TRACE: Tennessee Research and Creative TRACE: Tennessee Research and Creative Exchange Exchange Multi-model study of HTAP II on sulfur and nitrogen deposition Multi-model study of HTAP II on sulfur and nitrogen deposition

. This study uses multi-model ensemble results of 11 models from the second phase of Task Force Hemispheric Transport of Air Pollution (HTAP II) to calculate the global sulfur (S) and nitrogen (N) deposition in 2010. Modeled wet deposition is evaluated with observation networks in North America,


Introduction
Nitrogen (N) plays an important role in the balance of the global ecosystem. Human activities such as consumption of fossil fuels, production and usage of N fertilizers, and livestock cultivation disturb the N cycle in the ecosystem (Vitousek et al., 1997;Galloway et al., 2008). Estimation under the IPCC Special Report on Emissions Scenarios (SRES) A2 scenario predicts that the N deposition over land will increase by a factor of ∼ 2.5 from 2000 to 2100 (Lamarque et al., 2005). Elevated N deposition can cause exceedance of N critical loads on ecosystems . About 11 % of the world's natural vegetation has already received N deposition that exceeds the critical load in 2000 (Dentener et al., 2006). The most affected regions are eastern Europe (80 %), South Asia (60 %) and East Asia (40-50 %). This percentage will be 40 % for the world's protected areas in 2030 (Bleeker et al., 2011). Elevated S and N deposition are also associated with a host of environmental issues such as acidification and eutrophication of the terrestrial system (Bouwman et al., 2002), loss of ecosystem biodiversity (Bobbink et al., 2010), harming heterotrophic respiration and disturbing the soil decomposition process (Janssens et al., 2010), although some studies have found that increasing N deposition could benefit the carbon uptake by land processes (Reay et al., 2008;Holland et al., 1997). Similar to the terrestrial system, over-richness of S and N deposition are also threats to the aquatic system by acidification (Doney et al., 2007) and eutrophication of the ocean (Bergstrom and Jansson, 2006;Jickells, 2006;Jickells et al., 2017).
In order to understand S and N deposition, a number of global-scale studies have been conducted in the last decade. Dentener et al. (2006) investigated the current (2000) and future (2030) S and N deposition with multi-model ensemble results of ACCENT IPCC Fourth Assessment Report (AR4) experiment (PhotoComp). Model evaluation showed that 60-70 % of modeled wet deposition is within ±50 % of measurements in Europe and North America. NH x deposition was overestimated in South Asia, and NO y deposition was underestimated in East Asia. Eleven percent of the world's nature vegetation received N deposition that exceed the critical load in 2000, and this percentage would increase to 17 % under current air quality legislation and 25 % under the IPCC SRES A2 scenario in 2030. Sanderson et al. (2008) used the ensemble results of the first phase of the Task Force Hemispheric Transport of Air Pollution (HTAP I) to estimate the long-range transport of oxidized nitrogen between Europe, North America, South Asia and East Asia. Results showed that 8-15 % of NO x from source regions could be transported beyond the distance of 1000 km, which indicated the impact of intercontinental transport of air pollutants on deposition. Lamarque et al. (2013) calculated the S and N deposition in 2000 using a multi-model ensemble of the Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP). Model performance on NO − 3 wet depo-sition was comparable with PhotoComp and HTAP I, but NH + 4 wet deposition was not well simulated. Simulations with the projected emissions in 2100 under four Representative Concentration Pathways (RCPs) indicated that N deposition is likely to substantially increase in Latin America, Africa and parts of Asia (especially South Asia) in the future. Vet et al. (2014) conducted a comprehensive evaluation on the multi-model performance on deposition of HTAP I. The results underestimated the wet deposition at observation sites with high observed N deposition in North America, southern and northern Europe, and East Asia. Dry deposition in the Unites States was found to deviate with inferential dry deposition data. Kanakidou et al. (2016) used the ACCMIP simulation results under historical, RCP6.0 and RCP8.5 emission scenarios to estimate the changes in N deposition driven by human activity in the past (1850), present (2005) and future (2050). The results showed that organic nitrogen (ON) from primary emission and secondary organic aerosol (SOA) account for 20-30 % of total N deposition. The impact of human activity on N deposition has increased from 15 % in the past to 60 % in present years, and this impact is likely to persist in the future. Bian et al. (2017) examined the possible factors causing the inter-model diversity in simulating NO − 3 and NH + 4 deposition by comparing the results of nine models participating in the third phase of Aerosol Comparisons between Observations and Models (AeroCom III). The results showed that models have large differences in calculating the pH adjustment for the effective Henry's law constant, which could greatly influence the simulation of NH x wet deposition.
These studies give a clear view of S and N deposition in the early 2000s. However, large changes are seen in the global N emissions in the last decade (van der A et al., 2008), including a large increase in China (Q. van der A et al., 2006;Richter et al., 2005;Kurokawa et al., 2013;Zhang et al., 2007;Li et al., 2017) and general decreases in both Europe (Tørseth et al., 2012) and the eastern United States (Kim et al., 2006). In addition, ground observations and satellite measurements show large increases in the dry deposition in the western United States, eastern Europe and east China, together with decreases in the eastern United States, western Europe and Japan (Jia et al., 2016). Thus, a follow-up study is needed to update our knowledge about the S and N deposition with emission changes in the 21st century.
In this study, we use the multi-model mean (MMM) of 11 global models from the second phase of the HTAP (HTAP II) to calculate the S and N deposition in 2010. Section 2 gives a short description of the HTAP II and introduces the method to develop MMM and metrics for model evaluation. Section 3.1 evaluates MMM performance on wet deposition with observations from networks in North America, Europe and East Asia. The modeled dry deposition is compared with the inferential data in North America (see details in Sect. 3.1). We also compare the model performance of this study with previous studies in 2001 of PhotoComp (Dentener et al., 2006), HTAP I (Vet et al., 2014) and ACCMIP (Lamarque et al., 2013). Section 3.2 and 3.3 estimate the S and N deposition in continental, coastal and ocean regions in 2010. By comparing our results with deposition in 2001 of HTAP I, we investigate the changes of deposition in the past 10 years. We conclude with the findings in Sect. 4.

Model description and experiment setup
The HTAP was developed in 2005 aiming at understanding the long-range transport of air pollution and its impact on regional air quality. HTAP I has involved more than 20 global models with the base simulation year of 2001. A comprehensive assessment has been published to summarize the findings in HTAP I with respect to the long-range transport of (1) ozone and particulate matter, (2) mercury and (3) persistent organic pollutants (HTAP, 2010). The HTAP II was launched in 2012 with the base year of 2010. A prescribed emission inventory called HTAPv2.2 is used by models from different groups to facilitate a fair evaluation of the models' ability and uncertainty (Galmarini et al., 2017). It is a harmonized emission inventory formed by the best estimation of emissions from different organizations, including the Environmental Protection Agency (EPA) of the United States, the EPA and Environment Canada, the European Monitoring and Evaluation Programme (EMEP) and the Netherlands Organisation for Applied Scientific Research (TNO), the Model Intercomparison Study for Asia (MICS-Asia III) and the Emission Database for Global Atmospheric Research (EDGARv4.3). The development of the emission inventory is described in Janssens-Maenhout et al. (2015). The following are some highlights of the findings from HTAP II. Stjern et al. (2016) estimated the impact of domestic and foreign emission change of black carbon, organic carbon and SO 4 on regional radiative forcing. Huang et al. (2017) studied the impact of intercontinental outflow from East Asia to North America on O 3 pollution by simulating the regionalscale Sulfur Transport and dEposition Model (STEM) with boundary conditions provided by three global transport models. Jonson et al. (2018) conducted a source apportionment for O 3 pollution in Europe and calculated the global contributions of emissions. Tan et al. (2018) investigated the intercontinental export of sulfur and nitrogen emission and its impact on local deposition.
Among the 20 models participating in the HTAP II (configurations described in Stjern et al., 2016), 11 models (i.e., CAM-Chem, CHASER_re1, CHASER_t106, EMEP_rv48, GEMMACH, GEOS5, GEOSCHEMAJOINT,OsloCTM3v.2,GOCARTv5, submitted the model outputs of S and N deposition. To develop the MMM, all models are interpolated to a uniform 0.1 • × 0.1 • horizontal resolution (the same resolution as the emission inventory) by linear interpolation. Then the MMM of the emission/deposition quantities of each of S and N is calculated by averaging (arithmetic mean) all available model outputs. More details are demonstrated in Sect. 2.2.
The base year of simulation is 2010, with an additional 6month run as model spin-up. The administrative boundaries of 17 regions are shown in Fig. S1 in the Supplement. Details about the experiment setup can be found in Galmarini et al. (2017).

Method for calculating the MMM
To make the discussion clear, we define the terms as follows: the continental regions refer to all land regions including the Antarctic. The coastal regions are defined in Fig. S1. In Sect. 3.2 and 3.3, the S deposition contains gas phase SO 2 deposition and aerosol SO 2− 4 deposition. The N deposition includes oxidized nitrogen (NO y ) deposition and reduced nitrogen (NH x ) deposition. NO y deposition is composed of all oxidized nitrogen species except N 2 O. Based on the model outputs, NO y deposition mainly includes NO 2 , HNO 3 , aerosol NO − 3 , peroxyacyl nitrate (PAN) and other organic nitrates than PAN. NH x deposition consists of gas phase NH 3 deposition and aerosol NH + 4 deposition. Before constructing the MMM, we check the quality of model outputs using two criteria. First, we check the mass balance of each model by comparing its global deposition with its emission. A models is excluded if its deposition value falls outside the range of ±20 % of its emission values. The second criterion is to check if the result of a model is away from the mean value of all models. We adopt the median of models ±1.5 × interquartile by Vet et al. (2014) as upper and lower limits and check the values separately for all species of deposition and emission. The models used to develop the MMM and their values are summarized in Tables S1-S3 in the Supplement. After the quality check, we calculate the mean value of species using Eq. (1) with all available model outputs. Then, we combine all of the related species into total deposition/emission by Eq. (2).
For both Eqs.
(1) and (2), i is the individual model and j is the species of deposition/emission from model outputs. S i (j ) is the species j from model i, and S MMM (j ) is the MMM of species j .

Model evaluation metrics
To compare the model performance with previous projects consistently, we adopt the following metrics in Lamarque et al. (2013): linear fit slope, mean bias, mean observa-tion, mean model, correlation coefficient (R) and fraction (of model results) within ±50 % (of observations). In addition, we use four statistical metrics following Eqs. (3)-(6).  and NH + 4 wet deposition with observations at the NADP, EMEP and EANET stations. Performances of individual models can be found in Figs. S2-S4. The SO 2− 4 wet deposition comprises gas phase SO 2 and aerosol SO 2− 4 wet deposition. The NO − 3 wet deposition includes gas phase HNO 3 and aerosol NO − 3 wet deposition. The NH + 4 wet deposition contains gas phase NH 3 and aerosol NH + 4 wet deposition. Figure 2 displays the spatial distributions of MMM SO 2− 4 , NO − 3 and NH + 4 wet deposition (contours) with observations (filled circles). In terms of SO 2− 4 wet deposition, the MMM results are consistent with observations at the NADP stations with a slope close to 1 (0.9) and a high R value (0.8) (Fig. 1a). The MFB and MFE are 9 and 32 %, respectively, indicating slight overestimation. According to Fig. 2a, the observed SO 2− 4 wet deposition is highest in the northeastern United States, and this spatial distribution is well captured by MMM. The EMEP stations are well simulated with low MFB (−7 %) and MFE (25 %) (Fig. 1b). The MMM predictions are within ±50 % of observations at 87 % of the stations. According to Fig. 2b, one station in Poland and one station in Norway, with observed SO 2− 4 wet deposition of 1000 and 500 mg(S) m −2 yr −1 , respectively, are both underestimated by 50 %. We evaluate the model performances on simulating precipitation (Figs. S5 and S6). For the Norwegian site, the observed precipitation is 1566 mm yr −1 , and the MMM underestimated the precipitation by 49 %, which fits well for the 50 % underestimation of SO 2− 4 wet deposition at this site. For the Polish site, the observed precipitation is 1137 mm yr −1 and the MMM underestimated the precipitation by 21 %. The underestimation in precipitation could partly explain the negative model bias in simulating SO 2− 4 wet deposition. Another possible reason is the high topography of the sites. The Polish site is 1603 m a.s.l. (above sea level), which is one of the highest sites among the European sites. Similar to the Polish site, one site in Spain, which is at a height of 1360 m, is underestimated by 142 mg(S) m −2 yr −1 (59 %) for SO 2− 4 wet deposition, while its precipitation is well simulated with a slight positive model bias of 5 %. At the EANET stations, very high SO 2− 4 concentrations were measured at some stations, probably correlated with dust emission (Dentener et al., 2006). To make a consistent comparison with previous projects, we ignore the measurements coincident with measured calcium (Ca 2+ ) deposition larger than 20 mole m −2 yr −1 . The evaluation (Fig. 1c) shows that the SO 2− 4 wet deposition is generally underestimated at the EANET stations by 23 % (MFB) and 44 % (MFE). The stations in Korea and Vietnam are generally underestimated by more than 200 mg(S) m −2 yr −1 (Fig. 2c). On the other hand, the SO 2− 4 wet deposition is generally well simulated in Indonesia, the Philippines, Thailand and Japan. Overall, 76 % of the stations of all networks predicted quantities within ±50 % of observations. The EANET stations have the highest model bias among the three networks. It should be noted that for the three excluded stations (located in China) with high Ca 2+ deposition, the SO 2− 4 wet deposition is largely underestimated by more than 1000 mg(S) m −2 yr −1 (not shown in figures). If we include these stations in the model evaluation, the mean bias for East Asia changes from −160 to −300 mg(S) m −2 yr −1 . We also realize that the observation stations in China are mainly located along the eastern and southern coast, while the highest modeled deposition is found in the inland areas. Therefore, it is hard to conduct a comprehensive evaluation over this region due to unavailable measured data in the inland areas.
For NO − 3 wet deposition, the MMM results agree well with observations at the NADP stations, as shown by the linear regression line in Fig. 1d with a slope of 1.2 and R value of 0.9. However, the amount of deposition is overestimated by 33 % (MFB) and 36 % (MFE). According to Fig. 2d, there is a general tendency of overestimation throughout the stations in the United States, especially the stations located in the Midwest and southeast. At the EMEP stations, the NO − 3 wet deposition is well simulated with low MFB of −5 % and MFE of 24 % (Fig. 1e). The modeled deposition is within ±50 % of observed deposition at more than 90 % of the stations. The MMM results are close to the observations at stations with deposition lower than 400 mg(N) m −2 yr −1 but generally underestimate the deposition at stations with higher observations. According to Fig The modeled NH + 4 wet deposition agrees well with observations at the NADP stations with MFB of 7 % and MFE of 25 % (Fig. 1g). About 88 % of modeled deposition is within ±50 % of observations. The MMM has captured the high deposition in the United States Midwest well but slightly underestimates the deposition in the southeast ( Fig. 2g). At the EMEP stations, the NH + 4 wet deposition is well simulated with MFB of −1 % and MFE of 36 % (Fig. 1h). The MMM results are close to the observations at most stations and reproduce the high deposition in Germany and Italy well (Fig. 2h). Some stations in Norway and Poland are slightly underestimated by 100-200 mg(N) m −2 yr −1 . These stations all report higher observed deposition than 500 mg(N) m −2 yr −1 . The NH + 4 wet deposition is underestimated at the EANET stations by 10 % (MFB) and 50 % (MFE) (Fig. 1i). The MMM has captured the high deposition in eastern China and Indonesia well but generally underestimates the NH + 4 wet deposition at the Russian stations (Fig. 2i). In addition, the observed deposition at the three Korean stations is relatively high (∼ 500-600 mg(N) m −2 yr −1 ), but the MMM fails to reproduce any of them. There could be a missing emission source in that region. Overall, 81 % of the MMM predictions are within ±50 % of observations at stations of all networks. The NH + 4 wet deposition is somewhat underestimated in all three regions, especially in East Asia. Table 1 compares the model performance of this study (HTAP II) with previous projects of PhotoComp (Dentener et al., 2006), HTAP I (Vet et al., 2014) and ACCMIP (Lamarque et al., 2013). It should be noted that the emission inputs, simulation periods and participating groups of this study (year 2010) are different from those of the previous projects (year 2001). Although the observations are from the same networks, the previous projects used 3-year averaged observations of 2000-2002 and this study used those of 2009-2011. Due to these differences, the model performances may not be totally comparable. In terms of SO 2− 4 wet deposition, the model performance is similar to that for previous projects in North America, with 4-6 % higher percentage of stations within ±50 % of observations. Large improvement is found in Europe. The absolute mean bias decreases from 50-130 to 30 mg(S) m −2 yr −1 . There is a 10 % increase in the fraction of stations within ±50 % of observations. At the East Asian stations, the absolute mean bias decreases slightly from 180-290 to 160 mg(S) m −2 yr −1 . But the R value and fraction within ±50 % have somewhat declined. For NO − 3 wet deposition, HTAP II performs similar to the ensembles used in previous projects in North America but slightly better in Europe, with a lower mean bias and 5 % increase in the fraction within ±50 % of observations. The model mean bias at the East Asian stations has decreased significantly from ∼ 50 to ∼ 1 mg(N) m −2 yr −1 . However, the biases for individual models are large (Fig. S3). Large negative model bias is found in Southeast Asia, and improvements are needed in the future. In terms of NH + 4 wet deposition, HTAP II shows similar R values to those of ensembles used for the previous projects in North America, with slightly lower model bias. HTAP II shows considerable improvement in Europe. The slope of the regression line increases from 0.3-0.4 to 0.6, and the mean bias decreases from as large as −95 to −4 mg(N) m −2 yr −1 . For East Asia, the slope, mean bias and R values for HTAP II are all within the ranges of the previous projects, while the absolute mean bias decreases from 70-140 to 30 mg(N) m −2 yr −1 .

Dry deposition
The number of dry deposition measurements is limited due to difficulty in measuring the dry deposition directly by instruments. This study evaluates the dry deposition in the United States using information from the Clean Air Status and Trends Network (CASTNET). Instead of direct measurements, the data are produced by an "inferential" method, using calculations of the measured concentration of species and modeled dry deposition velocities. We use the 3-year average data of 2009-2011 from CASTNET and adopt the same selection criteria as we did for the wet-deposition measurements. Data from 81 stations out of 85 available stations are used for comparison. Figure 3 shows the scatterplots of the MMM SO 2 , SO 2− 4 , NO − 3 , HNO 3 and NH + 4 dry deposition with inferential data at the CASTNET stations. Performances of individual models can be found in Figs. S7-S11. The modeled SO 2 dry deposition is 240 (170 %) mg(S) m −2 yr −1 higher than the inferential data, and only 5 % of the stations are within ±50 % of the inferential values. There are smaller discrepancies for SO 2− 4 dry deposition (14 mg(S) m −2 yr −1 and 60 %) between model and inferential results. Modeled NO − 3 , HNO 3 and NH + 4 dry deposition is generally 50-100 % higher than the inferential data, and the fraction within ±50 % is about 15 %. Figure 4 shows the spatial distributions of MMM dry deposition (contours) with the inferential data (filled circles). The MMM results are consistent with the inferential data in the western United States, where the dry deposition is generally low. And both datasets predict high NO − 3 dry deposition in western California. Large disagreements are found in the eastern United States. In the Midwest (mainly Indiana and Ohio), although both results estimate higher N (NO − 3 , HNO 3 and NH + 4 ) dry deposition in this region than the others, the prediction of MMM is 20-30 mg(N) m −2 yr −1 higher than the inferential data at every station. In addition, the MMM estimates much higher deposition in the southern and northeastern United States than in the western United States, but this gradient is much weaker in the inferential data. Table 2 compares the model performance of this study (HTAP II) with that of the models used in HTAP I (Vet et al., 2014) and ACCMIP . HTAP I used the 2001 simulation results and compared them with 3-year average (2000)(2001)(2002) CASTNET data. ACCMIP used 10-year averages of both MMM and CASTNET data from 2000 to 2009. The N dry deposition values for all projects contain NO − 3 , NH + 4 and HNO 3 , and the S dry deposition includes SO 2 and SO 2− 4 . Both HTAP I and HTAP II overestimated the S and N dry deposition, but HTAP II has ∼ 100 mg(S) m −2 yr −1 and ∼ 80 mg(N) m −2 yr −1 lower mean bias than HTAP I. The comparison with ACCMIP results may not be solid since there are large differences in simulation periods. Generally, the HTAP II performance is similar to ACCMIP for NH + 4 , SO 2 and SO 2− 4 dry deposition but has a larger mean bias for HNO 3 dry deposition.
Since the CASTNET dry deposition is not actually measured but instead a calculation of measured concentration of species and modeled dry deposition velocities, it is necessary to investigate which factor of these two contributes to the model bias. We compare the modeled air pollutant concentrations with CASTNET measurements as shown in Tables S4-S8. The MMM overestimates the SO 2 , SO 2− 4 , HNO 3 , NO − 3 and NH + 4 concentrations by 394, 40, 217, 135 and 173 %, respectively. It should be noted that the CASTNET sites are generally located in rural regions that are away from emission sources (Sickles and Shadwick, 2008); thus the measured concentrations of air pollutants are relatively low compared with those of urban sites, while the resolutions of the HTAP II models range from 0.5 to 3 • and are not fine enough to reproduce the characteristic of some rural sites. The models with finer resolutions except CHASER_t106 model (i.e., EMEP_rv48 (0.5 • × 0.5 • ) and SPRINTARS (1.1 • × 1.1 • )) generally perform better than the others, while models with coarse resolutions (i.e., CHASER_re1 (2.8 • × 2.8 • ) and OsloCTM3.v2 (2.8 • × 2.8 • )) generally do not perform well for all species. This could explain the overestimation of air pollutant concentrations at the CASTNET sites.
In order to check the differences of modeled dry deposition velocity between CASTNET and HTAP II models, we adopt the general approach for calculating dry deposition velocity from Wesely (1989): where V d is the deposition velocity, F c is the dry deposition flux and C a is the concentration of species. The negative mark indicates the direction of the dry deposition velocity. This scheme has been widely adopted in global models (Wesely and Hicks, 2000) with modifications. We compare the calculated dry deposition velocity of models and CASTNET (Tables S9-S13). The mean biases of dry deposition velocities for MMM are −8, 0.3, 7, 19 and 2 % for SO 2 , SO 2− 4 , HNO 3 , NO − 3 and NH + 4 , respectively, which are much lower than those of air pollutants. The model bias for dry deposition at the CASTNET sites mainly comes from the model overprediction of air pollutant concentration.   The CASTNET data are 54 % lower for SO 2 dry deposition and 47 % lower for HNO 3 dry deposition than CAPMoN, mainly due to using different models to calculate the dry velocity. Table 3 lists the calculated amount of S emission and deposition in continents, coastal regions and oceans. We calculate the ratio of S deposition to S emission (Fig. 5c). Because it is not clear how dimethyl sulfide (DMS) emission will transfer to S deposition, this ratio does not represent the transformation of S emission to deposition. For continental non-coastal regions, the average ratio is 85 % (86 % if taking coastal regions into consideration). In highemission regions, this ratio can be viewed as the "scavenging" effect of S pollution by deposition. In major source regions of S emission (i.e., the North China Plain, the Midwest of the United States and India), the ratios are only slightly higher than 50 %, while in low-S-emission regions (< 10 mg(S) m −2 yr −1 ) the ratios could exceed 400 % (areas with white color in Fig. 5c). This result indicates that the deposition in these regions is largely affected by long-range transport of pollution from other regions. The impact of intercontinental transport of air pollutants on deposition can be quantified by the emission perturbation experiments in HATP II. Results from those experiments will be discussed in another paper (Tan et al., 2018).

Total S deposition
We compare the S emission and deposition in 2010 from HTAP II with those in 2001 from HTAP I (Vet et al., 2014) (Table 3). We re-calculate the HTAP I results according to the regions defined in HTAP II (Fig. S1), so the HTAP I results may look different from those in Table 2 of Vet et al. (2014). Because different models were used for each of the two ensembles compared, associated uncertainty is expected. In addition, emissions in HTAP I were not prescribed, so each modeling group used its own best estimation of emissions (Sanderson et al., 2008). Conversely, all models in HTAP II used the same anthropogenic emission (although there were still differences in natural emission). Globally, the S emission decreases by 5 Tg ( (Lu et al., 2010). For coastal regions, Europe has experienced a 2 Tg(S) (54 %) decrease, and East Asia has experienced a 1 Tg(S) (43 %) decrease in S emission. Other regions have relatively small (0-0.6 Tg(S)) changes. The global S deposition decreases by 2 Tg(S), with a 5 Tg(S) (11 %) decrease in continental non-coastal regions, 4 Tg(S) (16 %) increase in non-coastal ocean regions and 1 Tg(S) (5 %) decrease in coastal regions. The regions with the largest change in deposition coincide with those having big changes in emission. For instance, Europe experiences a 5 Tg(S) decrease in S deposition with 8 Tg(S) decrease in emission, and South Asia receives 2 Tg(S) more S deposition with a 2 Tg(S) increase in emission. Figure S13b compares the S deposition in HTAP II with that in HTAP I. Declined S deposition is found in large areas of the eastern United States and Europe (400-1500 mg(S) m −2 yr −1 ). Regions with increased S deposition are India and Indonesia (100-800 mg(S) m −2 yr −1 ). In China, there is a mixture of both increases and decreases in S deposition over different areas. The changes in S depositions agree well with changes in S emissions (Fig. S13a). During China's 11th FYP, one of the main technologies to control the SO 2 emission was to install flue-gas desulfurization (FGD) in power plants (Cao et al., 2009). The effectiveness of this technology in removing SO 2 emission varies considerably regionally, as a result of several factors, such as the coverage of FGD technology in power plants, local re-    About 38 % of global NO y deposits in the ocean, lower than 43 % in PhotoComp (Dentener et al., 2006) and 42 % in HTAP I (Vet et al., 2014), but higher than the 30 % estimated by Lamarque et al. (2005). It should be noted that these values are calculated by the land-ocean mask, which may differ among different studies. For non-coastal ocean regions, the NO y deposition is 13 Tg(N), accounting for 22 % of the global deposition, while the emission from oceans is only 2 Tg(N), about 4 % of global emission. The difference of 11 Tg(N) indicates NO y transport from continents to the open ocean. The Antarctic has near-zero NO x emission but receives 0.1 Tg(N) NO y deposition. Deposition has been a non-negligible pathway through which the human pollution contaminates nearly untouched areas.
We calculate the ratio of NO y deposition to NO x emission (Fig. 6c). In continental non-coastal regions, the average ratio is 74 % (81 % if taking coastal regions into consideration). In high NO x emission regions (i.e., North America, East Asia and South Asia), about 60-80 % of the NO y is removed by deposition, with large regional variations. For low-emission regions (i.e., North Africa and central Asia), the ratio can reach higher than 90 %. In coastal regions and the open ocean, the ratio is generally over 200 %. Instead of the local emission, the transport of air pollutants from elsewhere is the major source of deposition.

NH x deposition
The global NH x deposition is 54 Tg(N) in 2010, with 69 % in continental non-coastal regions, 19 % in non-coastal ocean regions and 13 % in coastal regions (Table 4). For continental non-coastal regions, South Asia receives 16 % of global NH x depositions, followed by East Asia (13 %). The whole Indian subcontinent receives higher NH x depositions than 2000 mg(N) m −2 yr −1 (Fig. 6e). Also, the Asian regions have several high-deposition areas: the North China Plain and Indonesia (1200-2000 mg(N) m −2 yr −1 ), Japan, Thailand, Vietnam and Myanmar (500-600 mg(N) m −2 yr −1 ). Other regions with high NH x deposition are the United States Midwest, Germany, France, northern Italy, southern Brazil and Ethiopia (400-800 mg(N) m −2 yr −1 ). Distributions of components of NH x deposition are shown in Fig. S15.
Coastal regions of Southeast Asia (3 %), East Asia (2 %) and South Asia (2 %) receive the largest NH x deposition (∼ 200-400 mg(N) m −2 yr −1 ). The east coasts of North America and Mexico also have high NH x deposition (150-200 mg(N) m −2 yr −1 ). Compared to NO y deposition, the NH x deposition in coastal regions is relatively lower. The ocean receives 17 Tg(N) of NH x deposition in 2010, within the range of 13-29 Tg(N) estimated by Duce et al. (2008) but lower than the 23.5 Tg(N) estimated by Dentener et al. (2006) and 21.4 Tg(N) estimated by Vet et al. (2014). About 31 % of NH 3 emission is deposited in ocean areas, similar to the 31 % estimated by Dentener et al. (2006) and 30 % estimated by Lamarque et al. (2005) but slightly lower than 37 % in Photo-Comp (Dentener et al., 2006) and 37 % in HTAP I (Vet et al., 2014). The ocean emitted 12 Tg(N) of NH 3 in 2010, which means that at least 5 Tg(N) of NH x deposition in oceans in 2010 came from continental regions. This value is considerably lower than the 13 Tg(N) of deposition-emission difference for NO y (including the 2 Tg(N) difference in coastal regions). A possible explanation is that NH 3 has a short lifetime in the atmosphere, which makes it more likely to deposit close to where it is emitted (Shen et al., 2016), while NO x can be oxidized to organic nitrate (Moxim et al., 1996), which facilitates the long-range transport from land to open ocean.
We calculate the ratio of NH x deposition to NH 3 emission (Fig. 6f). The average ratio is 87 % for continental noncoastal regions (92 % if also considers the coastal regions). The ratios are generally higher than those of NO y deposition (74 and 81 %), since a large proportion of NH x is deposited near the source. The ratios are generally over 400 % for coastal areas but less than 100 % in the open ocean (70-90 %). This is because there is less continental NH x transported to the open ocean than to coastal regions.

N deposition
The global N deposition in 2010 is 113 Tg(N), with 65 % of deposition in the continental non-coastal regions, 20 % in non-coastal oceans and 15 % in coastal regions (Table 4). East Asia (13 %) and South Asia (11 %) receive the largest amount of N deposition, consistent with the fact that they are also the largest N emission sources (16 and 13 %, respectively). The deposition reaches 3000 mg(N) m −2 yr −1 over eastern China (especially the North China Plain) and 2000 mg(N) m −2 yr −1 over India and Southeast Asia (Thailand, Vietnam and Malaysia). Other regions of high N deposition are the northeastern United States and western Europe (800-1200 mg(N) m −2 yr −1 ), Mexico, Central Amer-ica, Brazil, northern sub-Saharan Africa and the northeastern Middle East (500-600 mg(N) m −2 yr −1 ). For coastal regions, the east coast of the United States, all coasts of India and the east coast of East Asia are identified with relatively high deposition (> 600 mg(N) m −2 yr −1 ). Table 5 compares the N emission and deposition in HTAP II with HTAP I. The global N emission increases from 105 to 115 Tg(N), with a 12 Tg(N) (15 %) increase in continental non-coastal regions and a 2 Tg(N) (14 %) decrease in coastal regions. The change in the ocean is small due to increased NO y deposition but decreased NH x deposition. For continental non-coastal regions, increases in N emission are found in South Asia (5 Tg(N), 56 %), East Asia (4 Tg(N), 26 %) and Southeast Asia (2 Tg(N), 58 %), while the emission in Europe decreases by 1 Tg(N) (12 %). The emission changes in coastal regions are relatively small. The global N deposition increases by 7 Tg(N), with a 9 Tg(N) (14 %) increase in continental non-coastal regions and a 2 Tg(N) decrease in the ocean. Asian regions also have experienced the largest increases in deposition, and the amounts are identical with corresponding emission changes. Figure  The global N dry and wet deposition is 40 and 73 Tg(N) yr −1 in 2010, respectively. We calculate the ratio of dry deposition as dry deposition dry deposition + wet deposition ×100 %. For continental non-coastal regions, about 44 % (range from 35 to 61 %) of the N deposition comes from dry deposition (42 % if taking coastal regions into consideration). If the overestimation of N dry deposition in Sect. 3.1.2 is considered, this ratio could be even lower. Desert areas (e.g., the Sonoran, Mojave and Chihuahuan deserts near the west coast of North America; the Sahara in North Africa; the Arabian Desert in the Middle East; and the Great Victoria Desert in Australia) are seen with high ratios of dry deposition (80 %) (red-colored regions in Fig. 7c). This outcome is reasonable since these areas generally lack precipitation. Low fractions of dry deposition (30 %) are found in Russia, western China, Southeast Asia, Australia and Central America. Almost all coastal regions are dominated by wet deposition. A study by Jickells (2006) reported a dry deposition ratio of 21-45 % for the east coast of the United States, and a study by Baker et al. (2010) suggested a ratio of 15-22 % for the Atlantic Ocean. Our study receives similar ratios for these coastal regions. A study by Bey et al. (2001) found an outflow of NO y from Asia over the western Pacific Ocean through deposition. According to this study, about 70 % of this land-to-ocean export of NO y deposition is through wet deposition (Fig. 7a).
The NH x and NO y deposition is 54 and 59 Tg(N) yr −1 in 2010, respectively. The average ratio of NH x deposition (cal-culated as NH x deposition NH x deposition + NO y deposition ×100 %) for continental non-coastal regions is 47 % (45 % if coastal regions are taken into consideration). South Asia (71 %) and Southeast Asia (63 %) are dominated by NH x deposition, owing to high local NH 3 emission, while the Middle East (25 %) and North Africa (34 %) are dominated by NO y deposition. Figure 7f shows the global distribution of the ratio of NH x deposition. While a high ratio is found in the Indian subcontinent, Southeast Asia, southeastern Brazil, southern Argentina and New Zealand (70-80 %), and Eastern Asia (∼ 60 %), other continental non-coastal regions are mainly dominated by NO y deposition. This is consistent with finding by ACCMIP . We compare the ratio of NH x deposition in 2010 (HTAP II) with that in 2001 (HTAP I) (Fig. S17). Generally, we find a 10 % worldwide decrease in the ratio of NH x deposition from 2001 to 2010. In particular, a 30 % decrease is found in southeastern China, mainly due to the large increase in NO x emission during the last decade. On the other hand, the ratio of NH x deposition in California was 15-20 % in 2001 and increased to 40-60 % in 2010. The ratio in Alaska also increased from 30-40 to 50 %. There is a generally 5-10 % increase over the eastern United States. This is consistent with an observed large increase of the NH x depositions and decrease of NO y depositions in the northeastern United States from 1990s to 2010s (Du et al., 2014;Li et al., 2016). A possible explanation is that the implementation of emission control strategies such as the Clean Air Act (CAA) has resulted in a large reduction in NO x emissions, which lowered the NO y deposition in the United States (Lloret and Valiela, 2016). This benefit is compensated for by increasing NH x deposition because no limitation is implemented on NH 3 emission (Kanakidou et al., 2016;Li et al., 2016). Some regions have small increases in the ratio of NH x deposition, such as northern Europe (Norway) (5 %), Southeast Asia (10 %) and western Australia (10 %).

Conclusions
We calculate the S and N deposition in 2010 using the multimodel mean (MMM) of an 11-model ensemble from the HTAP II. The model performance on wet deposition is evaluated with the measurement networks NADP over North America, EMEP over Europe and EANET over East Asia. The modeled wet deposition compares favorably with the observations. About 76-83 % of stations are predicted within ±50 % of observations. SO 2− 4 wet deposition is underestimated in East Asia by 20 %, especially at three Chinese stations with high Ca 2+ concentration. Because the locations of the Chinese stations do not cover the areas with highest deposition, it is hard to provide a comprehensive evaluation over this region. For NO − 3 wet deposition, a 20 % positive model bias is generally found at stations in the eastern United States, while some European (Poland, Norway and Spain) and East Asian (in Southeast Asia) stations with  high observed deposition are underestimated by about 60-70 %. NH + 4 wet deposition is underestimated in Europe (especially in Norway and Poland) and East Asia (especially in Russia and Korea). An intercomparison is conducted with the previous projects PhotoComp, ACCMIP and HTAP I. HTAP II has significantly improved the estimation of both S and N deposition at European stations compared to that in previous projects. Improved estimates are also found in East Asia. Modeled dry deposition is compared with the inferential data from CASTNET in North America. The MMM results are generally higher than the inferential data by 50-170 %, which is also reported in ACCMIP and HTAP I studies.
We calculate the S and N depositions on lands, in coastal zones and in open oceans. The global S deposition is 84 Tg(S) in 2010, with 49 % deposits in continental noncoastal regions, 32 % deposits in non-coastal oceans and 19 % deposits in coastal regions. The global N deposition is 113 Tg(N) in 2010, of which 59 Tg(N) is NO y deposition and 64 Tg(N) is NH x deposition. About 65 % of N is deposited The percentage of dry deposition in wet+dry deposition for NO y , NH x and N (NO y + NH x ) deposition. The ratio is calculated as (dry deposition) / (dry + wet deposition) × 100 %. (d, e, f) The percentage of NH x deposition in N (NO y + NH x ) deposition for wet, dry and wet+dry deposition. The ratio is calculated as (NH x deposition) / (NO y + NH x deposition).
in continental non-coastal regions, and 35 % in oceans (including 15 % in coastal regions). For continental regions, high S deposition is found in Asia regions (East Asia, South Asia and Southeast Asia), the United States Midwest, Central America and eastern Europe. For N deposition, high deposition is also identified in the above-mentioned regions plus sub-Saharan Africa and Brazil. For coastal regions, the east coast of Asia, all coasts of India and Malaysia, and the east coast of the Unites States are seen with relatively high S and N deposition. According to our estimation, about 4 Tg(S) of S deposition and 18 Tg(N) of N deposition are exported from land to ocean, including 0.3 Tg(S) and 4 Tg(N) in coastal regions.
We compare the HTAP II results in 2010 with HTAP I in 2001 by using the same land-ocean mask. The S deposition decreases 2 Tg(S) from 2001 to 2010 -with significant decreases in Europe (5 Tg(S)), North America (3 Tg(S)) and Russia (2 Tg(S)) -and increases in South Asia (2 Tg(S)) and the Middle East (1 Tg(S)). East Asia does not have large net changes in its S deposition due to increased S emission from 2001 to 2005 and a continuous reduction in S emission starting from 2006 owing to the SO 2 control policies in China's 11th FYP. The N deposition increases by 7 Tg(N). The increased N emissions from South Asia (5 Tg(N)), East Asia (4 Tg(N)) and Southeast Asia (2 Tg(N)) lead to identical amounts of elevation in deposition in corresponding regions. We also compare the ratio of NH x deposition in total N deposition between HTAP I and HTAP II. The ratio has increased in some regions of North America, especially in California (∼ 20 %), Alaska (∼ 10 %) and the eastern United States (5-10 %), which agrees well with recent observational and mod-eling studies in the United States. A small increase in the ratio of NH x deposition is found in northern Europe (Norway) (5 %), Southeast Asia (10 %) and western Australia (10 %). On the other hand, NO y deposition has started to dominate in East Asia (especially China) due to increased NO x emission in recent years.
This study updates our knowledge about the global S and N deposition in 2010. We find that the global distributions of S and N depositions have changed considerably during the last 10 years, with decreases in North America and Europe and increases in Asian regions. Further studies could determine how much these changes could affect the sourcereceptor relationship on deposition between continents and the impact of this relationship on global agriculture and ecosystems.
Data availability. The observation data are publicly available. The model data can be downloaded from the AeroCom database (http: //iek8wikis.iek.fz-juelich.de/HTAPWiki/FrontPage, last access: 6 April 2018) upon request.
Competing interests. The authors declare that they have no conflict of interest.
Special issue statement. This article is part of the special issue "Global and regional assessment of intercontinental transport of air pollution: results from HTAP, AQMEII and MICS". It is not associated with a conference.