Remote Sensing of Environment 199 (2017) 93–106

Contents lists available at ScienceDirect

Remote Sensing of Environment

j ourna l homepage: www.e lsev ie r .com/ locate / rse

Estimation of monthly bulk nitrate deposition in China based on satelliteNO2 measurement by the Ozone Monitoring Instrument

Lei Liu a, Xiuying Zhang a,⁎, Wen Xu b, Xuejun Liu b, Xuehe Lu a, Dongmei Chen c, Xiaomin Zhang a,Shanqian Wang a,d, Wuting Zhang a

a Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, International Institute of Earth System Science, Nanjing University, Nanjing 210023, Chinab College of Resources and Environmental Sciences, Centre for Resources, Environment and Food Security, Key Lab of Plant-Soil Interactions of MOE, China Agricultural University, Beijing 100193,Chinac Department of Geography and Planning, Queen's University, Kingston, ON K7L 3N6, Canadad Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application, Nanjing 210023, China

⁎ Corresponding author.E-mail address: lzhxy77@163.com (X. Zhang).

http://dx.doi.org/10.1016/j.rse.2017.07.0050034-4257/© 2017 Elsevier Inc. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 26 March 2016Received in revised form 25 June 2017Accepted 8 July 2017Available online xxxx

Remote sensing technology has great potential to expand the observation of ground-level nitrogen depositionfrom local monitoring sites to a regional scale, with high spatial and temporal resolutions. A new methodologyis developed to estimate the spatial distribution of monthly bulk nitrogen deposition on a regional scale, basedon precipitation amounts and HNO3 and aerosol nitrate (NO3

−) columns, derived from OMI NO2 columns andthe relationship of NO2, HNO3 and NO3

− from MOZART. The accuracy assessment shows that the proposedmodel has achieved a reasonably high predictive power for monthly NO3

−-N deposition (slope= 0.96, intercept= 0.35, R = 0.83, RMSE = 0.72) across China. The spatial NO3

−-N deposition shows a significant gradient fromindustrial areas to undeveloped regions, ranging from 0.01 to 26.76 kg N ha−1 y−1 with an average of5.77 kg N ha−1 y−1 over China during 2010–2012. The bulk NO3

−-N deposition shows a clear seasonal variation,with high depositions occurring in thewarm season (March–November) and peaking in July andAugust, and lowbulk NO3

−-N deposition appearing in winter (December–February). This study proves for the first time that theatmospheric boundary layer (ABL) HNO3 and NO3

− columns and precipitation are powerful to predict the bulknitrate deposition.

© 2017 Elsevier Inc. All rights reserved.

Keywords:Bulk depositionNitrate depositionSpatial patternNitrogen

1. Introduction

Nitrogen (N) deposition (Ndep) is a worldwide problem (Sutton andBleeker, 2013; Van Damme et al., 2015), leading to N saturation(Phoenix et al., 2012), eutrophication (BERGSTRÖM and Jansson,2006; Horswill et al., 2008), loss of biomass productivity and vegetationcover (Bassin et al., 2007; Reich, 2009) and soil acidification (Hoegberget al., 2006; Stevens et al., 2009). China has been one of the hotspots ofthe N deposition (Jia et al., 2014; Liu et al., 2013; Pan et al., 2012),resulting from human activities such as industrial development, fertiliz-er application, biomass burning and agriculture expansion (Gu et al.,2012; Lee et al., 2011b). Therefore, it is essential to estimate Ndep onthe land surface and establish control measurements for improving en-vironmental quality.

Many studies concerned bulk nitrate deposition in precipitation(containing wet and dry depositions through particles and/or gas)based on ground measurements. The bulk nitrate deposition overChina was estimated in our previous work (Liu et al., 2016), based on

the data from papers published during 2003–2014. However, large un-certainties still existed due to the errors ofmeasurements fromdifferentstudies and the large time span of themeasurements. ChinaAgriculturalUniversity has organized a Nationwide Nitrogen DepositionMonitoringNetwork (NNDMN) since 2004, consisting of 43monitoring sites acrossChina (Liu et al., 2011; Xu et al., 2015). They reported the ground bulknitrate depositions based on 43 NNDMN observation sites throughoutChina, providing a great potential for understanding the temporalchange of the ground bulk nitrate depositions on a national scale (Xuet al., 2015). However, it is not easy to obtain the spatial distributionof the nitrate depositions across China only based on the ground mea-surements, due to the limited number of the monitoring sites andtheir ununiform distribution. This study tries to expand the groundbulk nitrate depositions at the monitoring sites in NNDMN to a nationalscale.

The spatial variation of Ndep is influenced by energy consumption(E), N fertilizer use (FN) and precipitation amount (P). Based on previ-ous studies on estimating Ndep on a regional scale, the Ndep has linearor logarithmical relationships with P, FN, and E (Jia et al., 2014; Liu etal., 2016; Zhan et al., 2015; Zhu et al., 2015), confirming that these fac-tors could be used to estimate the spatial variations of Ndep on a regional

94 L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

scale. The studies mentioned above were based on statistical data (FNand E) from the China Statistical Yearbook on a provincial scale, whichmay not reflect the spatial variations of N emissions due to the influenceof atmospheric transmission and complex air flows. Thus, it is very dif-ficult to estimate Ndep with high spatial or temporal resolutions usingsuch coarse datasets.

The remote sensing technology provides a new means to indicatethe temporal and spatial variations of NO2 concentrations in the atmo-sphere due to its advantages of high spatial and temporal resolutions.The NO2 columns retrieved from Scanning Imaging Absorption spec-trometer for Atmospheric Chartography (SCIAMACHY), Global OzoneMonitoring Experiment (GOME) and GOME-2 and Ozone MonitoringInstrument (OMI) indicate the spatial and temporal variations of NO2

in the atmosphere (Ma et al., 2013). In recent years, the satellite datahave been widely used to calculate pollutant concentrations due to itshigh spatial and temporal resolutions. The tropospheric NO2 verticalcolumn densities (VCDs) have been employed to study the NO2 issueson regional scales, including emission trends and variations in pollutedhotspots (Hayn et al., 2009; Lin, 2012). Satellite NO2 columns were alsoused to assess the temporal change of the air quality (Jiang et al., 2006),the effects of emission control measures (Wang et al., 2007a), and dailyambient NO2 concentration predictions (Lee and Koutrakis, 2014).Moreover, some studies used the NO2 columns to estimate dry gaseousNO2 deposition on a regional or global scale (Cheng et al., 2013; Lu et al.,2013; Nowlan et al., 2014). However, to our knowledge, no study has re-ported on estimating bulk nitrate deposition using remotely sensedNO2

columns, since there is still difficulty in estimating wet/bulk Ndep usingremotely sensed NO2 columns.

The difficulty in using NO2 columns to estimate bulk nitrate deposi-tions is mainly due to that the contribution of NO2 to dissolved NO3

−-Nin precipitation is very limited as NO2 is not very directly soluble, butin indirect way. The fundamental theory is that NO2 reacts with O3 toform NO3

− and then highly soluble N2O5 (NO2 + NO3 → N2O5, N2O5

+ H2O→ 2 HNO3) (Barrie, 1985). Thus, most of the bulk NO3−-N in pre-

cipitation originates from HNO3 and aerosol nitrate (NO3−), which can

also tracked in similar related studies (Ervens, 2015; Sutton et al.,2011). However, no convincing high spatiotemporal HNO3 and NO3

−

columns can be retrieved based on known satellite instruments. Thefirst global distribution of satellite HNO3 column was gained from IASI(Infrared Atmospheric Sounding Interferometer) at a resolution of 1°latitude × 1° longitude (Wespes et al., 2009). This coarse resolution in-dicates that even a province in China has few pixels, which made it dif-ficult to use the data for estimation of bulk nitrate deposition in China.Currently, the IASI HNO3 product are not available to the public andare still under validation (Ronsmans et al., 2016). The same issue existsfor satellite NO3

− due to data availability and coarse resolution. Facingthis problem, our current work tried to use OMI NO2 columns and therelationship of NO2, HNO3 and NO3

− columns from an atmosphericchemistry transport model (CTM, MOZART-4) that has been adoptedwidely by atmospheric sciences (Brasseur et al., 1998; Emmons et al.,2010), to retrieve the HNO3 and NO3

− columns. Then, a statisticalmodel is established based on measured bulk nitrate depositions, theHNO3 and NO3

− columns, and the precipitation amount. Finally, themonthly bulk nitrate deposition over China is estimated based on theconstructed model.

2. Data

2.1. Monthly ground-level NO3−-N deposition

The monthly ground measurements of NO3−-N deposition from the

Chinese Nationwide Nitrogen Deposition Monitoring Network(NNDMN) from 2010 to 2012 are used to estimate the bulk NO3

−-N de-position. During this study period (2010−2012), 33 sites in NNDMN(Fig. 1) have complete monitoring data, and the ground measureddata from these 33 sites are used to construct the model. The detailed

information on the description of themonitoring sites and the observa-tion timespan of the sites are shown in the Supplementary Table S2 inXuet al. (2015). The detailed information onmeasuring bulkNO3

−-Nde-position can be found in Xu et al. (2015) and Liu et al. (2011).

The monthly ground-based NO3−-N depositions at 7 sites (Fig. 1)

from the Acid Deposition Monitoring Network in East Asia (EANET)were also used to examine the predictive performance of the construct-ed model in this study, which are considered as the independent mon-itoring network for model validation. Compared with the NNDMN, theEANET sites are mainly located in less polluted area, and the bulk/wetnitrate depositions are reasonably lower than those in NNDMN (Xu etal., 2015). The details of the monitoring techniques including sampling,chemical analysis, definitions of data completeness, quality control andquality assurance, were documented in the EANET monitoring manual(http://www.eanet.asia/product/index.html).

2.2. Tropospheric vertical column density (VCD) of NO2

The tropospheric vertical column density (VCD) of NO2 is retrievedfrom the Ozone Monitoring Instrument (OMI), boarded on NASA'sEOS Aura satellite. The DOMINO (Dutch Finnish Ozone Monitoring In-strument) product (v2.0) is used in this study (http://www.temis.nl).The algorithmon retrievingNO2 columns is described in themanuscript(Boersma et al., 2007). The recently updated version can be trackedfrom the DOMINO Product Specification Document. We use individualpixel clear sky (cloud radiance fraction b0.5) Level 2 OMI data. TheOMI NO2 columns have been well validated, with a bias of +20%/−30% accounting for different regions (Boersma et al., 2008; Irie et al.,2008; Irie et al., 2009), and have also been widely applied in environ-mental-related studies and for the support of emission control policy(Russell et al., 2012; Zhao and Wang, 2009). We found some grid cellswith a very small percentage (b0.003%) had outlier values (“NULL” or“tropospheric NO2 VCD b0 molecules cm−2”), which should be re-moved before the usage for estimating bulk deposition. To generatecontinuous maps, we performed the Kriging interpolation(Geostatistical tool in ArcGIS) to estimate the values at these cellswith outlier values. The cross validation of Kriging interpolation(0.125° latitude × 0.125° longitude) showed that the average relativeerror by month ranged from −2.97% to 4.17%. Notably, the resolutionof interpolationmaps is consistent with the original resolution of DOM-INO NO2 (0.125° latitude × 0.125° longitude), indicating that this inter-polation procedure mainly produces uncertainty at the grid cells (withoutlier values).

2.3. MOZART-4 output of the NO2, NO3− and HNO3 data

In this study the gaseous NO2, NO3− and HNO3 data outputted from

the Model for Ozone and Related chemical Tracers, version 4 (MO-ZART-4) during 2010–2012 were used. TheMOZART-4 has been widelyused as a global chemical transport model, which can be run at any res-olution, relying on the input meteorological fields and the computermemory limitations (Emmons et al., 2010). The standard MOZART-4mechanism contains 39 photolysis, 12 bulk aerosol compounds, 85gas-phase species and 157 gas-phase reactions. The MOZART-4 outputof the NO2, NO3

− and HNO3 data was temporally varying (6 h) with 1.9latitude × 2.5 longitude horizontal resolution and 56 vertical levelsfrom the ground. Detailed descriptions on the MOZART simulations ofNO2, HNO3 and NO3

− data can be found in previous studies (Emmonset al., 2010; Emmons et al., 2012; Tie et al., 2007).

2.4. Atmospheric boundary layer (ABL) height

The monthly average of atmospheric boundary layer (ABL) height,obtained from ERA-Interim (http://www.ecmwf.int/services/archive),is used in this study. The ERA-Interim Archive is a part of ECMWF's (Eu-ropean Centre for Medium-Range Weather Forecasts) Meteorological

Fig. 1. Spatial locations of groundmeasurements of wet/bulk nitrate depositions during 2010–2012. NNDMNwas used formodel construction and cross validationwhile EANETwas usedfor independentmodel validation. The GDP (gross domestic product) datawere obtained from China National Earth System Science Data Sharing Infrastructure (http://www.geodata.cn).

95L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

Archive and Retrieval System (MARS). A comprehensive documenta-tion of the ERA-Interim reanalysis system has been published as anopen-access article in the Quarterly Journal of the Royal MeteorologicalSociety (Dee et al., 2011).

2.5. Monthly precipitation

Themonthly precipitation data from2010 to 2012 throughout Chinais supplied by Centre for Atmosphere Watch and Services (CAWAS),China Meteorological Administration (CMA). The original gridded pre-cipitation data from CAWAS were used directly in the present work(http://data.cma.cn/), which are considered to be the most accuratedataset in China among similar products such as the National Centersfor Environmental Prediction (NCEP) and European Centre for Medi-um-Range Weather Forecasts (ECMWF). The CAWAS precipitationdata were produced by two dimensional thin plate smoothing splineswith the degree of data smoothing determined by minimizing the gen-eralized cross validation (Hutchinson, 1998; Wan et al., 2011; Yan,2004). The detailed discussion on the data set and processing can befound at China Meteorological Data Sharing Service System (http://data.cma.cn/).

2.6. Land use map

In this study, 30m digital land-use data is obtained from the LandsatThematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+)data in 2010. The land cover types used in this study refer to theTsinghua University 30 m resolution global land cover product of 2010(http://data.ess.tsinghua.edu.cn). Those land-cover classes are groupedfurther into seven aggregated classes of land cover in this study: waterarea, grassland, cropland, forestry land, built-up land, permanent wet-land and barren valeted land. The detailed description and accuracycontrol and assessment of this data can be found in the manuscripts(Gong et al., 2013; Yu et al., 2013).

3. Methodology

3.1. N compounds in the atmosphere

There are many N species including NO2, NO, N2O N2O5, HNO3,HONO and NO3

− in the atmosphere. The contribution of NO2 to dissolvedN inprecipitation is small as this gas is not very soluble relative toHNO3 orNO3

− (Mentel et al., 1996). We firstly derived the HNO3 or NO3− columns

based onOMINO2 columns and their relationship fromMOZART.Notably,we used the OMI NO2 columns with local time between 13:00 and 14:00whereas theMOZART output NO2 data used in the current work are tem-porally varying 6 h every day (00, 06, 12 and 18). We have two steps toderive the HNO3 or NO3

− columns. First, we established the relationshipbetween monthly NO2 columns (at 12:00) and the average HNO3

columns {[HNO3(00)+HNO3(6)+HNO3(12)+HNO3(18)]/4} as well asthe average NO3

− columns {[NO3(00)+NO3(6)+NO3(12)+NO3(18)]/4} from MOZART. NO2 columns are significantly (p b 0.01) correlatedwith both HNO3 and NO3

− columns from MOZART, as shown in Fig. 2and Fig. S2. Second,we derived themonthlymeanHNO3 or NO3

− columnsbased on the OMI NO2 columns and the relationship established in thefirst step.

3.2. Atmospheric boundary layer (ABL) HNO3 and NO3− columns

To calculate bulk nitrate deposition in the form of precipitation, it ismore reasonable to use the HNO3 and NO3

− columns below the precipi-tation height instead of the tropospheric HNO3 and NO3

− columns (Fig.3a) because the scavenging effect on N compounds is from the top pre-cipitation height rather than the top troposphere height (Racette et al.,2009), which is much higher than that of the production of cloudwater (Garratt, 1994). However, it is not feasible tomodel the precipita-tion height because the precipitation height depends on the status ofthe cloud which is extremely ephemeral and variable in both spaceand time (Bruintjes et al., 2010). Alternatively, we use ABL height as itis well modeled and validated (Dee et al., 2011). There are two majorreasons supporting this alternative. First, the precipitation height

Fig. 2. The relationship of NO2 columns (1013 molec. cm−2) at 12 h and HNO3 columns (1013 molec. cm−2) averaged by 00, 06, 12 and 18 h in China calculated from MOZART.

96 L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

(commonly below 2–3 km) is higher than but close to ABL height (com-monly below 2 km, Fig. 3b) (relative to the top of the troposphere, com-monly 10–20 km). Second, more importantly, the HNO3 and NO3

−

columns, are mainly within ABL from the profile shapes (Rozanov etal., 2005; Russell et al., 2011), indicating there is no large difference be-tween the HNO3 and NO3

− columns under precipitation height andunder ABL height. Thus, the N compounds within the ABL contributeto thedominated percentage of bulk nitrate depositions byprecipitationscavenging effects, and the tropospheric HNO3 and NO3

− columns wereconverted to ABL HNO3 and NO3

− columns to calculate the nitratedeposition.

Atmospheric chemistry transport model (CTM) can produce profilesfor aerosol (Liu et al., 2004; VanDonkelaar et al., 2010; VanDonkelaar etal., 2006), NO2 (Hudman et al., 2007; Lamsal et al., 2008; Nowlan et al.,

Fig. 3. Chemical processing inwet nitrate deposition (left) and atmospheric boundary layer struindicates the atmospheric boundary layer (ABL). Z is height and h is the boundary layer depth

2014) and SO2 (Lee et al., 2011a; Nowlan et al., 2014). We used the re-lationship between the tropospheric HNO3/NO3

− columns and ABLHNO3/NO3

− columns in MOZART-4 to convert the satellite-derived tro-pospheric HNO3/NO3

− columns to ABL HNO3/NO3− columns.

The MOZART-4 output of HNO3 and NO3− includes 56 vertical levels

from the ground and the vertical HNO3 and NO3− profiles fromMOZART

are simulated in every grid cell by Gauss function:

f xð Þ ¼ ∑ni¼1aie

− x−bið Þ2c2i ð1Þ

where n= 2, 3, 4, 5 or 6; x represents the height from the surface; f(x)represents the HNO3 or NO3

− concentration at a height of x (theHNO3 orNO3

− profiles). ai, bi and ci represent the constants of Gauss function. As

cture (right) (Garratt, 1994). (a) indicates the troposphere that the satellite indicated; (b).

97L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

measures of accuracy and precision of the five models (n= 2, 3, 4, 5 or6), we use the coefficient of determination (R2) and root mean squareerror (RMSE) to determine the best simulation function.

The tropospheric HNO3 or NO3− columns fromMOZART are calculat-

ed according to the vertical profiles as:

F htrop� � ¼

Z htrop

0f xð Þdx ð2Þ

where F(htrop) is MOZART tropospheric HNO3 or NO3− columns; htrop

represents the height of troposphere; f(x) represents the HNO3 orNO3

− profiles.The satellite-derived ABL HNO3 or NO3

− columns are calculated as:

SABL ¼ Strop � F hABLð ÞF htrop� � ð3Þ

where Strop represents the tropospheric HNO3 or NO3− columns (de-

scribed in Section 3.1); F(hABL) is the ABL HNO3 or NO3− columns from

MOZART; hABL represents the height of atmospheric boundary layer.

3.3. Statistical model of estimating bulk nitrate depositions

The statistical model tries to estimate bulk nitrate deposition fromABL HNO3 or NO3 columns and precipitation on a monthly scale ateach site, which is summarized by the following equations:

Nmn ¼ αfix þ βfix Smn− εn þ εfix� �� � ð4Þ

where

Smn ¼ Pmn � SABLð Þmn

SABLð Þmn ¼ γ � HNO3ABLð Þmn þ δ� NO3ABLð Þmn

εfix ¼ α−β Smnð Þ

Nmn is the measured NO3−-N deposition in month m, at site n;

(SABL)mn is an indicator of combining effect of HNO3 and NO3;(HNO3ABL)mn and (NO3ABL)mn are the ABL HNO3 and NO3

− columns inmonthm, at site n, respectively;γ and δ are the constants; Pmn is thepre-cipitation in monthm, at site n; Smn is an indicator of a combining effectof precipitation and N compounds in the atmosphere; βfix and αfix arethe slope and intercept; εn indicates random error or site bias for siten; εfix indicates the fixed error derived from the long term measure-ments of Smn at sites, indicating on the average effect of Smn on theNO3

−-Nmeasurements; In addition, the site bias represents themonthlyrandom error in the Smn-NO3

−-N relationship because the NO3−-N mea-

sured at a site may not be representative of a Smn value in a 0.125° lati-tude × 0.125° longitude grid cell. In brief, the bias may result from theirland use, spatial locations ormeteorological condition (Lee et al., 2011b;Yap and Hashim, 2013).

If all the independent variables including (HNO3ABL)mn, (NO3ABL)mn

and Pmn and the dependent variable Nmn are taken into Eq. (4), the con-stants include γ, δ, α, β as well as site bias εn. All the constants are fixedand can be easily gained except the εn, which is variable geographically.We use the controlling covariant (Smn) combining the ground εn to gen-erate the continuous εn map by the Co-Krigingmethod, where the rela-tionship between εn and Smn is statistically significant with slope=−3× 10−7 and intercept = 2 × 10−7 (p b 0.001). Once the εn map is gen-erated, we can gain the continuous monthly bulk nitrate depositionsusing the established model.

3.4. Model validation

The statistical model of estimating bulk nitrate deposition was vali-dated by a widely used cross validation (CV) technique, known as

Monte Carlo cross-validation, in evaluating the statistical model perfor-mance (Dubitzky et al., 2007; Lee et al., 2011b; Lee and Koutrakis, 2014;Varma and Simon, 2006). This technique randomly splits the datasetinto validation and training data. The model is established using thetraining data and the accuracy of the model performance is evaluatedusing the validation data (Varma and Simon, 2006).

In k-fold cross-validation, the original dataset is randomly dividedinto k equal sized subgroups. Of the k subgroups, one subgroup is treat-ed as the validation data for evaluating the performance of the statisticalmodel, and the remaining k-1 subgroups are used as training data. Inthis study, the 5-fold cross validation technique is used to validate thestatistical model of estimating nitrate deposition. The observation dataare divided into 5 randomly subgroups, among which 4 subgroups(training set) are used to construct a model, and 1 subgroup (test set)is used to evaluate the estimation performance. The process of 5-foldcross validation is then repeated 5 times, with each of the 5 subgroupsused exactly once. The CVmethod over repeated random subgroups en-sures that all observation data are used for both validation and training,and every observation data is exactly once used for model validation(Varma and Simon, 2006). To evaluate the relationship between themeasured and predicted NO3

−-N, the Pearson correlation coefficientsand the coefficient of determination (R2) are used. Also, the root meansquare error (RMSE) is applied to evaluate the accuracy. The validationis very important to assess the spatial accuracy of the predicted bulkNO3

−-N depositions.

3.5. Clarifications on methodology thoughts

Given the aforementioned explanations, figures, associated conceptsand method descriptions, the implementation seems not easy. We aimto generate accurate datasets of bulk nitrate depositions over China,and hence will provide basic input parameters for process-basedmodels of N biogeochemistry in agricultural or other ecosystems to de-termine or quantify their effects on ecosystemhealth, biological diversi-ty and greenhouse gas balances. Thus, the national groundmeasurements of bulk nitrate depositions (NNDMN) were used formodel constructions to make this estimation less biased with the realstatus of bulk nitrate depositions. Due to limited spatial representativesof ground measurements, satellite observations with high spatial andtemporal resolutions, and an atmospheric chemistry transport model(from MOZART in this study) with detailed vertical profiles for N com-pounds (HNO3 or NO3

−), were combined to expand the ground mea-surements to a national scale. The thought of this study can alsotracked in some studies in estimating particular matter (PM) (Kloog etal., 2015; Yap and Hashim, 2013) or some pollution gases (Lee andKoutrakis, 2014). In such methods, three steps are generally included:(1) as many as ground measurements of target objects should be pre-pared for both model construction and validation; (2) the most relatedfactors influencing the target objects should be selected, and the mostimportant point is that selected factors should be equippedwith charac-ters of high spatiotemporal resolutions; (3) these ground measure-ments of target objects and selected factors are then used for modelconstruction and retrieval of high spatiotemporal target objects on a re-gional scale.

On the analogy of this, the methodology also includes three similarsteps. For the first step, this study selected the national ground mea-surements of bulk nitrate depositions (NNDMN) over China for modelconstruction and validation. Second, N compounds (HNO3 and NO3

−)as well as precipitation were selected as the most important factorsinfluencing the nitrate depositions. By this step, HNO3 and NO3

− col-umns were generated by OMI NO2 columns and their relationshipwith HNO3 and NO3

− from MOZART. Third, a statistical model was con-structed for estimating nitrate depositions combining generated HNO3

and NO3− columns and precipitation. Key procedures of this study are

given in Fig. 4, and detailed procedures step by step have been describedin Section 2 in the Supplementary material.

Fig. 4. The basic steps in the current work: (a) main steps for generating continuous tropospheric NO2 VCDs; (b) procedures for generating the ABL HNO3 and NO3− columns and (c) key

procedures for calculating the NO3−-N deposition according to the statistical model constructed in Section 3.3.

98 L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

3.6. Sensitivity analysis by Monte Carlo

The analysis of the uncertainty degree due to the errors or uncer-tainty coming from the input data to drive themodel, which is in par-ticular associated with the regional or large scale modeling wherethe inputs are usually derived from simulations (such as the precip-itation data) or satellite observations (such as NO2 columns data inthe present work). The Monte Carlo simulations have the ability toquantify the sensitivity of these input factors, in which several possi-ble sceneries are considered for every input parameter by using therandom values within an acceptable specific range. Therefore, thecorresponding estimated dependent values can be generated, andthen the statistically possible distributions and range of the modelsimulations with the change of independent input variables can begained. Actually, the Most Significant Factor (MSF) simplified byMonte Carlo simulations (due to expensive computation) is widelyused in many large scale simulations (Giltrap et al., 2010; Kieseet al., 2005; Li et al., 1996), in which the extreme values of theinput parameters are used to producing the variations of the targetestimated object. In this study, we also used the MSF simulations toexplore the sensitive parameters on estimating the bulk nitrate de-positions in the constructed model.

We referred to the methods in a previous study (Li et al., 1996) byvarying a single input parameter and keeping other input parametersunchanging to explore the impact of the input parameters on the bulknitrate depositions. We set the single varying parameter by 10%, 20%and 30% (the extreme value in MSF) higher than the actual data todrive the model, and gained three additional estimated bulk nitrate de-positions to compare with the baseline (using the actual data to drivethe model). After repeating the same procedures one by one for allinput parameters, variations of the bulk nitrate depositions can be quan-tified, which will be used to determine which parameter is the mostsensitive factor.

4. Results and discussions

4.1. NO3−-N prediction and accuracy assessment

HNO3 andNO3− columnswerefirstly derived fromOMINO2 columns

based on their relationship from MOZART (Fig. 2 and S2). Then, tropo-spheric HNO3 orNO3

− columnswere converted toABLHNO3 orNO3− col-

umns based on their vertical profiles fromMOZART. The vertical profileswerewell simulated by Gauss function in each grid cell for HNO3 (Fig. 5)or NO3

− (Fig. S3).The spatial pattern of the random errors (εn) across China was gen-

erated from 2010 to 2012 in the mixed effects models (Fig. 6). The εnmapwas generated usingCo-Krigingmethod constrained by Smn as stat-ed in Section 3.3. We used the mean predictive error (MPE) and rootmean square standardized error (RMSSE) to assess the interpolationperformance. The MPE indicates the degree of predictive bias and thepredictive values are unbiased if this value is equal to 0, while theRMSSE close to 1 represents accurate error distribution (Jia et al.,2014). The MPE and RMSSE of εn were 0.03 and 0.98, respectively(Table S2), showing the interpolation performed well and the εn mapwas normally distributed. The developed areas have a high negativesite bias and the southeastern and northwestern China have a high pos-itive site bias, showing that densely developed areas have anunderestimated NO3

−-N and the developing areas have anoverestimated NO3

−-N. Thus, it is essential to add the site bias into themixed effect model to do the adjustment. The spatial pattern of thesite bias indicates the influence of the land uses over China. Land usepartly explains the NO3

−-N variability because there are different emis-sion sources in different developed areas, suggesting the simulatedNO3

−-N deposition is sensitive to emission estimates. The negative re-gions, corresponding to the cropland or Natural Vegetation Mosaicland, denote underestimated values from Smn, so that enhancement isneeded. This is mainly due to the fact that these regions are being

Fig. 5. R2 and RMSE (molec. cm−2) simulated by Gauss function (68–142°E, 5–55°N) for HNO3 vertical profiles from MOZART in each grid cell in 2012.

99L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

surrounded by intensive anthropogenic activities and rapid develop-ments along with urbanization, which shorten the transport distanceof NOx emission from fossil fuel combustion (Xu et al., 2015). Similarly,the positive regions, corresponding to Grasslands, Forests or Barren orSparsely Vegetated land, indicate an overestimation in values from Smn

that need to be trimmed down.The monthly bulk NO3

−-N depositions were generated over Chinaduring the period 2010–2012 based on the constructed statisticalmodel. Based on the NNDMNmeasurements, the 5-fold cross validation

results are shown in Fig. 7, explaining the variability in monthly NO3−-N

depositions (slope=0.96, intercept=0.35, R=0.83, RMSE=0.72) fora period of three years (i.e. 2010 to 2012) in China. Moreover, 7 site inEANET representing the independent measurements were also com-pared with the estimation (slope = 0.98, intercept = 0.02, R = 0.85,RMSE = 0.17) during 2010–2012 in China (Fig. 8). This indicated thatthe ABL HNO3, NO3

− columns and precipitation amounts are powerfulto predict the bulk nitrate depositions, and the constructed modelcould be used to estimate the bulk nitrate deposition on a regional scale.

Fig. 6. Spatial distribution of randomerror (a) and landuse (b) inChina. Note: The landuse (b) could partly explain the randomerror (a): positive values in (a) corresponding to the Barrenand Sparsely Vegetated and Grasslands; the negative values in (a) corresponding to the croplands and forests.

100 L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

4.2. Spatial patterns of bulk NO3−-N deposition

This study expanded the ground-level NNDMN measurements to anational scale and obtained a reliable estimation of bulk NO3

−-N deposi-tion for three years, with the aid of the HNO3, NO3

− columns and precip-itation. To our knowledge, this is the first time to obtain the bulk NO3

−-Ndeposition with such high spatial and temporal resolutions throughoutChina.

Themonthly bulk NO3−-N deposition from January to December dur-

ing 2010–2012 is shown in Fig. 9, and the annual bulk NO3−-N deposi-

tion over China during 2010–2012 was also demonstrated in Fig. 10.The bulk NO3

−-N deposition map showed an obvious gradient fromthe developed regions to undeveloped regions in China, ranging from0.01 to 26.76 kgN ha−1 y−1 with an average of 5.77 kg N ha−1 y−1 dur-ing 2010–2012 (Fig. 10a). Based on the measurements in NNDMN, in

Fig. 7.Accuracy assessment of themonthly estimated bulk nitrate depositions (kg N ha−1 m−1)in Section 2 in the Supplementary material.

spite of some dry deposition (e.g. dust and/or particulates) included,thewet (-only) deposition is themajor part (78–90%) of bulk depositionaccording to a previous study (Zhang et al., 2008).

From the monthly spatial maps (Fig. 9), we clearly found the bulkNO3

−-N deposition by month (month−1) exceeding 1.5 kg N ha−1 m−1

mostly occurred in Northern China (in July and August) including Bei-jing, Shandong, Shaanxi, Henan and Liaoning provinces, as well asSouthern China (especially in December, January, February andMarch) including the Jiangsu, Anhui, Zhejiang, Jiangxi, Hunan, Sichuanand Guangdong provinces. From the annual spatial map (Fig. 10), highbulk nitrate deposition occurred in three main regions: (1) NorthChina, including Hebei, Liaoning, Shandong, Tianjin, Henan and Beijingprovinces; (2) Yangtze delta region, including Jiangsu, Anhui, Zhejiang,Shanghai, Jiangxi, Hunan and Hubei provinces; (3) Sichuan Basin, in-cluding Sichuan and Chongqing provinces. It is not surprising that

in NNDMNusing 5-fold cross-validation. For detailed steps for this validation can be found

Fig. 8. Comparison between the measured and estimated NO3−-N using EANET,

representing the independent monthly ground bulk nitrate depositions. The monitoringsites in EANET are distributed in Fig. 1 where most of sites were located in less pollutedarea.

101L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

North China Plain, Yangtze delta region and Sichuan Basin had high bulknitrate deposition due to high NOx emissions, which can be clearly seenfrom the HNO3 map (Fig. 10c). High HNO3 (as a highly soluble N com-pound) columns, occur in these regions, and hence largely be scavengedby precipitation. In some regions in South China, such as Hunan andJiangxi provinces, small areas of high HNO3 columns were found,while large areas of bulk nitrate depositions were found in these re-gions. This is mainly because the precipitation amounts in these regionswere much higher than those in other regions (Fig. 10b, annual precip-itation is 1.4–6.5 times than that in other geographical areas).

Fig. 9. Spatial distribution of the monthly bulk NO3−-N depos

The Western China, including Tibetan plateau (TP) and NorthwestChina (NW), had low bulk nitrate depositions compared with theother areas of China, with the bulk NO3

−-N deposition below0.5 kg N ha−1 m−1 even for all months (Fig. 9). This comes from tworeasons. First, NW and TP were much less populated (Fig. 10d) and in-dustrialized, and natural emissions were supposed to dominate the Ncompounds (e.g., soil emissions and lighting),while, in central and east-ern coastal regions, HNO3 columns (Fig. 10c) are mainly distributedwith dense population and high industrialization level. Second, moreimportantly, the precipitation amounts in Western China (annual pre-cipitation was about 200–300 mm) were much lower than that inother regions, which were about 1/2–1/6 times than that in other geo-graphical regions.

4.3. Seasonal variations of the bulk NO3−-N depositions

The monthly average bulk NO3−-N deposition from 2010 to 2012 is

shown in Fig. 11a,which is consistentwith the temporal changes of pre-cipitation amounts over China (Fig. 11b). Higher bulkNO3

−-N depositionoccurred in summer, when China received on average 330.51 mm ofrain during the summer, accounting for 51.29% of total annual precipita-tion. The minimum bulk NO3

−-N deposition occurred in winter, whichwas attributable to the low precipitation (b50.00 mm, about 1/7 ofthe precipitation amounts in summer). This indicates that the precipita-tion was an important controlling effect on the temporal change of bulkNO3

−-N deposition. The consistent effects of precipitation on the bulk ni-trate deposition have also been reported by other studies (Pan et al.,2012; Xu et al., 2015).

The seasonal trend of bulk nitrate deposition can also be influencedby the N compounds in the atmosphere, which can be seen from thetemporal change of ABL HNO3 columns (Fig. 11c), as an indicator ofhighly soluble N compounds (N2O5, HNO3, or NO3

−) in the atmosphere.The scatter plots of precipitation versus the ABL HNO3 columns are

ition (kg N ha−1 m−1) from 2010 to 2012 across China.

Fig. 10. Spatial distribution of the annual bulk NO3−-Ndeposition (kgNha−1 y−1) from2010 to 2012 across China (a). (b) denotes the annual precipitation from2010 to 2012 across China;

(c) represents the annual HNO3 columns from 2010 to 2012 across China and (d) shows the distribution of population (people/km2) in 2010 in China.

102 L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

illustrated in Fig. 12. A log model was created to simulate the relation-ship between precipitation and N compounds in the atmosphere. Thisshowed that when the rainfall was low, the soluble N compounds(such as N2O5, HNO3, or NO3

−) in the atmosphere decreased with in-creasing precipitation, mainly because of the scavenging of pollutants.The scavenging of N compounds by precipitation is an important pro-cess influencing the seasonal change of bulk NO3

−-N depositions (Liuet al., 2015). Thus, the seasonal trends of bulk NO3

−-N deposition arecaused by a combination of precipitation and the ABL HNO3 columns,which can explain the seasonal variation of the bulk NO3

−-N depositionswith correlation coefficients of 0.91 (Fig. 12b).

It should be noted here that this analysis ignored the spatial varia-tions of the bulk NO3

−-N depositions, precipitation and ABL HNO3 col-umns. We used the average bulk NO3

−-N depositions, precipitationand ABL HNO3 columns throughout China by month to do the analysisof the seasonal variations. There is no doubt that the spatial variationsof the bulk NO3

−-N depositions result from combining effects of the pre-cipitation and the ABL HNO3 columns.

4.4. Different methods to estimate bulk nitrate depositions

Generally, there were three methods to quantify atmospheric bulknitrate/wet deposition on the national scale: atmospheric chemistrytransport model (CTM) simulations (Appel et al., 2011; Dentener etal., 2006; Ge et al., 2014; Zhang et al., 2012; Zhao et al., 2017),

geostatistical methods (such as Kriging interpolation) (Jia et al., 2014;Lü and Tian, 2007; Liu et al., 2016; Zhu et al., 2015) and averagemethods based on long-term monitoring networks (Liu et al., 2013;Pan et al., 2012; Xu et al., 2015). The average method is the simplestmethod, and the national nitrate deposition can be easily gained by av-eraging the observation data at the monitoring sites. There was nodoubt that the average method at the site scale in the long-term moni-toring networks such as the Chinese Nationwide Nitrogen DepositionMonitoring Network (NNDMN), Chinese Ecosystem Research Network(CERN), Acid Deposition Monitoring Network in East Asia (EANET), Eu-ropean Monitoring and Evaluation Programme, Europe (EMEP), CleanAir Status and Trends Network/the National Atmospheric DepositionProgram, United States (CASTNET/NADP) and Canadian Air and Precip-itationMonitoring Network, Canada (CAPMoN)were themost accurateand valuable datasets, which have been widely used to validate themodeling bulk/wet nitrate deposition by CTMs (Appel et al., 2011;Dentener et al., 2006; Ge et al., 2014; Zhang et al., 2012). At the nationalscale, the averagemethod used to estimate the bulk/wet nitrate deposi-tion by averaging the observation data at the monitoring sites possiblyshould be careful and may cause large uncertainties due to ignoringthe relatively unbalanced levels of economic development in differentregions and the inhomogeneous distribution of observation sites(Fig. 1).

Geostatistical methods (such as Kriging interpolation) can generaterelatively more accurate results than average method because of

Fig. 11.Monthly bulk NO3−-N deposition (kg N ha−1 m−1), precipitation (mm) and ABL HNO3 (1013 molec./cm2) for the time period 2010–2012 in China. The vertical bars represent the

error of monthly bulk NO3−-N deposition, precipitation and ABL NO2 for the three years of values. (a), (b) and (c) shows themonthly bulk nitrate deposition, precipitation and ABL HNO3

columns. J, F, M, A, M, J, J, A, S, O, N, D represent the months from January to December.

103L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

considering the surrounding area around observation sites as long aswecan derive the bulk/wet nitrate deposition data with a relatively suffi-cient number of ground sites which are distributed throughout China(Jia et al., 2014; Lü and Tian, 2007; Liu et al., 2016; Zhu et al., 2015).Jia et al. (2014) used the Kriging methods to obtain the spatial distribu-tions of yearly bulk nitrate depositions based on published papers be-tween 1980s and 2010s; Liu et al. (2016) summarized the existingspatial distributions of yearly bulk nitrate deposition since 2003, and ex-plored the potential influencing factors. However, the above studies areconstrained by different sampling procedures and standards of

Fig. 12. The relationship between precipitation (mm), ABLHNO3 (1013molec./cm2) and bulkNOthe relationship between precipitation (mm) and ABLHNO3 (1013molec./cm2). (b) shows the ca monthly scale.

measured bulk nitrate depositions resulting from different studies.Lack of standards for sampling procedures and concept of atmosphericbulk nitrate depositions leaded to incomparable results by differentstudies, and worsens the existed difficulties related to data integration.The geostatistical methods can hardly estimate the long-term nitratedeposition due to the limited continuous monitoring sites (43 sites inNNDMN, 41 sites in CERN) in China.

CTM methods can be used to estimate the bulk/wet nitrate deposi-tion, and the nitrate deposition maps on a national or global scalehave been generated (Appel et al., 2011; Brasseur et al., 1998; Lu and

3−-Ndeposition (kgNha−1m−1) on amonthly scale in China from2010 to 2012. (a) showsombination effect of precipitation andABL HNO3 columns on the bulk nitrate deposition on

Fig. 13. Sensitivity of bulk nitrate depositions (kg N ha−1 m−1) to NO2 columns (1013

molec. cm−2), precipitation (mm) and ABL height (m) in July 2012. The single varyingparameter was set 10%, 20% and 30% (the extreme value in MSF) higher than the actualdata to drive the model in a simulation.

104 L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

Tian, 2014). CTM can estimate the wet/bulk nitrate deposition based onthe physical and chemical mechanisms involved in the atmosphericprocess or transport and in cloud scavenging or rainout removal. Thescheme including in-cloud and below-cloud scavenging and the scav-enging in convective updrafts from large-scale precipitation has beendescribed in the previous studies (Amos et al., 2012; Liu et al., 2001;Mari et al., 2000). Wet/bulk nitrate depositions from CTMs have largespatial resolutions. Resolutions of different CTMs are varying from 1°latitude × 1° longitude to 5° latitude × 4° longitude, and further to beimproved to b1° latitude × 1° longitude in recent years (Appel et al.,2011; Dentener et al., 2006; Vet et al., 2014; Zhang et al., 2012).Lü and Tian (2007) found the CTM modeling results during the year of1993 (Dentener, 2006) significantly underestimated the real nitrate de-positionwith 71.3% ofmodeled valueswithin±50% of the groundmea-surements of nitrate deposition. Vet et al. (2014) also found thatmoderate underestimation (the modeling results accounting for 70%of the measured) in Asia, implying that the CTMs were not completelyeffective in deriving the spatiotemporal nitrate deposition throughoutthe country. The reason for the underestimation in Asia may be due tothe underestimation of emission,which has been suggested by previousstudies (Wang et al., 2012;Wang et al., 2007b; Zhang et al., 2007). In ad-dition, many chemical transport processes are simplified and parame-terized in CTMs (uncertainties in chemistry schemes and errors of theemission data), and consequently CTMs cannot provide all the informa-tion of reality (Du and Liu, 2014). The operational performance of atmo-spheric chemistry model in predicting nitrate deposition needs to befurther validated by comparison with field measurements. Further im-provement of CTMmethods in predicting the bulk/wet nitrate depositionneed to be corrected ormodified by asmany as possible observation dataof nitrate deposition. In recent years, the spatial resolutions of CTM in es-timating bulk/wet nitrate depositions has been improved such as 1° × 1°resolution in the globe (note that the emission resolution are larger than1° × 1°) by Dentener et al. (2006) and Vet et al. (2014), 1/2° × 2/3° reso-lution in the USA by Zhang et al. (2012) and 30 km by 30 km in China byGe et al. (2014). Ge et al. (2014) assessed the accuracy of an improvedCTM named as the Nested Air Quality Prediction Modeling System(NAQPMS), improved the spatial resolutions into 30 km by 30 km (notethat the emission data at 0.5° × 0.5° resolution and the precipitation at0.25° × 0.25°), and found the overall correlation (R) between the model-ing and observationwere about 0.74 across China (slightly lower than theaccuracy with R = 0.81 in this study).

We are supposed to clarify in particular that we do not aim at im-proving the scheme of CTM using the traditional way including in-cloud and below-cloud scavenging to estimate the bulk/wet nitrate de-position, but enhancing themethods proposed by the recent studies (Jiaet al., 2014; Zhan et al., 2015; Zhu et al., 2015) from the statistical per-spective. Jia et al. (2014) found that FN, E and P combined contributed79% of the spatial variation of Ndep. Zhan et al. (2015) concluded thatFN, E, and P jointly explained 84.3% of the spatial pattern of Ndep, inwhich FN (27.2%) was the most important, followed by E (24.8%) andP (9.3%). Zhu et al. (2015) also found P and FN could explain 80–91%of the spatial variation of Ndep. These studies confirmed that these fac-tors could be used to estimate the spatial variations of Ndep on a regionalscale, but were confined effectively to provincial, municipal or county-level scale due to the limited statistical data from the China StatisticalYearbook, which may not denote the real status of spatial distributionof N species resulting from the impact of the complicated air atmo-spheric transport. Facing this problem, we applied the satellite-derivedcolumns with high spatial and temporal resolutions to replace the sta-tistical data in estimating the bulk nitrate deposition over China, andtest whether the NO2 columns retrieved from remotely-sensed tech-niques could be used to estimate the bulk/wet NO3

−-N on a monthlyscale. The bulk nitrate deposition estimated from the proposed methodwere compared with the ground measurements during 2010–2012 inChina. The results showed this method is powerful to estimate the spa-tial distribution of bulk nitrate deposition with high-resolution on

national and regional scales in China.We consider that themethodologywill be promising in investigating and measuring the bulk/wet deposi-tion from a satellite perspective on a regional scale.

4.5. Model uncertainty

We conducted both 5-fold cross validation using NNDMNmeasure-ments and independent validation using the EANET measurements be-tween the estimated and measured bulk nitrate depositions, showingthe modeling simulations can still generate estimations with uncer-tainties. This may result from both the input dataset used for runningthe model and the constructed model itself. The estimation may pro-duce relatively enormous uncertainty with inaccurate input data, indi-cating it is critically to conduct a sensitive analysis on how thedifferent input data impact the estimation of bulk nitrate depositions.We presented here the sensitive analysis by using the MSF method tobetter understand the key influencing factors on the model perfor-mance for the regional simulations.

As described in Section 3 on themethodology, the key input param-eters are NO2 columns, precipitation and ABL height (NO2 columns andABL height are key inputs to derive the process variables of ABL HNO3

and NO3−). To examine the sensitivity of each parameters to the bulk ni-

trate depositions, we took amonth (July in 2012) as a case study by con-trolling a single parameter varying and other input parametersunchanging to drive the model in a single simulation. Based on theMSF method by setting the extreme value as 10%, 20% and 30% higherthan the actual data for each parameter, we found the precipitation isthe most sensitive parameter to the bulk nitrate depositions as shownin Fig. 13. A 30% increase in precipitation leaded to 19.61% increase inbulk nitrate deposition. Varying NO2 columns had moderate impact on

105L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

the bulk nitrate deposition. A 30% increase in NO2 columns produced a13.73% increase in bulk nitrate deposition. Compared with precipitationandNO2 columns, ABL had less effect on the bulk nitrate deposition, anda 30% increase in ABL height generated a 3.92% increase in bulk nitratedeposition.

In addition to the uncertainty coming from the input data, the con-structedmodel itself withmultiple assumptionsmay also arouse uncer-tainty in estimating bulk nitrate depositions. First, we used ABL heightinstead of the top precipitation height to calculate HNO3 and NO3

− col-umns scavenged by precipitation due to the difficulty in modeling thetop precipitation height. Second, the discrepancy of resolutions be-tweenOMI andMOZART aswell as the relationship ofMOZARTNO2 col-umns vs. HNO3 columns, and NO2 columns vs. NO3

− columns may alsolead to uncertainty in estimating the bulk nitrate depositions. Third,our constructed model was not only based on satellite NO2 columns,MOZART simulations but also measured NNDMN sites, indicating thatin general more continuous monitoring sites with uniform monitoringmethods should be enhanced in future.

5. Conclusion

In order to better estimate long-term monthly NO3−-N deposition,

we have developed and validated a model to estimate the monthlybulk NO3

−-N deposition with a 0.125° latitudes × 0.125° longitudesacross China. A good model performance has been shown across thewide range of geo-climatic characteristics of China. The 5-fold cross val-idation in NNDMN shows that the monthly estimated and measuredNO3

−-N deposition are generally correlated with a slope = 0.96 and R= 0.83, and the independent validation in EANET denotes the monthlyestimated and measured NO3

−-N deposition are also significantly corre-lated with a slope = 0.98 and R = 0.85.

The NO3−-N deposition map indicated a significant gradient from

east to west in China, ranging from 0.01 to 26.76 kg N ha−1 y−1 withan average of 5.77 kg N ha−1 y−1. The spatial pattern of NO3

−-N deposi-tion by province showed that Beijing, Hebei, Shandong, Jiangsu, Guang-dong and Zhejiang, were the largest contributors, while theundeveloped areas in the west had little contribution to nationalNO3

−-N deposition. Also, high bulk NO3−-N deposition occurred in sum-

mer, followed in order by spring, autumn and winter.The generated bulk nitrate deposition is able to reflect the status of

bulk nitrate deposition in China, and to provide basic information forassessing the influence of N enrichment on regional biogeochemical cy-cles, water resources and climate. The satellite-retrieved bulk nitratedeposition is particularly helpful in assessing the status of nitrate depo-sition without ground measurements, and the results support the de-velopment of a satellite-based N deposition monitoring plan in thefuture over China.

Acknowledgements

This study is supported by the National Natural Science Foundationof China (No. 41471343, 40425007 and 41101315). The authors wouldlike to thank the anonymous reviewers for their constructive commentsand suggestions.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.rse.2017.07.005.

References

Amos, H.M., Jacob, D.J., Holmes, C.D., Fisher, J.A., Wang, Q., Yantosca, R.M., Corbitt, E.S.,Galarneau, E., Rutter, A.P., Gustin, M.S., 2012. Gas-particle partitioning of atmosphericHg(II) and its effect on global mercury deposition. Atmospheric Chemistry & Physics11, 29441–29477.

Appel, K., Foley, K., Bash, J., Pinder, R., Dennis, R., Allen, D., Pickering, K., 2011. A multi-res-olution assessment of the Community Multiscale Air Quality (CMAQ) model v4.7 wetdeposition estimates for 2002–2006. Geosci. Model Dev. 4, 357–371.

Barrie, L., 1985. Scavenging ratios, wet deposition, and in-cloud oxidation: an applicationto the oxides of sulphur and nitrogen. Journal of Geophysical Research: Atmospheres90, 5789–5799.

Bassin, S., Volk, M., Suter, M., Buchmann, N., Fuhrer, J., 2007. Nitrogen deposition but notozone affects productivity and community composition of subalpine grassland after3 yr of treatment. New Phytol. 175, 523–534.

BERGSTRÖM, A.K., Jansson, M., 2006. Atmospheric nitrogen deposition has caused nitro-gen enrichment and eutrophication of lakes in the northern hemisphere. Glob.Chang. Biol. 12, 635–643.

Boersma, K., Eskes, H., Veefkind, J.P., Brinksma, E., Van Der, A.R., Sneep, M., Van Den Oord,G., Levelt, P., Stammes, P., Gleason, J., 2007. Near-real time retrieval of troposphericNO2 from OMI. Atmos. Chem. Phys. 7, 2103–2118.

Boersma, K., Jacob, D.J., Bucsela, E., Perring, A., Dirksen, R., Yantosca, R., Park, R., Wenig, M.,Bertram, T., Cohen, R., 2008. Validation of OMI tropospheric NO2 observations duringINTEX-B and application to constrain NOx emissions over the eastern United Statesand Mexico. Atmos. Environ. 42, 4480–4497.

Brasseur, G., Hauglustaine, D., Walters, S., Rasch, P., Müller, J.F., Granier, C., Tie, X., 1998.MOZART, a global chemical transport model for ozone and related chemical tracers:1. Model description. Journal of Geophysical Research: Atmospheres 103,28265–28289 (1984–2012).

Bruintjes, R.T., Clark, T.L., Hall, W.D., 2010. Interactions between topographic airflow andcloud/precipitation development during the passage of a winter storm in Arizona.J. Atmos. Sci. 51, 48–67.

Cheng, M., Jiang, H., Guo, Z., Zhang, X., Lu, X., 2013. Estimating NO2 dry deposition usingsatellite data in eastern China. Int. J. Remote Sens. 34, 2548–2565.

Dee, D., Uppala, S., Simmons, A., Berrisford, P., Poli, P., Kobayashi, S., Andrae, U.,Balmaseda, M., Balsamo, G., Bauer, P., 2011. The ERA-Interim reanalysis: configura-tion and performance of the data assimilation system. Q. J. R. Meteorol. Soc. 137,553–597.

Dentener, F.J., 2006. Global Maps of Atmospheric Nitrogen Deposition, 1860, 1993, and2050. Data set. Available on-line [http://daac.ornl.gov/] from Oak Ridge National Lab-oratory Distributed Active Archive Center, Oak Ridge, Tennessee, U.S.A. http://dx.doi.org/10.3334/ORNLDAAC/830.

Dentener, F., Drevet, J., Lamarque, J., Bey, I., Eickhout, B., Fiore, A., Hauglustaine, D.,Horowitz, L., Krol, M., Kulshrestha, U., 2006. Nitrogen and sulfur deposition on region-al and global scales: a multimodel evaluation. Glob. Biogeochem. Cycles 20.

Du, E., Liu, X., 2014. High rates of wet nitrogen deposition in China: a synthesis. NitrogenDeposition, Critical Loads and Biodiversity. Springer, pp. 49–56.

Dubitzky, W., Granzow, M., Berrar, D.P., 2007. Fundamentals of Data Mining in Genomicsand Proteomics. Springer Science & Business Media.

Emmons, L., Walters, S., Hess, P., Lamarque, J.-F., Pfister, G., Fillmore, D., Granier, C.,Guenther, A., Kinnison, D., Laepple, T., 2010. Description and evaluation of theModel for Ozone and Related chemical Tracers, version 4 (MOZART-4). Geosci.Model Dev. 3, 43–67.

Emmons, L.K., Hess, P.G., Lamarque, J.F., Pfister, G.G., 2012. Tagged ozone mechanism forMOZART-4, CAM-chem and other chemical transport models. Geoscientific ModelDevelopment Discussions 5, 1531–1542.

Ervens, B., 2015. Modeling the processing of aerosol and trace gases in clouds and fogs.Chemical Reviews 115, 4157–4198.

Garratt, J.R., 1994. The Atmospheric Boundary Layer. Cambridge University press.Ge, B.Z., Wang, Z.F., Xu, X.B., Wu, J.B., Yu, X.L., Li, J., 2014. Wet deposition of acidifying sub-

stances in different regions of China and the rest of East Asia: modeling with updatedNAQPMS. Environ. Pollut. 187, 10–21.

Giltrap, D.L., Li, C., Saggar, S., 2010. DNDC: a process-basedmodel of greenhouse gas fluxesfrom agricultural soils. Agric. Ecosyst. Environ. 136, 292–300.

Gong, P., Wang, J., Yu, L., Zhao, Y., Zhao, Y., Liang, L., Niu, Z., Huang, X., Fu, H., Liu, S., 2013.Finer resolution observation and monitoring of global land cover: first mapping re-sults with Landsat TM and ETM+ data. Int. J. Remote Sens. 34, 2607–2654.

Gu, B., Ge, Y., Ren, Y., Xu, B., Luo, W., Jiang, H., Gu, B., Chang, J., 2012. Atmospheric reactivenitrogen in China: sources, recent trends, and damage costs. Environmental Science &Technology 46, 9420–9427.

Hayn, M., Beirle, S., Hamprecht, F.A., Platt, U., Menze, B.H., Wagner, T., 2009. Analysingspatio-temporal patterns of the global NO2-distribution retrieved from GOME satelliteobservations using a generalized additive model. Atmos. Chem. Phys. 9, 6459–6477.

Hoegberg, P., Fan, H., Quist, M., Binkley, D., Tamm, C.O., 2006. Tree growth and soil acid-ification in response to 30 years of experimental nitrogen loading on boreal forest.Glob. Chang. Biol. 12, 489–499.

Horswill, P., O'Sullivan, O., Phoenix, G.K., Lee, J.A., Leake, J.R., 2008. Base cation depletion,eutrophication and acidification of species-rich grasslands in response to long-termsimulated nitrogen deposition. Environ. Pollut. 155, 336–349.

Hudman, R.C., Jacob, D.J., Turquety, S., Leibensperger, E.M., Murray, L.T., Wu, S., Gilliland,A.B., Avery, M., Bertram, T.H., Brune, W., Cohen, R.C., Dibb, J.E., Flocke, F.M., Fried, A.,Holloway, J., Neuman, J.A., Orville, R., Perring, A., Ren, X., Sachse, G.W., Singh, H.B.,Swanson, A., Wooldridge, P.J., 2007. Surface and lightning sources of nitrogen oxidesover the United States: magnitudes, chemical evolution, and outflow. Journal of Geo-physical Research: Atmospheres 112, 1–14.

Hutchinson, M.F., 1998. Interpolation of rainfall data with thin plate smoothing splines -part I: two dimensional smoothing of data with short range correlation. Journal ofGeographic Information & Decision Analysis 2, 153–167.

Irie, H., Kanaya, Y., Akimoto, H., Tanimoto, H., Wang, Z., Gleason, J., Bucsela, E., 2008. Val-idation of OMI tropospheric NO2 column data using MAX-DOAS measurements deepinside the North China Plain in June 2006: Mount Tai Experiment 2006. Atmos. Chem.Phys. 8, 6577–6586.

106 L. Liu et al. / Remote Sensing of Environment 199 (2017) 93–106

Irie, H., Kanaya, Y., Takashima, H., Gleason, J.F., Wang, Z., 2009. Characterization of OMItropospheric NO2 measurements in East Asia based on a robust validation compari-son. SOLA 5, 117–120.

Jia, Y., Yu, G., He, N., Zhan, X., Fang, H., Sheng,W., Zuo, Y., Zhang, D., Wang, Q., 2014. Spatialand decadal variations in inorganic nitrogen wet deposition in China induced byhuman activity. Sci Rep 4, 1–7.

Jiang,W.,Ma, J., Yan, P., Richter, A., Burrows, J.P., Nü, H., 2006. Characterization of NO2 pol-lution changes in Beijing using GOME satellite data. J. Applied Meteor. Sci 17, 67–72.

Kiese, R., Li, C., Hilbert, D.W., Papen, H., Butterbach-Bahl, K., 2005. Regional application ofPnET-N-DNDC for estimating the N2O source strength of tropical rainforests in theWet Tropics of Australia. Glob. Chang. Biol. 11, 128–144.

Kloog, I., Sorek-Hamer, M., Lyapustin, A., Coull, B., Wang, Y., Just, A.C., Schwartz, J., Broday,D.M., 2015. Estimating daily PM 2.5 and PM 10 across the complex geo-climate re-gion of Israel using MAIAC satellite-based AOD data. Atmos. Environ. 122, 409–416.

Lamsal, L.N., Martin, R.V., van Donkelaar, A., Steinbacher, M., Celarier, E.A., Bucsela, E.,Dunlea, E.J., Pinto, J.P., 2008. Ground-level nitrogen dioxide concentrations inferredfrom the satellite-borne Ozone Monitoring Instrument. Journal of Geophysical Re-search: Atmospheres 113, 1–15.

Lee, H.J., Koutrakis, P., 2014. Daily ambient NO2 concentration predictions using satelliteozone monitoring instrument NO2 data and land use regression. Environmental Sci-ence & Technology 48, 2305–2311.

Lee, C., Martin, R.V., van Donkelaar, A., Lee, H., Dickerson, R.R., Hains, J.C., Krotkov, N.,Richter, A., Vinnikov, K., Schwab, J.J., 2011a. SO2 emissions and lifetimes: estimatesfrom inverse modeling using in situ and global, space-based (SCIAMACHY andOMI) observations. Journal of Geophysical Research: Atmospheres 116, 1–13.

Lee, H., Liu, Y., Coull, B., Schwartz, J., Koutrakis, P., 2011b. A novel calibration approach ofMODIS AOD data to predict PM2.5 concentrations. Atmos. Chem. Phys 11,7991–8002.

Li, C., Narayanan, V., Harriss, R.C., 1996. Model estimates of nitrous oxide emissions fromagricultural lands in the United States. Glob. Biogeochem. Cycles 10, 297–306.

Lin, J.-T., 2012. Satellite constraint for emissions of nitrogen oxides from anthropogenic,lightning and soil sources over East China on a high-resolution grid. Atmos. Chem.Phys. 12, 2881–2898.

Liu, H., Jacob, D.J., Bey, I., Yantosca, R.M., 2001. Constraints from 210Pb and 7Be onwet de-position and transport in a global three-dimensional chemical tracer model driven byassimilatedmeteorological fields. Journal of Geophysical Research: Atmospheres 106,12109–12128.

Liu, Y., Park, R.J., Jacob, D.J., Li, Q., Kilaru, V., Sarnat, J.A., 2004. Mapping annual meanground-level PM2.5 concentrations using Multiangle Imaging Spectroradiometeraerosol optical thickness over the contiguous United States. Journal of GeophysicalResearch: Atmospheres 109, 1–10.

Liu, X., Duan, L., Mo, J., Du, E., Shen, J., Lu, X., Zhang, Y., Zhou, X., He, C., Zhang, F., 2011. Ni-trogen deposition and its ecological impact in China: an overview. Environ. Pollut.159, 2251–2264.

Liu, X., Zhang, Y., Han, W., Tang, A., Shen, J., Cui, Z., Vitousek, P., Erisman, J.W., Goulding, K.,Christie, P., 2013. Enhanced nitrogen deposition over China. Nature 494, 459–462.

Liu, L., Zhang, X., Lu, X., 2015. The composition, seasonal variation, and potential sourcesof the atmospheric wet sulfur (S) and nitrogen (N) deposition in the southwest ofChina. Environ. Sci. Pollut. Res. 1–13.

Liu, L., Zhang, X., Wang, S., Lu, X., Ouyang, X., 2016. A review of spatial variation of inor-ganic nitrogen (N) wet deposition in China. PLoS One 11.

Lü, C., Tian, H., 2007. Spatial and temporal patterns of nitrogen deposition in China: syn-thesis of observational data. Journal of Geophysical Research: Atmospheres 1984–2012, 112.

Lu, C., Tian, H., 2014. Half-century nitrogen deposition increase across China: a gridded time-series data set for regional environmental assessments. Atmos. Environ. 97, 68–74.

Lu, X., Jiang, H., Zhang, X., Liu, J., Zhang, Z., Jin, J., Wang, Y., Xu, J., Cheng, M., 2013. Estimat-ed global nitrogen deposition using NO2 column density. Int. J. Remote Sens. 34,8893–8906.

Ma, J., Beirle, S., Jin, J., Shaiganfar, R., Yan, P., Wagner, T., 2013. Tropospheric NO2 verticalcolumn densities over Beijing: results of the first three years of ground-based MAX-DOAS measurements (2008–2011) and satellite validation. Atmos. Chem. Phys. 13,1547–1567.

Mari, C., Jacob, D.J., Bechtold, P., 2000. Transport and scavenging of soluble gases in a deepconvective cloud. Journal of Geophysical Research Atmospheres 105, 22255–22268.

Mentel, T.F., Bleilebens, D., Wahner, A., 1996. A study of nighttime nitrogen oxide oxida-tion in a large reaction chamber—the fate of NO2, N2O5, HNO3, and O3 at different hu-midities. Atmos. Environ. 30, 4007–4020.

Nowlan, C., Martin, R., Philip, S., Lamsal, L., Krotkov, N., Marais, E., Wang, S., Zhang, Q.,2014. Global dry deposition of nitrogen dioxide and sulfur dioxide inferred fromspace-based measurements. Glob. Biogeochem. Cycles 28, 1025–1043.

Pan, Y., Wang, Y., Tang, G., Wu, D., 2012. Wet and dry deposition of atmospheric nitrogenat ten sites in Northern China. Atmos. Chem. Phys. 12, 6515–6535.

Phoenix, G.K., Emmett, B.A., Britton, A.J., Caporn, S.J., Dise, N.B., Helliwell, R., Jones, L.,Leake, J.R., Leith, I.D., Sheppard, L.J., 2012. Impacts of atmospheric nitrogen deposi-tion: responses of multiple plant and soil parameters across contrasting ecosystemsin long-term field experiments. Glob. Chang. Biol. 18, 1197–1215.

Racette, P., Adler, R.F.,Wang, J.R., Gasiewski, A.J., Jakson, D.M., Zacharias, D.S., 2009. An air-borne millimeter-wave imaging radiometer for cloud, precipitation, and atmosphericwater vapor studies. Journal of Atmospheric & Oceanic Technology 13, 1426–1428.

Reich, P.B., 2009. Elevated CO2 reduces losses of plant diversity caused by nitrogen depo-sition. Science 326, 1399–1402.

Ronsmans, G., Langerock, B., Wespes, C., Hannigan, J.W., Hase, F., Kerzenmacher, T.,Mahieu, E., Schneider, M., Smale, D., Hurtmans, D., De Mazière, M., Clerbaux, C.,Coheur, P.F., 2016. First characterization and validation of FORLI-HNO3 vertical pro-files retrieved from IASI/Metop. Atmos. Meas. Tech. 9, 4783–4801.

Rozanov, A., Bovensmann, H., Bracher, A., Hrechanyy, S., Rozanov, V., Sinnhuber, M., Stroh,F., Burrows, J., 2005. NO2 and BrO vertical profile retrieval from SCIAMACHY limbmeasurements: sensitivity studies. Adv. Space Res. 36, 846–854.

Russell, A., Perring, A., Valin, L., Bucsela, E., Browne, E., Wooldridge, P., Cohen, R., 2011. Ahigh spatial resolution retrieval of NO2 column densities from OMI:method and eval-uation. Atmos. Chem. Phys. 11, 8543–8554.

Russell, A., Valin, L., Cohen, R., 2012. Trends in OMI NO2 observations over the UnitedStates: effects of emission control technology and the economic recession. Atmos.Chem. Phys. 12, 12197–12209.

Stevens, C.J., Dise, N.B., Gowing, D.J., 2009. Regional trends in soil acidification and ex-changeable metal concentrations in relation to acid deposition rates. Environ. Pollut.157, 313–319.

Sutton, M.A., Bleeker, A., 2013. Environmental science: the shape of nitrogen to come. Na-ture 494, 435–437.

Sutton, M.A., Howard, C.M., Erisman, J.W., Billen, G., Bleeker, A., Grennfelt, P., vanGrinsven, H., Grizzetti, B., 2011. The European Nitrogen Assessment: Sources, Effectsand Policy Perspectives. Cambridge University Press.

Tie, X., Chandra, S., Ziemke, J.R., Granier, C., Brasseur, G.P., 2007. Satellite measurements oftropospheric column O3 and NO2 in Eastern and Southeastern Asia: comparison witha global model (MOZART-2). J. Atmos. Chem. 56, 105–125.

Van Damme, M., Erisman, J.W., Clarisse, L., Dammers, E., Whitburn, S., Clerbaux, C.,Dolman, A.J., Coheur, P.F., 2015. Worldwide spatiotemporal atmospheric ammonia(NH3) columns variability revealed by satellite. Geophys. Res. Lett. 42, 8660–8668.

Van Donkelaar, A., Martin, R.V., Park, R.J., 2006. Estimating ground-level PM2.5 usingaerosol optical depth determined from satellite remote sensing. Journal of Geophys-ical Research: Atmospheres 111.

Van Donkelaar, A., Martin, R.V., Brauer, M., Kahn, R., Levy, R., Verduzco, C., Villeneuve, P.J.,2010. Global estimates of ambient fine particulate matter concentrations from satel-lite-based aerosol optical depth: development and application. Environ. HealthPerspect. 118, 847.

Varma, S., Simon, R., 2006. Bias in error estimation when using cross-validation for modelselection. BMC Bioinformatics 7 (91-91).

Vet, R., Artz, R.S., Carou, S., Shaw, M., Ro, C.-U., Aas, W., Baker, A., Bowersox, V.C., Dentener,F., Galy-Lacaux, C., Hou, A., Pienaar, J.J., Gillett, R., Forti, M.C., Gromov, S., Hara, H.,Khodzher, T., Mahowald, N.M., Nickovic, S., Rao, P.S.P., Reid, N.W., 2014. A global as-sessment of precipitation chemistry and deposition of sulfur, nitrogen, sea salt, basecations, organic acids, acidity and pH, and phosphorus. Atmos. Environ. 93, 3–100.

Wan, L., Ma, Q., Zhang, J., Fu, Y., Zhang, X., 2011. Precise comparison of spatial interpola-tion for precipitation using KRIGING and TPS (thin plate smoothing spline) methodsin Loess Plateau. Science of Soil & Water Conservation 09, 79–87.

Wang, Y., McElroy, M.B., Boersma, K.F., Eskes, H.J., Veefkind, J.P., 2007a. Traffic restrictionsassociated with the Sino-African summit: reductions of NOx detected from space.Geophysical Research Letters 34.

Wang, Y., McElroy, M.B., Martin, R.V., Streets, D.G., Zhang, Q., Fu, T.-M., 2007b. Seasonalvariability of NOx emissions over east China constrained by satellite observations: im-plications for combustion andmicrobial sources. Journal of Geophysical Research: At-mospheres 112 (n/a-n/a).

Wang, S.W., Zhang, Q., Streets, D.G., He, K.B., Martin, R.V., Lamsal, L.N., Chen, D., Lei, Y., Lu,Z., 2012. Growth in NOx emissions from power plants in China: bottom-up estimatesand satellite observations. Atmos. Chem. Phys. 12, 4429–4447.

Wespes, C., Hurtmans, D., Clerbaux, C., Santee, M.L., 2009. Global distributions of nitric acidfrom IASI/MetOP measurements. Atmospheric Chemistry & Physics 9, 7949–7962.

Xu, W., Luo, X.S., Pan, Y.P., Zhang, L., Tang, A.H., Shen, J.L., Zhang, Y., Li, K.H., Wu, Q.H., Yang,D.W., Zhang, Y.Y., Xue, J., Li, W.Q., Li, Q.Q., Tang, L., Lv, S.H., Liang, T., Tong, Y.A., Liu, P.,Zhang, Q., Xiong, Z.Q., Shi, X.J., Wu, L.H., Shi, W.Q., Tian, K., Zhong, X.H., Shi, K., Tang,Q.Y., Zhang, L.J., Huang, J.L., He, C.E., Kuang, F.H., Zhu, B., Liu, H., Jin, X., Xin, Y.J., SHi,X.K., Du, E.Z., Dore, A.J., Tang, S., Collett Jr., J.L., Goulding, K., Sun, Y.X., Ren, J., Zhang,F.S., Liu, X.J., 2015. Quantifying atmospheric nitrogen deposition through a nationwidemonitoring network across China. Atmos. Chem. Phys. 15, 12345–12360.

Yan, H., 2004. Modeling spatial distribution of climate in china using thin plate smoothingspline interpolation. Scientia Geographica Sinica 24, 163–169.

Yap, X., Hashim, M., 2013. A robust calibration approach for PM10 prediction fromMODISaerosol optical depth. Atmos. Chem. Phys. 13, 3517–3526.

Yu, L., Wang, J., Gong, P., 2013. Improving 30 m global land-cover map FROM-GLC withtime series MODIS and auxiliary data sets: a segmentation-based approach. Int.J. Remote Sens. 34, 5851–5867.

Zhan, X., Yu, G., He, N., Jia, B., Zhou, M., Wang, C., Zhang, J., Zhao, G., Wang, S., Liu, Y., 2015.Inorganic nitrogen wet deposition: evidence from the north-south transect of EasternChina. Environ. Pollut. 204, 1–8.

Zhang, Q., Streets, D.G., He, K., Wang, Y., Richter, A., Burrows, J.P., Uno, I., Jang, C.J., Chen, D.,Yao, Z., Lei, Y., 2007. NOx emission trends for China, 1995–2004: the view from theground and the view from space. Journal of Geophysical Research: Atmospheres112 (n/a-n/a).

Zhang, Y., Liu, X., Fangmeier, A., Goulding, K., Zhang, F., 2008. Nitrogen inputs and isotopesin precipitation in the North China Plain. Atmos. Environ. 42, 1436–1448.

Zhang, L., Jacob, D.J., Knipping, E.M., Kumar, N., Munger, J.W., Carouge, C., Van Donkelaar,A., Wang, Y., Chen, D., 2012. Nitrogen deposition to the United States: distribution,sources, and processes. Atmos. Chem. Phys. 12, 4539–4554.

Zhao, C., Wang, Y., 2009. Assimilated inversion of NOx emissions over east Asia using OMINO2 column measurements. Geophys. Res. Lett. 36, 1–5.

Zhao, Y., Zhang, L., Chen, Y., Liu, X., Xu, W., Pan, Y., Duan, L., 2017. Atmospheric nitrogendeposition to China: a model analysis on nitrogen budget and critical load exceed-ance. Atmos. Environ. 153, 32–40.

Zhu, J., He, N., Wang, Q., Yuan, G., Wen, D., Yu, G., Jia, Y., 2015. The composition, spatialpatterns, and influencing factors of atmospheric wet nitrogen deposition in Chineseterrestrial ecosystems. Sci. Total Environ. 511, 777–785.