644.74 ha), gravelly loam (3667 ha),landslide (897 ha), and gravelly sandy loam (124.97ha) type (Bgure 3a). Spatial variability in theintensity of precipitation is the most importantfactor aAecting soil erosion. However, only a fewmetrological stations are found in the study area.To overcome this problem, the gridded rainfalldata of CHIRPS with station data for the last 15yrs was used. The annual precipitation ranged from415 to 2600 mm (Bgure 3b). For identifying the soilerosion prone area it is important to know thepresent land-use, as the proposed site falls on theeastern Himalayan zone which is characterized byundulating topography, hence it is proned tolandslides during rainy days and causes heavylosses. LULC map was prepared following super-vised classiBcation using Sentinel-2 images and hasbeen classiBed into 10 LULC classes (Bgure 4a).

J. Earth Syst. Sci. (2020) 129:94 of 18 94

ftp://chg.ucsb.edu/pub/org/chg/products/CHIRP/monthly/india/

ftp://chg.ucsb.edu/pub/org/chg/products/CHIRP/monthly/india/

https://vertex.daac.asf.alaska.edu/

https://scihub.copernicus.eu/dhus

2.3 Geomorphology, slope and relative relief

Geomorphologymapping involves the identiBcationand characterization of different landform types.Landform genesis is inCuenced by various factorssuch as slope, existing drainage pattern and under-lying lithology of the region. The study area com-prised of rocky terrain. Six geomorphology classeswere extracted from level-2 geomorphology datasetof NRSC Bhuvan data portal (Bgure 4b). The studyarea has the dominance of structural origin forma-tions of moderately dissected hills and valleys. Theslope is an angle formed at the junction of ant surfaceof the earth and a horizontal datum. The slope isan important driver of soil erosion process. Theslope was calculated from high-resolution DEM

(resolution of 12.5 m) of ALOS PALSAR and was

categorized into Bve (Bgure 4c) classes (gentile

(0–5�), marginal (5–15�), moderate (15–25�), high(25–35�), and very high ([35�)). Relative relief

represents the altitudinal difference at a given point.

There is a decrease in safety factor with an increase

in altitude. Thus, between two slopes having similar

properties except for elevation, higher the slope,

higher will be the probability of landslide and soil

erosion. In this area, runoA is higher and inBltration

is lower due to its steepness. Besides this, the slope

saturation also enhances the shear forces through

drag by reducing the shear resistance of the slope

(Nagarajan et al. 2000; Pandey et al. 2008). The

neighborhood function of spatial extension tool of

Figure 1. Location map of the study area.


ArcGIS using ALOS PALSAR DEM was used tocreate the relative relief (Bgure 4d).

2.4 Drainage density and drainage frequency

Drainage network inCuences the soil erosionprocess by eroding and transporting the sedimentsthrough streams. It is an important variableinCuencing landslide and soil erosion processes inmountainous regions (Sarkar and Kanungo 2004).

The area having greater density signiBes morenumber of streams thereby higher possible rate ofsoil erosion. Drainage network of the study areawas extracted from the ALOS PALSAR DEM(Bgure 5a) to prepare the drainage density map.Study area is classiBed into Bve drainage classes,viz., very low (0–500 m/km2), low (501–1000m/km2), moderate (1001–1500 m/km2), high(1501–2000 m/km2), and very high ([2000m/km2). The number of streams per unit area isthe drainage frequency. The unit area in this

Figure 2. Flow chart for steps involved in the soil hazard zonation mapping.

Figure 3. Soil texture (a) and precipitation map (b) of the study area.


study is 1 km2 and drainage frequency rangesfrom 0 to 35 per km2. The study area was clas-siBed into three drainage (Bgure 5b) frequencyzones, viz., low (\10), moderate (10–20), and high([20).

2.5 Lineament density and frequency

Lineaments represent the crustal structure likefaults, fracture, crack which controls the movement

of surface-runoA and groundwater recharge. The

lineament of the study area was extracted from

PALSAR DEM derived shaded relief of differentsolar (Bgure 5c) azimuths (0�, 45�, 90�, 135�, 180�,225�, 270�, 315�) with solar elevation of 30� andambient light setting of 0.20 constant were taken.

Two different sets of shaded relief [(0�, 45�, 90�,135�) and (180�, 225�, 270�, 315�)] were prepared

by overlay operation. The automatic lineamentextraction technique of PCI Geomatica was used to

map all the lineaments covering the study region(Abdullah et al. 2010; Das and Singh 2016; Das et al.

Figure 4. (a) LULC; (b) Geomorphology; (c) Slope; (d) Relative relief map of the study area.


2018). The lineament validation process was carriedout by overlaying the two sets of lineaments on top of

the drainage network of study area and we foundthat the Brst set with azimuth have the negative

relationshipwith the lineaments and later set has thepositive relationship, hence used in the study. The

Bve lineament density classes were used (Bgure 5c),i.e., low (0.500 m/km2), low (501–1000 m/km2),

moderate (1001–1500 m/km2), high (1501–2000m/km2), and very high ([2000 m/km2). Lineament

frequency of the study area ranged from 10 to 54. Itwas classiBed into Bve classes (Bgure 5d), viz., verylow (\10), low (10–20), moderate (20–30), high(30–40), and very high ([40).

2.6 Soil erosion models

The spatial mapping of soil erosion process wasestimated using the RUSLE model and it relies onvarious factors, viz., rainfall-runoA erosivity (RE),

Figure 5. Drainage density (a), drainage frequency (b), lineament density (c), and lineament frequency (d) map of the studyarea.


soil erodibility (E S), slope length (L S), cover andmanagement (CM), and conservation practices(PC). Due to the simplicity of the RUSLE model, itis the widely used soil erosion model and can beapplied to a diverse region with multifaceted landuse sectors (Renard et al. 1997). RUSLE model canalso be applied at a continental scale (Van DerKnijA et al. 2000), on a territorial scale (Van DerKnijA et al. 1999) and at the basin scale (Bala-subramani et al. 2015; Toubal et al. 2018). Theequation adopted for the RUSLE model in thecurrent study of soil erosion is as follows:

A ¼ RE � ES � LS � CM � PC; ð1Þ

where,A is the estimated average annual soil loss (t/ha/yr), RE is the rainfall-runoA erosivity factor (MJmm/ha/h/y), E S is the soil erodibility factor (t/ha/MJ/mm), L S is the slope length factor, CM is thecovermanagement factor, andPC is the conservationsupport practices factor for the study area. The LS,CM, and PC factor values are dimension less values.To predict the spatial attribute of different

parameters of RUSLE model, equation (1) wasintegrated into the map algebra tool of the spatialanalysis toolbox of ArcGIS 10.3, where all thefactors were multiplied to estimate the soil erosionrate of the study area. This model shows variedvalues for land and water bodies, hence, soileroded in water bodies were excluded from theanalysis.

2.7 Rainfall-runoA erosivity factor (RE)

RE calculation needs rainfall data on a daily basis.However, it was considered as a major challengingtask due to unavailability of high-resolution rainfalldatabase for the area. In such circumstances, vari-ous empirical models were used for RE factor cal-culation using rainfall data (Millward and Mersey1999; Lee and Heo 2011). In the current study, thelinear relationship developed by Choudhury andNayak (2003) was used to evaluate the RE factor.

RE ¼ 79þ 0:363 �Xa; ð2Þ

where Xa is the average annual precipitation.

2.8 Soil erodibility factor (E S)

E S indicates the different rate at which soils erodein a region. The different soils have a different rateof soil erosion because of varied soil characteristics.

However, soil erodibilty is mainly inCuenced bysoil texture and organic substances as reportedby Wischmeier and Smith (1978). ES factor forthe present study was estimated based on soiltexture class correlation with organic matter asgiven by Schwab et al. (1981) and presented intable 1.

2.9 Slope length and steepness factor (LS)

It determines the steepness of the landscape andprovides the eAect of topography on soil loss. LS

factor was estimated by applying two equationsusing PALSAR DEM data (Shinde et al. 2011):

L ¼ 0:4 � SP þ 40; ð3Þ

where, L is the slope length and SP is the slopesteepness (%).The equation of slopes LS up to 21% was

calculated:

LS1 ¼ L

22:1

� �� 65:41 � ðSin hÞ2 þ 4:56 � Sin hþ 0:065� �

:

ð4Þ

However, equation for[21% slopes is as follows:

LS2 ¼ L

22:1

� �0:7

� 6:432 � ðSin hÞ7 � Cos h� �

; ð5Þ

where, h is the angle of the slope and L is the lengthof slope (m). Further, two different resulted LS

factor values were integrated into one single LS

map for the study area.

2.10 Cover management factor (CM)

Vegetal cover, in general, minimizes the soilerosion through obstructing direct rainfall on soil

Table 1. Soil erodibility factor for different texturalclass.

Textural class

Organic matter (%)

\0.5 2

Coarse loamy 0.27 0.24

Fine loamy 0.35 0.30

Gravelly loam 0.12 0.10

Gravelly sandy loam 0.27 0.24

Loamy 0.38 0.34

Sandy clay loam 0.27 0.25

Sandy loam 0.27 0.24

Landslide – –

Rocky mountain – –


particles. CM factor is the ratio of soil loss fromcropped land under the speciBc condition to theconsequent loss from continuous fallow land. CM

factor is determined through the satellite-derivednormalize differential vegetation index (NDVI)using Sentinel-2 data following the equationdeveloped by Van Der KnijA et al. (1999).

CM ¼ exp �a � NDVI

b�NDVIð Þ

� �; ð6Þ

where a and b are the shape of NDVI-C curve, thevalue for a and b is 2 and –1, respectively. CM

values of the present study ranges from 0 to 1.

2.11 Support practice factor (P S)

P S factors are related to the agronomial practices(contour tillage, strip cropping, and terrace) thatmodiBes the surface runoA Cow pattern, andminimizes the runoA rate and hence prevents thesoil erosion (Renard and Foster 1983). The P S

factor is the ratio of soil loss with speciBc supportpractice to the equivalent loss with upslope anddownslope tillage (Renard et al. 1997). P S factorwas estimated from Sentinel-2 derived LULCmap, and values for different LULCs wereassigned following the list provided in USDAhandbook (1981). P S value for the rocky area wasconsidered as zero as these areas are devoid of soilerosion.

2.12 Analytical hierarchy process (AHP)

AHP technique is applied in the present study. It isa multi-criteria decision analysis method whichanalyzes different environmental parameters andinvolves ranking based on certain criteria to solve avery complex decision-making process (Saaty andVargas 2001). Hierarchy of an environmentalparameter was based on its interdependency andinterrelationship between other parameter andhierarchy was done through higher to the lowerlevel. This technique uses both subjective andobjective factors in the decision-making processes(Vijith et al. 2012; Yasser et al. 2013). Further, ituses a pairwise comparison of a different parameterwhich inCuences the soil erosion process. Thepairwise comparison was done by the help of amatrix of the parameter and giving weight to theparameter with respect to their inCuence on otherparameter expressed in a numerical scale and also

consistency ratio (Saaty 1980). The pairwisecomparison matrix can be represented as:

M ¼

1 a12 a13 . . . a1na21 1 a22 . . . a2na31 a32 1 . . . . . .

..

. ... ..

. ... ..

.

an1 an2 . . . . . . 1

2666664

3777775;

where

aij ¼Weight of attribute i

Weight of attribute j: ð7Þ

The weight may vary in the numerical scale of 1and 9. The element of 1/1 in the matrix have equalimportance for both parameters i and j. Thepresent study considered eight individualparameters such as LULC, geomorphology, soil,slope, drainage density, drainage frequency,lineament density, and lineament frequency of thestudy area. Though geology of study area hassignificant inCuence on the soil erosion, it was notconsidered in the present study. However, soilerosion is also dependent on the slope and currentland use and land cover prevailing in the studyarea. Apart from these, majority of the study areais covered with dense forest cover with undulatinghilly terrain. The calculation of the weight matrixwas done using pair-wise comparison matrix. Theeigenvector method was used for determining theweight of attributes. The primary eigenvector ofmatrix oAers the weight value of each factor and itwas computed by adding each column values ofpair-wise comparison matrix, and afterward, thevalues in each cell were divided by the summedvalues of the same factor (Saaty 1980). Theestimation of the consistency level of the matrixof order n through the consistency index (CI)calculation (equation 8):

CI ¼ kmax � n

n � 1; ð8Þ

where, kmax is the highest eigenvalue of the matrix.The kmax will be n if the pair-wise comparison didnot have any consistency and resulting CI to bezero. The consistency ratio (CR) was calculatedto measure the consistency of the pairwise-comparison matrix (Kurttila et al. 2000)

CR ¼ CI

RI; ð9Þ


where, RI is the average of resultant consistencyindex of randomly generated comparison and itvaries according to order of the matrix (Saaty1980). The CR of the matrix should be less than orequal to 0.1 to be accepted as consistent. Thechanges in variables in analysis depend on the CRvalue. The CR [ 0.1 endorses the omission ofvariable and CR\ 0.1 will allow the variables inthe analysis.

2.13 Soil erosion susceptibility validation

Accuracy of the prediction model is a crucialrequirement in susceptibility modelling approach,as it is the best available technique to identify thefuture probable zones of hazard and its extent inthe study area. Hence, for the present study, theaccuracy of the model can be done by validatingthe soil erosion susceptibility map with the his-torical known soil erosion locations occurred in thestudy area as well as overlaying the soil erosionlocation on recent satellite image. The validity ofsoil erosion susceptibility map can also be graphi-cally done by applying area under curve (AUC)method (Pradhan et al. 2010; Feizizadeh et al.2014; Tehrany et al. 2017). The AUC method is thewidely used approach to determine the accuracy ofpredictive model by interpreting its ability topredict the hazard locations (Feizizadeh et al.2014).

3. Results and discussion

3.1 Rainfall-runoA erosivity factor (RE)

The analysis of CHRIP rainfall data revealed thatthe average yearly rainfall of the study area rangedbetween 415 and 2600 mm. The spatial distributionof observed rainfall erosivity varied greatly acrossthe study region. The average estimated RE valuewas 421 mm/ha/h/Yr and it ranged between 212and 851 MJ mm/ha/h/Yr (Bgure 6a). The south-east part of the study area is having higher RE

factor and it decreases gradually in a northwesterndirection. This could be mainly due to the south-eastern leeward side of the mountain acted as abarrier to the south-west monsoonal rainfall comesfrom the Bay of Bengal which might have causedhigher rainfall in a different part of the site. Theobserved RE value can be correlated with theBndings of 1062–2014 MJ mm ha/h/Yr from Karsowatershed of Damodar Barakar catchment and 267and 694 MJ mm ha/h/Yr in the Cat area and themountainous steep area (Alexakis et al. 2013). Thelower RE value (333.6–412 MJ mm ha/h/Yr) incomparison to present Bndings was also reportedfrom Andipatti watershed of Tamil Nadu (Pandeyet al. 2007; Balasubramani et al. 2015). However,much higher RE value (1514.66 MJ mm ha/h/Yr)was observed by Prasannakumar et al. (2012) inthe mountainous watershed, Kerala.

Figure 6. Rainfall erosivity (a) and soil erodibility (b) factor map of the study area.


3.2 Soil erodibility factor (E S)

Soil texture map was used for extracting soilerodibility factor (E S) factor and showed thediverse type of the soil texture in the present studysites such as coarse-loamy, Bne loamy, gravellyloam, gravelly sandy loam, loamy, sandy clayloam, and sandy loam. E S values of the study arearanged from 0 to 0.38 t ha/MJ/mm with a meanvalue of 0.27 t ha MJ/mm (Bgure 6b). The presentvalue of E S factor can be compared with the valuesreported by various researchers. E S factor rangedbetween 0.16 and 0.28 t ha/MJ mm in the hillyterrain of Tirap district of Arunachal Pradesh (Daset al. 2018). The E S factor from 0.09 to 0.48 t ha/MJ/mm was reported from the mountainouswatershed of lower Himalayas (Sheikh et al. 2011)and 0.14–0.37 t ha/MJ/mm from the agriculturalwatershed of Western Ghats region (Pradeep et al.2014).

3.3 Slope length and steepness factor (LS)

PALSAR-DEM derived slope data was used toestimate the LS factor and the values varied from0 to 39.57 (Bgure 7a). It was visualized that theslopes having an angle of 0–15� have the lower LS

values (0–5). Similarly, higher slope degree in thestudy area showed a higher LS factor. Based on the

result, it can be corroborated with the LS factorreported by various researchers. A lower LS value(0–8.22) for the iron ore mining areas of Jharkhandwas reported (Kayet et al. 2018) and the values(0–20) reported by Parveen and Kumar (2012) iscomparatively low with respect to values reportedin the present Bndings. LS factor values rangingfrom 0.07 to 58.59 was reported from the southernWestern Ghats river basin (Thomas et al. 2018).Similar LS values from the present Bndings rangingfrom 0 to 44.62 were reported by Das et al. (2018)from the hilly terrain of Arunachal Pradesh.However, much higher LS values (0–107.28) werereported by Demirci and Karaburun (2012) thanthe values resulted from the present study.

3.4 Cover management factor (CM)

CM factor is another significant variable after theLS factor and it is expressed as an eAect of croppingand management practices upon soil surface. Sen-tinel-2 derived NDVI was used to quantify the CM

factor and its spatial variability. The NDVI is animportant indicator as it shows the type of cover onthe soil layer. Calculated CM factor ranges from 0to 1.28 (Bgure 7b). The lower value represented thevegetation cover and higher values represent therocky area or barren land covered in the studyregion.

Figure 7. Slope length (a) and cover management (b) factor map of the study area.


3.5 Support practice factor (P S)

Spatial distribution of conservation supportpractice on soil loss ranges between 0 and 1(Bgure 8). The lower P S value represents theavailability of higher conservation practice on thesoil in case of settlement, and higher values of P S

represent lesser non-availability of conservationpractice. Conservation support practice in thepresent study showed that the northern part wascovered by the rocky mountain, snow-coveredareas and have the higher P S value of 1. Also, thedense forest and barren land area showed the P S

value of 1 as no management supportive practiceswere absent in the region. The land covered by thewater was not included in the study as it has the P S

value of 0 (table 2).

3.6 Estimation of soil erosion rate

The spatial modelling of annual soil erosion ratewas estimated integrating all the Bve RUSLE fac-tors with the geo-statistical techniques. The pre-dicted average soil erosion rate was 124.21 t/ha/Yrand the values ranged between 0 and 5817.17 t/ha/Yr (Bgure 9). The predicted erosion rate for thestudy area is high as compared to the values(0–87.62 t/ha/Yr) reported by Das et al. (2018)from hilly terrain of Tirap District, 0–61.4 t/ha/Yrfrom Himalayan watershed (Sheikh et al. 2011) and134 Mg/ha/Yr from Kenyan highland (Angimaet al. 2003). A similar range of annual soil erosionrate (0–4227 t/ha/Yr) was also reported which

corroborates with the values in the present study(Pradeep et al. 2014). However, Bndings of thepresent study reveal quite lower soil erosion ratewhen compared with the Bndings of Alexakis et al.(2013) from the island of Cyprus (0–6394 t/ha/Yr).Predicted average soil erosion for the differentLULC classes revealed higher (417.54 t/ha/Yr) soilerosion rate for the sandy areas followed by rockymountain (214.82 t/ha/Yr), dense forest (180.50t/ha/Yr) and settlement (29.62 t/ha/Yr) areas(table 3), and this could be mainly due tothe intensity of management practices over theparticular land use and land cover type.

3.7 Soil erosion hazard assessment

The spatial map of soil erosion potential zone wasdelineated using AHP plugin in the ArcGIS plat-form and RUSLE approach. The eight differentgeo-environmental variables (LULC, geomorphol-ogy, slope, relative relief, drainage density, drai-nage frequency, lineament density, and lineamentfrequency) of the study area were statisticallyassessed and their inCuences on soil erosion wereanalyzed through AHP technique to predictthe vulnerable soil erosion areas of the study site.The AHP analysis starts with the analysis of dif-ferent individual parameter rating by manualassessing of the imported value. The pair-wisecomparison matrix normalizes eigenvector differ-ent feature layer is given in table 4. Normalizeeigenvector (ratings) values ranged from 0.02 to0.20. Among different LULC classes, the majorityof the area was covered by dense vegetation havingeigenvector of 0.08. The most dominant landformwas hills and valleys of structural origin and prin-cipal eigenvector values ranged between 0.04 and0.39 with higher eigenvector was observed for thedenudational slope (0.39) and least eigenvector forthe water body (0.04).The slope of the study area inCuences the soil

erosion process by controlling surface Cow. Slopeswere grouped into Bve classes and it showed thatarea having[35� slope angles is having maximumeigenvector of 0.44 which decreases with thedecrease in slope angle. Hence, it is inferred thatthese areas are having higher vulnerability to soilerosion. The relative relief had also a significantrole in soil erosion process similar to that of theslope. Pair-wise comparison of relative reliefrevealed that the principal eigenvector rangedbetween 0.04 and 0.36 and higher eigenvector was

Figure 8. Map showing spatial variability of support practicefactor in the study area.


observed for relative relief class of 1000 m per km2

and more. The pair-wise comparison of drainagedensity depicted that area having higher densityshowed higher principal eigenvector (0.40), thushigher vulnerability to soil erosion. Similarly,drainage frequency of the study area also showed asimilar trend and greater eigenvector (0.54) wasobserved for [20 frequency class. Among the

various lineament density class of the study, high-density class (2000 m per km2 and more) have thehigh principal eigenvector (0.36). The lineamentfrequency had a similar trend to that of lineamentdensity and greater eigenvector was observed for[40 frequency class. The pair-wise comparison ofthe different parameters was shown in table 4. TheAHP analysis of different individual parameters

Table 2. Area statistics (ha) and predicted average soil erosion rate for different landuseclasses.

LULC class

Area

p

Average soil erosion

rate (t/ha/Yr)(Ha) (%)

Agriculture land 2321 0.47 0.5 73.59

Degraded forest 42026 8.54 0.8 48.27

Dense forest 216983 44.11 1.0 180.50

Grass land 30332 6.17 0.9 107.54

Open forest 134722 27.39 0.8 44.34

Rocky area 40948 8.33 1.0 214.82

Sandy area 684 0.14 1.0 417.54

Settlement 5342 1.09 0.1 29.62

Snow area 14754 3.00 1.0 80.89

Water body 3750 0.76 0.0 –

Figure 9. Spatial distribution of average annual soil erosion rate in the study area.


Table 3. Pair-wise comparison matrix and eigenvector (rating) of the different feature layer.

Factors

LULC WB DGF DNF GL OF RA SA STM SNA AL Ratings

Water body (WB) 1 0.14 0.33 0.50 0.20 1.00 1.00 1 1 0.14 0.04Degraded forest (DGF) 7 1 2.33 3.50 1.40 7.00 7.00 7 7 1 0.24Dense forest (DNF) 3 0.43 1 1.50 0.60 3.00 3.00 3 3 0.43 0.10Grass land (GL) 2 0.29 0.67 1 0.40 2.00 2.00 2 1 0.14 0.06Open forest (OF) 5 0.71 1.67 2.50 1 5.00 5.00 5 5 0.71 0.17Rocky area (RA) 1 0.14 0.33 0.50 0.20 1 1.00 1 1 0.14 0.04Sandy area (SA) 7 0.14 0.33 0.50 0.20 1.00 1 1 1 0.14 0.04Settlement (STM) 1 0.14 0.33 0.50 0.20 1.00 1.00 1 1 0.14 0.04Snow area (SNA) 1 0.14 0.33 0.50 0.20 1.00 1.00 1 1 0.14 0.04Agriculture land (AL) 7 1.00 2.33 3.50 1.40 7.00 7.00 7 7 1 0.24CR 0.019

Geomorphology WB FO LHV MHV HHV DOWater body (WB) 1 0.50 0.25 0.20 0.17 0.11 0.04Fluvial origin (FO) 2 1 0.50 0.40 0.33 0.22 0.09Structural origin (LHV) 4 2 1 0.80 0.67 0.44 0.17Structural origin (MHV) 5 2.50 1.25 1 0.83 0.56 0.22Structural origin (HHV) 6 3 1.50 1.20 1 0.67 0.26Denudation origin (DO) 9 4.50 2.25 1.80 1.50 1 0.39CR 0.0210

Slope A B C D E\ 5 (A) 1 0.33 0.20 0.14 0.11 0.055–15 (B) 3 1 0.60 0.43 0.33 0.1515–25 (C) 5 1.67 1 0.71 0.56 0.2425–35 (D) 7 2.33 1.40 1 0.78 0.34[35 (E) 9 3 1.80 1.29 1 0.44CR 0.036

Relative relief A B C D E\ 250 (A) 1 0.33 0.20 0.14 0.11 0.04251–500 (B) 3 1 0.60 0.43 0.33 0.12501–750 (C) 5 1.67 1 0.71 0.56 0.20751–1000 (D) 7 2.33 1.40 1 0.78 0.28[ 1000 (E) 9 3 1.80 1.29 1 0.36CR 0.0175

Drainage density A B C D E\ 500 (A) 1 0.33 0.20 0.14 0.11 0.05501–1000 (B) 3 1 0.60 0.43 0.33 0.141001–1500 (C) 5 1.66 1 0.71 0.56 0.221501–2000 (D) 7 2.33 1.40 1 0.78 0.31[2000 (E) 9 3 1.80 1.29 1 0.40CR 0.0348

Drainage frequency A B C\ 10 (A) 1 0.20 0.14 0.0810–20 (B) 5 1 0.71 0.38[20 (C) 7 1.4 1 0.54CR 0.0022

Lineament density A B C D E\ 500 (A) 1 0.33 0.20 0.14 0.11 0.04501–1000 (B) 3 1 0.60 0.43 0.33 0.121001–1500 (C) 5 1.67 1 0.71 0.56 0.201501–2000 (D) 7 2.33 1.40 1 0.78 0.28[ 2000 (E) 9 3.00 1.80 1.29 1 0.36CR 0.0003


Table 3. (Continued.)

Factors

LULC WB DGF DNF GL OF RA SA STM SNA AL Ratings

Lineament frequency A B C D E\ 10 (A) 1 0.33 0.20 0.14 0.11 0.0510–20 (B) 3 1 0.60 0.43 0.33 0.1520–30 (C) 5 1.67 1 0.71 0.56 0.2530–40 (D) 7 2.33 1.40 1 0.78 0.35[ 40 (E) 9 3.00 1.80 1.29 1 0.46CR 0.0270

Table 4. Pair-wise comparison matrix of different parameters.

Factor

Parameter A B C D E F G H Ratings

LULC 1 1.80 0.90 1.28 4.50 1.80 3.00 9.00 0.21

Geomorphology 0.56 1 0.50 0.71 2.50 1.00 1.67 5.00 0.12

Slope 1.11 2.00 1 1.43 5.00 2.00 3.33 10.00 0.23

Relative relief 0.78 0.70 1.40 1 3.50 1.40 2.33 7.00 0.17

Drainage density 0.22 0.29 0.20 0.40 1 0.40 0.67 2.00 0.08

Drainage frequency 0.56 2.50 0.71 0.50 1.00 1 1.67 5.00 0.13

Lineament density 0.33 0.60 1.50 0.43 0.30 0.60 1 3.00 0.08

Lineament frequency 0.11 0.33 0.20 0.50 0.14 0.10 0.20 1 0.03

CR 0.078

Figure 10. Soil erosion potential zone map of the study area.


showed that the slope has the maximum normal-ized eigenvector (0.23) followed by the LULC,relative relief, drainage frequency, and lineamentfrequency with consistency ratio of 0.07 as both theslope and type of LULC on the study region governthe soil erosion process. The potential zone of soilerosion hazard was delineated (Bgure 10) and cat-egorized into Bve classes, viz., nil, low, moderate,high and critical based on the weightage andranking of all the parameters included in study.Soil erosion hazard zonation map showed that thestudy area is moderate to critical vulnerable to soilerosion which could be mainly due to the fact thatmost of the study area cover is having forest cover(table 5). Areas having high to critical soil erosionpotential were mostly located on a higher slopewith higher drainage and a greater number of lin-eament distributions. Soil erosion hazard zonationarea statistics showed that the moderate class hashigh areal coverage (279,311 ha) followed by higherosion hazard class (126,272 ha) and least for thenil soil erosion class.

3.8 AUC-based model validation

The performance of the soil erosion assessment ofAHPmethod showed that AUC for the present studywas 0.84 (84%) (Bgure 11), which indicates a betteraccuracy of the AHP model in the present study.

AUC observed in the present study can be comparedwith the AUC reported by 77.1% by Hembram andSaha (2018) and 83.1% by Tehrany et al. (2017). Theaccuracy assessment result of the present studyrevealed that the AHP modelling approach of soilerosion susceptibility for the study area explain thesoil erosion hazard occurrence locations in the bestpossible way by considering selected physicalparameters of the selected watershed.

4. Conclusions

The results indicated that most of the drivers takeninto account of the present study are among keyfactors driving to soil erosion. Higher rainfall-runoA erosivity is observed in the southeast zone ofwatershed area and soil texture showed greatvariability from sandy loam to gravelly sandyloam. Soil erodibility increases with an increase inslope angle and predicted average soil erosion rateis 124.21 t/ha/Yr. The area having more slope,relative relief, drainage density, and lineamentdensity, and frequency is highly vulnerable to soilerosion. Low soil erosion hazard in some area couldbe mainly due to the presence of forest cover.Majority of watershed areas are prone to soil ero-sion thus require appropriate sustainable manage-ment practices. Further, to minimize the rate ofsoil erosion it is a pre-requisite to review theexisting scientiBc management techniques andevolve suitable conservation measures viable topresent watershed area. These activities will notonly minimize the erosion but also improve the soilhealth and crops productivity thereby livelihood ofthe peoples residing in the study area will also beimproved.

Acknowledgements

Authors are thankful to the Department of Scienceand Technology, New Delhi for partial Bnancialsupport through AICP carbon sequestration pro-ject. We are also thankful to the Director, NERISTand Head, Department of Forestry, NERIST forextending all laboratory facilities and Departmentof Environment and Forest, Govt. of ArunachalPradesh for all logistic and Beld support during thecourse of this study. Authors are also thankful toall free database and satellite data providers whosedata were downloaded from their web portal andused in the present study.

Table 5. Area statistics of soil erosion zone.

Soil erosion

hazard class

Area

(Ha)

Nil 6879.85

Low 61222.90

Moderate 279311.11

High 126271.83

Critical 10297.49

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1Cum

ula

tive soil

ero

sio

n o

ccure

nce

(%)

Cumulative area (%) under susceptibility

AUC= 84.9%

Figure 11. Area under curve (AUC) for validation of predic-tion Soil erosion susceptibility.


References

Abbasi P, Ahmadi H, Khan M and Moeini A 2017 Evaluationof soil and water conservation projects through estimationof erosion intensity by geomorphological modelling (casestudy of Safaroud watershed, Mazandaran province, Iran);Appl. Ecol. Environ. Res. 15(3) 1739–1751.

Abdullah A, Akhir J M and Abdullah I 2010 Automaticmapping of lineaments using shaded relief images derivedfrom Digital Elevation Model (DEMs) in the Maran-SungiLembing area, Malaysia; Electron. J. Geotech. Eng.15 949–957.

Alexakis D D, Hadjimitsis D G and Agapiou A 2013 Integrateduse of remote sensing, GIS and precipitation data for theassessment of soil erosion rate in the catchment area ofYialias in Cyprus; Atmos. Res. 131 108–124.

Angima S D, Stott D E, O’Neill M K, Ong C K and Weesies GA 2003 Soil erosion prediction using RUSLE for centralKenyan highland conditions; Agr. Ecosyst. Environ.97(1–3) 295–308.

Balasubramani K, Veena M, Kumaraswamy K and Saravan-abavan V 2015 Estimation of soil erosion in a semi-aridwatershed of Tamil Nadu (India) using revised universalsoil loss equation (RUSLE) model through GIS; Model.Earth Syst. Environ. 1(3) 1–17.

Bera A 2017 Assessment of soil loss by universal soil lossequation (USLE) model using GIS techniques: A case studyof Gumti river basin, Tripura, India; Model. Earth Syst.Environ. 3 29.

Cerd�a A, Keesstra S D and Rodrigo-Comino J et al. 2017RunoA initiation, soil detachment and connectivity areenhanced as a consequence of vineyards plantations;J. Environ. Manag. 202 268–275.

Choudhury M K and Nayak T 2003 Estimation of soil erosionin Sagar lake catchment of central India; In: Proceedings ofthe International Conference on Water and Environment,Dec 15–18, Bhopal, India, pp. 387–392.

Das B and Singh S K 2016 Ground water potential zonemapping of semi-arid region of Kalaburgi and Yadgirdistricts of North Karnataka: A geospatial analysisapproach; Int. J. Curr. Res. 8(3) 28,797–28,807.

Das B, Paul A, Bordoloi R, Tripathi O P and Pandey P 2018Soil erosion risk assessment of hilly terrain throughintegrated approach of RUSLE and geospatial technology:A case study of Tirap District, Arunachal Pradesh; Model.Earth Syst. Environ. 4(1) 373–381.

Das S 2018 Geographic information system and AHP-basedCood hazard zonation of Vaitarna basin, Maharashtra,India; Arab. J. Geosci. 11 576.

Das S 2019a Geospatial mapping of Cood susceptibility andhydro-geomorphic response to the Coods in Ulhas basin,India; Remote Sens. Appl.: Soc. Environ. 14 60–74.

Das S 2019b Comparison among inCuencing factor, frequencyratio, and analytical hierarchy process techniques forgroundwater potential zonation in Vaitarna basin, Maha-rashtra, India; Groundw. Sustain. Dev. 8 617–629.

Demirci A and Karaburun A 2012 Estimation of soil erosionusing RUSLE in a GIS framework: A case study in theBuyukcekmece Lake watershed, northwest Turkey; Envi-ron. Earth Sci. 66(3) 903–913.

Eltner A, Mulsow C and Maas H 2013 Quantitative measure-ment of soil erosion from TLS and UAV data, International

Archives of the Photogrammetry; Remote Sens. SpatialInform. Sci. XL-1/W2(September):4–6 119–124.

Feizizadeh B, Roodposhti M S, Jankowski P and Blaschke T

2014 A GIS-based extended fuzzy multi-criteria evaluation

for landslide susceptibility mapping; Comput. Geosci. 73

208–221.

Gronsten H A and Lundekvam H 2006 Prediction ofsurface runoA and soil loss in southeastern Norway usingthe WEPP Hill slope model; Soil Till. Res. 85(1–2)186–199.

Hembram T K and Saha S 2018 Geo-environmental evaluationfor exploring potential soil erosion areas of Jainti Riverbasin using AHP model, eastern India; Univ. J. Environ.Res. Tech. 7(1) 38–55.

Joshi V U 2018 Soil loss estimation based on RUSLE along theCentral Hunter Valley Region, NSW, Austrailia; J. Geol.Soc. India 91 554–562.

Kayet N, Pathak K, Chakrabarty A and Sahoo S 2018

Evaluation of soil loss estimation using the RUSLE model

and SCS–CN method in hill slope mining areas; Int. SoilWater Conserv. Res. 6(1) 31–42.

Kheir R B, Cerdan O and Abdallah C 2006 Regional soilerosion risk mapping in Lebanon; Geomorphology 82(3–4)347–359.

Kirkby M J and Morgan R P C 1980 Soil Erosion; Wiley, NewYork.

Kurttila M, Pesonen M, Kangas J and Kajanus M 2000Utilizing the analytic hierarchy process (AHP) in SWOTanalysis – a hybrid method and its application to aforest–certiBcation case; For. Policy Econ. 1(1) 41–52.

Lal R 2001 Soil degradation by erosion; Land Degrad. Dev.12(6) 519–539.

Lee J H and Heo J H 2011 Evaluation of estimation methodsfor rainfall erosivity based on annual precipitation in Korea;J. Hydrol. 409(1–2) 30–48.

Li X, Zhou Y, Tian B, Kuang R and Wang L 2015 GIS basedmethodology for erosion risk assessment of the muddy coastin the Yangtze delta; Ocean Coast. Manag. 108 97–108.

Millward A A and Mersey J E 1999 Adapting the RUSLE tomodel soil erosion potential in a mountainous tropicalwatershed; Catena 38(2) 109–129.

Morgan R P C, Quinton J N and Rickson J R J 1992 Soilerosion prediction model for the Europian Community, GB-ISCO-WASWC.

Nagarajan R, Roy A, Vinod Kumar R, Mukherjee A and KhireM V 2000 Landslide hazard susceptibility mapping basedon terrain and climatic factors for tropical monsoon regions;Bull. Eng. Environ. 58 275–287.

Nasiri M 2013 GIS modelling for locating the risk zone of soilerosion in a deciduous forest; J. For. Sci. 59(2) 87–91.

Oh H J, Lee S, Chotikasathien W, Kim C H and Kwon J H2009 Predictive landslide susceptibility mapping usingspatial information in the Pechabun area of Thailand;Environ. Geol. 57(3) 641.

Oldeman L R, Hakkeling R T A and Sombrok W G 1992World map of the status of human-induced soil degrada-tion; Glasod 52(1) 1–35.

Onyando J O, Kisoyan P and Chemelil M C 2005 Estimationof potential soil erosion for river Perkerra catchment inKenya; Water Resour. Manag. 19(2) 133–143.

Pandey A, Chowdary V M and Mal B C 2007 IdentiBcation ofcritical erosion prone areas in the small agricultural


watershed using USLE, GIS and remote sensing; WaterResour. Manag. 21(4) 729–746.

Pandey A, Dabral P P, Chowdary V M and Yadav N K 2008Landslide hazard zonation using remote sensing and GIS: Acase study of Dikrong river basin, Arunachal Pradesh,India; Environ. Geol. 54 1517–1529.

Parveen R and Kumar U 2012 Integrated Approach of UniversalSoil Loss Equation (USLE) and Geographical InformationSystem (GIS) for soil loss risk assessment in Upper South KoelBasin, Jharkhand; J. Geogr. Inf. Syst. 4(6) 588–596.

Pradeep G S, Krishnan M V N and Vijith H 2014 IdentiBca-tion of critical soil erosion prone areas and annual averagesoil loss in an upland agricultural watershed of WesternGhats, using analytical hierarchy process (AHP) andRUSLE techniques; Arab. J. Geosci. 8(6) 3697–3711.

Pradhan B, Oh H J and Buchroithner M 2010 Weights-of-evidence model applied to landslide susceptibility mapping ina tropical hilly area; Geomat. Nat. Hazards Risk 1 199–223.

Prasannakumar V, Vijith H, Abinod S and Geetha N 2012Estimation of soil erosion risk within a small mountainoussub-watershed in Kerala, India, using Revised UniversalSoil Loss Equation (RUSLE) and geo-information technol-ogy; Geosci. Front. 3(2) 209–215.

Prasuhn V, Liniger H, Gisler S, Herweg K, Candinas A andCl�ement J P 2013 A high-resolution soil erosion risk map ofSwitzerland as strategic policy support system; Land UsePolicy 32 281–291.

Rahman M R, Shi Z H and Chongfa C 2009 Soil erosion hazardevaluation – An integrated use of remote sensing, GIS andstatistical approaches with biophysical parameters towardsmanagement strategies;Ecol.Model. 220(13–14)1724–1734.

Renard K, Foster G, Weesies G, McCool D and Yoder D 1997Predicting soil erosion by water: A guide to conservationplanning with the Revised Universal Soil Loss Equa-tion (RUSLE); Agric. Handbook 703 404.

Renard K G and Foster G R 1983 Soil conservation –

principles of erosion by water; In: Dryland Agriculture (eds)Dregne H E and Wills W O, American Society of Agron-omy, Soil Science Society of America, Madison, WI, USA,pp. 155–176.

Rozos D, Bathrellos G D and Skillodimou H D 2011 Compar-ison of the implementation of rock engineering system andanalytic hierarchy process methods, upon landslide suscep-tibility mapping, using GIS: A case study from the EasternAchaia County of Peloponnesus, Greece; Environ. EarthSci. 63 49–63.

Saaty T L 1980 The Analytic Hierarchy Process, Education,pp. 1–11.

Saaty T L and Vargas L G 2001 Models, Methods, Concepts,and Applications of the Analytic Hierarchy Process; 1st edn,Kluwer Academic, Boston, 333p.

Sarkar S and Kanungo D P 2004 An integrated approach forlandslide susceptibility mapping using remote sensing andGIS; Photogr. Eng. Rem. Sens. 70(5) 617–625.

Schwab G O, Frevert R K, Edminster T W and Barnes K K1981 Soil and Water Conservation Engineering; 3rd edn,Wiley, New York.

Sheikh A H, Palria S and Alam A 2011 Integration of GIS andUniversal Soil Loss Equation (USLE) for soil loss estima-tion in Himalayan watershed; Recent Res. Sci. Technol.3(23) 51–57.

Shinde V, Sharma A, Tiwari K N and Singh M 2011Quantitative determination of soil erosion and prioritiza-tion of micro-watershed using remote sensing and GIS;J. Ind. Soc. Remote Sens. 39(2) 181–192.

Sinha D and Joshi V U 2012 Application of universal soil lossequation (USLE) to recently reclaimed badland along theAdula and Mahalungi Rivers, Pravara basin, Maharashtra;J. Geol. Soc. India 80 341–350.

Tehrany M S, Shabani F, Javier D N and Kumar L 2017 Soilerosion susceptibility mapping for current and 2100 climateconditions using evidential belief function and frequencyratio; Geomat. Nat. Hazards. Risk 8(2) 1695–1714.

ThomasJ,JosephSandThrivikramjiKP2018Assessment of soilerosion in amonsoon-dominatedmountain river basin in Indiausing RUSLE-SDR and AHP; Hydrol. Sci. J. 63(4) 542–560.

Toubal A K, Achite M, Ouillon S and Dehni A 2018 Soilerodibility mapping using the RUSLE model to prioritizeerosion control in the Wadi Sahouat basin, north-west ofAlgeria; Environ. Monit. Assess. 190 210.

USDA 1981 Rainfall Erosion Losses from Cropland East of theRocky Mountain. Handbook no. 282. US Department ofAgriculture, Washington, DC.

Van Der KnijA J, Jones R J A, Montanarella L and Van derKnijA J M 1999 Soil Erosion Risk Assessment in Italy;Luxemb OA. Publ. Eur. Communities EUR 19022(EN):32.

Van der KnijA J M, Jones R J A and Montanarella L 2000 Soilerosion risk assessment in Europe; European Soil BureauResearch Report, EUR 19044 EN, 34p.

Vijith H, Suma M, Rekha V B, Shiju C and Rejith P G 2012An assessment of soil erosion probability and erosion rate ina tropical mountainous watershed using remote sensing andGIS; Arab. J. Geosci. 5(4) 797–805.

Wischmeier W H and Smith D D 1978 Predicting rainfallerosion losses, Agriculture Handbook No. 537; Agr. Hand-book 537 285–291.

Yasser M, Jahangir K and Mohmmad A 2013 Earth dam siteselection using the analytic hierarchy process (AHP): A casestudy in the west of Iran; Arab. J. Geosci. 6(9) 3417–3426.

Zhou P, Luukkanen O, Tokola T and Nieminen J 2008 EAectof vegetation cover on soil erosion in a mountainouswatershed; Catena 75(3) 319–325.

Corresponding editor: ARKOPROVO BISWAS


an integrated approach of gis, rusle and ahp to model soil

Documents