The HTAP was created in 2004 under the Convention on Long-range Transboundary Air Pollution (CLRTAP). It involves efforts from international scientists aiming at understanding the hemispheric transport of air pollutants and its impact on regional and global air quality, public health and near-term climate change. Until now, two phases of HTAP experiments have been conducted successfully. HTAP I involved more than 20 models from international modeling groups, with 2001 as the base year for modeling studies. A comprehensive report of the major findings of HTAP I was released in 2010 and could be downloaded from http://www.htap.org/ (last access: 6 April 2018). The HTAP II was launched in 2012, with 2010 as the base simulation year. HTAP II required all models to use the same prescribed anthropogenic emissions instead of using the best estimates of emissions by each modeling group as HTAP I. This facilitates an inter-model comparison between models. In addition, HTAP II had a refined definition for the boundaries of regions, which enabled an update in the S–R relationships among regions.

This study uses the ensemble of 11 global models from HTAP II (including CAM-Chem, CHASER_re1, CHASER_t106, EMEP_rv48, GEMMACH, GEOS5, GEOSCHEMAJOINT, OsloCTM3v.2, GOCARTv5, SPRINTARS and C-IFS_v2). A detailed description of the experimental setup can be found in Galmarini et al. (2017). The S deposition includes SO2 deposition and SO42- deposition. The N deposition is categorized by NOy deposition and NHx deposition. The NOy deposition is a sum of all oxidized N except N2O, including the deposition of NO2, HNO3, NO3-, PAN and other organic nitrates other than PAN (Orgn). The NHx deposition includes NH3 deposition and NH4+ deposition. To form the multi-model ensemble, we regrid all models to a uniformed horizontal resolution of 0.1∘×0.1∘. We use the multi-model mean (MMM) of all models to present the ensemble results, a procedure which has been proven previously to have a better agreement with observations than single model result (Dentener et al., 2006; Tan et al., 2018). The MMM values of the compositions of S or N deposition are calculated separately and then combined to compute the total S or N deposition. More details can be found in Tan et al. (2018).

The base simulation uses anthropogenic emissions in 2010 (Janssens-Maenhout et al., 2015), which is called “base case” in this study. The MMM performance on wet deposition has been evaluated with observations from the National Atmospheric Deposition Program (NADP) (http://nadp.sws.uiuc.edu/, last access: 6 April 2018) for the NA region, European Monitoring and Evaluation Programme (EMEP) Chemical Coordinating Centre (CCC) reports (http://www.nilu.no/projects/ccc/reports.html, last access: 6 April 2018) for the EU region and the Acid Deposition Monitoring Network in East Asia (EANET; http://www.eanet.asia/, last access: 6 April 2018) for the EA region in the previous study of Tan et al. (2018). The following are some brief results of the evaluation. Modeled sum of gas-phase SO2 wet deposition and aerosol SO42- wet deposition is evaluated with observed SO42- wet deposition. About 76 % of the stations of all networks are predicted within ±50 % of observation. Negative model biases (-20 %) are found at some East Asian stations. Modeled sum of gas-phase HNO3 wet deposition and aerosol NO3- wet deposition is compared with observed NO3- wet deposition. About 83 % of the stations of all networks are predicted within ±50 % of observation. The European and Southeast Asian stations with high observed NO3- wet deposition are somewhat underestimated. Modeled sum of gas-phase NH3 wet deposition and aerosol NH4+ wet deposition is compared with observed NH4+ deposition. About 81 % of modeled NH4+ wet deposition at stations of all networks are predicted within ±50 % of observation. A general underestimation is found in modeled NH4+ wet deposition, especially at East Asian stations. In terms of dry deposition, due to the lack of directly measured data, we compare the modeled dry deposition with inferential data from the Clean Air Status and Trends Network (CASTNET) over US. The CASTNET data are calculated with observed aerosol concentration and modeled dry deposition velocity, therefore it might have high uncertainties. Comparison shows that the modeled dry deposition is generally higher than the CASTNET inferential data by a factor of 1–2. This is a common feature of many global and regional models (WMO, 2017). According to the analysis, the model bias for dry deposition mainly comes from the model over-prediction of air pollutant concentration (Tan et al., 2018). The CASTNET sites are generally located in remote regions with relatively lower air pollutant concentrations than urban regions, but the models fail to represent this characteristic with coarse spatial resolution (Tan et al., 2018).

HTAP II defines the boundaries of 17 regions as shown in Fig. 1. The emission perturbation experiments are conducted separately for six regions (regions with color in Fig. 1) with high priority: NA, EU, SA, EA, ME and RU. In the perturbation experiments, the anthropogenic emissions (including NOx, SO2, NH3, VOC, CO and PM) of a specific region are reduced by 20 % from the amounts in the base case simulation, while the emissions in other regions are kept constant. In addition, a global perturbation experiment referred to as “GLO” is conducted with 20 % reduction of global anthropogenic emissions. We estimate the impact of hemispheric transport on deposition by comparing the model results under perturbation experiments with those under the base case simulation. In order to validate the quality of model outputs, we check the mass balance between emissions and deposition at the global scale. The mass balance check for the base case simulation is shown in Tan et al. (2018), therefore we show the mass balance check for perturbation experiments in this study. We compare the global total amounts of changes in deposition (Δ Depo) with changes in emissions (Δ Emis) for all perturbation cases (Table S1). Model results are excluded if their global Δ Depo values fall outside the range of ±20 % of their global Δ Emis. According to our results, the amounts of Δ Depo are all within the scopes for all perturbation cases except the NHx deposition under the EA case. The Δ Depo of NHx deposition under EA case is not available due to lack of modeled NH4+ wet deposition under 20 % emission perturbation in EA.

This section studies the export of S and N emissions and oxidation products from the source regions. Table 1 shows the S–R relationship of S and N deposition among the six regions. The numbers are the sensitivity (SENr→s) of deposition in the receptor or source regions to emission changes in the source regions (Sanderson et al., 2008). The metric is calculated as Δ Depo in the receptor or source regions divided by Δ Emis in the source regions following Eq. (1). SENr→s=ΔDepo(r/s)ΔEmis(s)×100%, where s is the source region and r is the receptor region. Δ Depo (r/s) is the deposition change in the receptor or source regions and Δ Emis (s) is the emission change in the source regions. The metric indicates the percentages of emissions of the source regions that are deposited locally or exported to foreign regions.

The numbers outside of the parenthesis in Table 1 are for coastal and non-coastal regions together, and the numbers in the parenthesis are specifically for coastal regions (defined in Fig. 1). “Others” means other regions of the world other than the six regions (white color in Fig. 1). The NA region has 69 % of its S emissions deposited within itself, including 9 % deposited on its coastal region. The remaining 31 % is exported to the other regions, mostly to the “Others” and less than 3 % is deposited on the other five regions (EU, SA, EA, ME and RU). A relatively large fraction (14 %) of European S emissions are exported to the RU region. Other major pathways of export of S emissions and reaction products are from SA to EA (9 %), from EA to RU (5 %), and from RU to EU (7 %) and EA (5 %). ME has considerable high percentages of S emissions exported to its nearby regions such as SA (8 %), EA (5 %) and RU (5 %). The S–R relationship of NOy deposition is similar to that of S deposition, except that EU and ME have 66 % and 54 % of NOx emissions deposited within the source region, which are 6 % and 12 % higher than those of S emissions, likely due to somewhat longer lifetimes of S emissions compared to NOx emissions and the high emission altitude of S emissions. In terms of NHx deposition, about 20 % more NH3 emissions are deposited within the source regions (except ME) compared with S and NOx emissions, due to its short lifetime in the atmosphere.

The seasonal variations in the export of S and N emissions from source regions are shown in Fig. S1. In terms of S emissions, there are 5 %–10 % seasonal differences around the annual average export fractions for all regions except SA. SA exports almost half of its S emissions in spring, which is twice the number in summer (20 %) and fall (25 %), related to the specific dry period and monsoon circulation. The seasonal export fractions of NOx and NH3 emissions are similar to that of S emissions in the pattern, but generally lower in values in all seasons. Generally, the source regions export the highest percentage of their emissions in winter and spring and lowest in summer. More proficient oxidation chemistry in summer results in more soluble components, and local weather systems, especially the episodes of precipitation have a large influence on this seasonality. For most continental regions, the wet deposition accounts for 50 %–70 % of total deposition (Tan et al., 2018; Vet et al., 2014; Dentener et al., 2006). Therefore, local precipitation plays an important role in the removal process of local pollution. On the other hand, for regions with low local precipitation like ME, the percentage of emissions removed within own region would be lower than the other regions. In addition, the strong westerly winds in winter and spring favor the hemispheric transport in the mid-latitudes of the Northern Hemisphere. While the rapid vertical convections in summer slowdown the zonal transport of air flows and accelerates the local removal process.

A comparison is conducted with previous studies. Bey et al. (2001) estimated that 70 % of the emitted NOx from Asia is lost within its boundary by HNO3 deposition in spring. The estimations in our study are 70 % for EA and 61 % for SA, close to Bey's result. Li et al. (2004) reported that about 20 % of anthropogenic NOx emitted by NA is deposited out of its boundary (about 1000 km offshore). Stohl et al. (2002) calculated that 9 %–22 % of surface NOx emissions are exported out of the boundary of NA. Our estimation is about 30 %, higher than Li's and Stohl's results. The HTAP I study by Sanderson et al. (2008) developed a S–R relationship for NOy deposition among NA, EU, SA and EA. The results showed that about 12 %–24 % of the emitted NOx is deposited out of the source regions. This study of HTAP II finds a higher percentage of export (26 %–34 %). It should be noted that the estimations of different studies are not fully comparable since they are influenced by several factors: (1) definition of the boundaries of the source and receptor regions. For instance, Li et al. (2004) defined the boundary of NA by a squared domain: 25–55∘ N, 65–130∘ W, while we use the continental boundaries defined by HTAP II. There are also changes in the boundaries of regions from HTAP I to HTAP II. For instance, Mexico and Central America are included in NA in HTAP I, but they are defined as a separate region in HTAP II (region 12 in Fig. 1). The boundary of EU is also changed in HTAP II. (2) Design of the perturbation simulations. The perturbation simulations in HTAP I only change the NOx emissions, but HTAP II simulations also reduce the other anthropogenic emissions, including SO2 and PM. The joint effects of controlling multiple species may result in more reduction in NOy deposition, but it is hard to estimate the influence in this study.

This section investigates the impact of hemispheric transport on deposition in the receptor regions. Figure 2 shows the annual response of S deposition to 20 % emission reduction in the source regions calculated as Eq. (2). Response=ΔDepoperturbationDepobase×100%, where Δ Depo (perturbation) is the Δ Depo between perturbation case and base case. Depo (base) is the deposition under base case. The negative values mean that the deposition decreases with reduction in emissions. Table S2 summarizes the regional median deposition fluxes under base case and emission perturbation cases. Figure 2a shows the global response of S deposition to 20 % emission reduction in NA. The largest deposition change is found in the source region (NA), with a 4 %–20 % decrease in S deposition in the non-coastal region and 14 %–16 % decrease around the east coast. The impact on the North Atlantic Ocean declines gradually from the near-coastal region (12 %–14 %) to the open ocean (2 %–12 %) and Eurasia (<1 %). Figure 2b shows the global response of S deposition to 20 % emission reduction in EU. Although the impact on continental non-coastal regions is also high (6 %–18 %), the impact on the coastal regions is generally less than 6 %, much lower than NA's impact on its east coast (14 %–16 %). The deposition in North Africa, central Asia and western RU is affected by 2 %–6 %. The 20 % emission reduction in SA (Fig. 2c) shows large influences over its southwest coast and the Arabian Sea (4 %–12 %). The SA's outflow affects the deposition in southeastern ME and eastern Sub-Saharan Africa by 1 %–4 % and western EA and Southeast Asia (mainly Bangladesh) by 2 %–6 %. Figure 2d shows the impact on S deposition from 20 % emission reduction in EA. On one hand, the impact is strong on the east coast of China (12 %–16 %) and decreases gradually over the North Pacific Ocean (4 %–10 %). Although the majority of S emissions are deposited on the North Pacific Ocean, the influence on western NA can still reach 4 %–6 %. On the other hand, the impacts on Southeast Asia and SA are much lower (2 %–5 % and <1 %), due to the block of air flows by the Himalaya mountains (Fig. S4). The 20 % emission reduction in ME mainly affects the S deposition in Africa by 2 %–10 % and western EA by 2 %–4 %. Figure 2f shows the change of S deposition with 20 % emission reduction in RU. The regions of impact are mainly located at high latitudes of the Northern Hemisphere, including northern EU (2 %–6 %) and western Arctic Circle (1 %–4 %). The Russian flow enters the Arctic in the lower troposphere in the winter season (Stohl, 2006).

Figure 3 shows the impacts of reducing NOx emissions on NOy deposition. The overall impact is qualitatively similar to that of S emissions in the spatial pattern, with some differences in the values. Some regions receive lower impact on NOy deposition than S deposition. For instance, SA's impact on eastern Africa is about 1 %–4 % on S deposition, but is <1 % on NOy deposition. ME's impact on the western Africa and Gulf of Guinea is about 2 %–4 % on S deposition, but is <1 % on NOy deposition. These smaller sensitivities reflect differences in lifetimes, and the lower formation of aerosol NO3- under warm conditions in tropical regions. Under the NA perturbation case (Fig. 3a), 8 %–12 % changes of NOy deposition are found on the west coast of California, due to high NOx emissions in California from mobile sources, which is not seen in S deposition. The impacts of emission reduction in EU and EA on their coastal regions are generally 2 %–4 % higher for NOy deposition than S deposition (Fig. 3b and d). The impact on NHx deposition is similar to that on NOy deposition (Fig. S2). It should be noted that this is the result of 20 % emission reduction in the source regions, therefore the actual impact (100 % emission reduction) could be 5 times higher when assuming a linear relationship between 20 % and 100 % emission reduction on deposition.

We quantify the amount of deposition carried by hemispheric transport and study its seasonality. Figure 4 shows the monthly changes in S deposition for 20 % emission reductions in the source regions. The values are meridional sum with a west–east resolution of 0.1∘, and display well the locations of the source regions. The negative values indicate the amounts of pollutants transported from the source regions to the receptor regions. According to Fig. 4a, NA has about (1-10)×104 kg (S) month-1 per 0.1∘ longitude of its S emissions transported and deposited over the North Atlantic Ocean (15–75∘ W) throughout the whole year. We also find about (1-3)×104 kg (S) month-1 per 0.1∘ longitude decrease in S deposition at about 90 and 120∘ E in spring and fall, which gives evidence to NA's influence on Eurasia via the transatlantic flow, although this amount accounts for less than 1 % of local S deposition (white in Eurasia in Fig. 2a). Figure 4b shows that about (1-3)×104 kg (S) month-1 per 0.1∘ longitude of EU's emissions are transported and deposited at 30–60∘ E in RU throughout the whole year and at 100–120∘ E in EA in spring and fall. According to Fig. 4c, SA exports its S emissions to 30–60∘ E in ME and eastern Africa in early spring and to 90–180∘ E in EA and the North Pacific Ocean from late spring until fall. In particular, the influence on EA can reach (5-10)×104 kg (S) month-1 per 0.1∘ longitude in mid-spring. According to Fig. 4d, EA's S emissions are widely transported and deposited over the North Pacific Ocean throughout the whole year. The Asian outflow arrives at the west coast of NA (∼130∘ W) in all seasons except summer, but only reaches far in western NA (∼90∘ W) in spring and brings about 1×104 kg (S) month-1 per 0.1∘ longitude of S deposition. The export to SA is only found during the Asia winter monsoon. Figure 5 shows the monthly changes in NOy deposition with perturbation experiments. Compared to S deposition, the change in NOy deposition by hemispheric transport is generally smaller. For instance, the NA's impact on Eurasia is (1-3)×104 kg (S) month-1 per 0.1∘ longitude for S deposition, but is less than 0.5×104 kg (N) month-1 per 0.1∘ longitude for NOy deposition. The SA's impact on EA (90–120∘ E) can reach (5-10)×104 kg (S) month-1 per 0.1∘ longitude for S deposition, but the amount is 4 times lower for NOy deposition. This result is in accordance with the S–R results in Sect. 3.1 that more S emissions are transported out of the source regions than N emissions, probably due to longer chemical lifetimes and higher emission altitudes. Patterns similar to NOy deposition are found in the monthly changes in NHx deposition (Fig. S3).

The deposition changes due to transport between neighboring regions are found throughout the whole year and are slightly stronger in winter, such as between EU and RU (∼30∘ E; Fig. 4b and f) and from EA to the North Pacific Ocean (∼130∘ E; Fig. 4d). This is consistent with the seasonality we found for the export of emissions from source regions in Sect. 3.1. In addition, most source regions reduce more S and NOx emissions in winter than the other seasons (Table S3), thus more emissions are exported abroad in this season. On the contrary, the deposition change by transport over long distance mainly occurs in spring and fall, especially the hemispheric transport from NA to EU, from EU to EA and from EA to NA. The seasonality of long-range transport for NA, EU and EA well fits the characteristic of westerlies, which is the prevailing winds in the mid-latitude of the Northern Hemisphere. This agrees with the seasonality of the transpacific, transatlantic and trans-Eurasia flows of air pollutants that spring is the most efficient season for long-range transport in the mid-latitudes (Holzer et al., 2005; Liu et al., 2005, 2003; Liang et al., 2004; Brown-Steiner and Hess, 2011; Li et al., 2014; Auvray and Bey, 2005; Wild et al., 2004). Although the westerlies are also strong in winter, the concentrations of air pollutants in the air flows are lower than spring. The formations of secondary species like PAN are suppressed by slow oxidation in the cold environment (Berntsen et al., 1999; Deolal et al., 2013; Moxim et al., 1996), which plays an important role as a reservoir for NOx in the long-range transport of air plumes (Lin et al., 2010; Hudman et al., 2004).

This section compares the contributions of hemispheric transport and own-region emission control on local deposition. A metric called RERER is calculated by dividing the Δ Depo due to foreign emission reduction by Δ Depo due to global (foreign + own region) emission control following Eq. (3). RERERi=ΔDepoi(foreign)ΔDepoi(global), where i is the region of focus. Δ Depoi (foreign) is the Δ Depo in region i due to 20 % foreign emission reduction. It is calculated by subtracting the Δ Depo due to 20 % own-region emission control from Δ Depo due to 20 % global emission reduction. Δ Depoi (global) is the Δ Depo in region i due to 20 % global emission reduction. This metric indicates the importance of foreign or local emissions on local deposition. A low RERER value (close to 0) indicates a predominant effect of own-region emissions on local deposition, while a high RERER value (close to 1) means strong impacts from hemispheric transport.

Table 2 shows the RERER values for total (include both non-coastal and coastal regions), non-coastal and coastal regions. For both non-coastal and coastal regions, the own-region impact includes control of both its coastal and non-coastal regions, and the foreign impact comes from emission reduction of both foreign coastal and foreign non-coastal regions. As we expected, NA (0.07, 0.17 and 0.17), SA (0.04, 0.17 and 0.18) and EA (0.16 and 0.16) regions have relatively low RERER values, due to large local emissions compared to the foreign contributions. EU (0.12, 0.34 and 0.36) and ME (0.32, 0.32 and 0.42) have relatively higher RERER values. RU is the only region with all RERER values (0.55, 0.59 and 0.61) higher than 0.5, which means its deposition is almost equally sensitive to the foreign exported air pollution and own-region control. The RERER values of NOy deposition are of similar magnitudes to those of S deposition, while the RERER values of NHx deposition are 0.1 lower than those two, probably due to the lack of data for the EA perturbation case, so that the contributions from EA are not included.

The RERER values of coastal regions are generally 0.1–0.3 higher than those of non-coastal regions. Even regions with low non-coastal RERER values such as NA and SA have high RERER values in coastal regions. For instance, the RERER values of NA reaches 0.3–0.4 for its coastal region, more than double of the RERER values in its non-coastal regions (0.05–0.12). Coastal regions receive high proportions of deposition from the hemispheric transport. Except large-scale circulation like prevailing westerlies, the coastal regions are featured with complex small-scale circulations. For instance, the low-level jet (zonal winds with high speed) contributes to rainfall in coastal regions in Asia (Xavier et al., 2018). The orographic effects enhance the precipitation over coastal mountain regions such as the west coast of NA and EU, and the southeast coast of RU (James and Houze, 2005). According to Table 1, EA exports 5 % of its S and N emissions to RU, almost half of which are deposited on RU's coastal regions. RU exports 7 %–12 % of S and N emissions to EU, 30 % of which are deposited on EU's coastal regions. The impact of hemispheric transport is identical or even larger than the effect of controlling own-region emissions for some coastal or near-coastal regions. According to Fig. 2, 20 % emission reduction in EA can reduce 2 %–6 % of S deposition in the west coast of NA. This effect is even larger than 20 % emission reduction in its own region (<1 %). Similarly, 20 % emission reduction in NA can change 2 %–5 % of S deposition in the west coast of EU, which is almost identical to the effect of 20 % emission control in EU. On one hand, the emissions in western NA and western EU are relatively low, thus the effect of own-region control is not significant. On the other hand, these coastal regions are in the downwind directions of eastern EA and eastern NA, which are the main source regions of S and N emissions. Coastal regions serve as transit places for air–sea exchanges with vulnerable ecosystems (Jickells, 2006; Jickells et al., 2017). The over-richness of deposition in coastal water and ecosystems can evoke a number of environmental issues, of which some are specifically for coastal regions such as threats to coastal benthic and planktonic system and sustainability of fishery (Paerl, 2002; Doney et al., 2007).

Figure 6 shows the source concentrations of deposition from hemispheric transport and own-region emissions. Other (OTH, pattern fill in the figure) is calculated as Δ Depo(GLO)-∑-Δ Depo(case) (case = 6, including NA, EU, SA, EA, ME and RU). It indicates the deposition change due to other reasons than the total effects of separate emission reduction in the six regions. It could come from the emission reduction in the rest of the world, especially nearby regions such as from Central Asia and North Africa to EU and ME. It could also come from the joint effects of emission control in multiple source regions, which possibly change the oxidant chemistry, atmospheric mixing and lifetimes of reactive pollutants. However, the model simulations do not allow us to separate these contributions in this study. For regions with low RERER values (NA, SA and EA), the own-region emissions dominate the local deposition by more than 80 %. For these regions, determined by vicinity and transport patterns, the foreign impact is somewhat dominated by certain source regions, such as from EA to NA (2 %–4 % out of 4 %–5 %), from ME to SA (5 %–6 % out of 7 %–11 %) and from SA to EA (3 %–4 % out of 4 %–7 %). For regions with higher RERER values (i.e., EU and ME), there is an about 20 % contribution from “OTH”. Besides this, RU contributes 4 %–5 % to EU's deposition and EU contributes 5 % to ME's deposition. The high RERER region (i.e., RU) has a different pattern than the other regions. The contributions of hemispheric transport from the other five regions (23 %–45 %) are almost equivalent to its own-region emission control (39 %–45 %), with significant impacts from EA (20 %–24 %) and EU (13 %–15 %).

Figure 7 shows the inter-model variations on simulating the Δ Depo of S, NOy and NHx deposition under emission perturbation cases, separately for wet and dry deposition. The values are global integrated changes in the components of deposition between base case and perturbation experiments from MMM results, with error bars showing the maximum and minimum values of models. The figure only shows the main compositions of S and N deposition, which together account for more than 95 % of total deposition. In terms of S deposition (Fig. 7a), the modeled Δ Depo by multiple models (defined as maximum value of multi-model minus minimum value of multi-model) ranges from 0.06 to 0.23 Tg(N) yr-1 and from 0.01 to 0.22 Tg(N) yr-1 for SO2 dry and wet deposition, respectively, and from 0.01 to 0.03 Tg(N) yr-1 and from 0.009 to 0.17 Tg(N) yr-1 for SO42- dry and wet deposition, respectively (ranges are for different cases). High uncertainty is found in the EA perturbation case, where the model divergences are mainly from SO2 wet and dry deposition and SO42- wet deposition. In terms of NOy deposition (Fig. 7b), the differences among models range from 0.003 to 0.07 Tg(N) yr-1 for NO2 dry deposition, and from 0.07 to 0.55 Tg(N) yr-1 and from 0.03 to 0.75 Tg(N) yr-1 for NO3- dry and wet deposition, respectively. The EA perturbation case also has the largest inter-model variation, with high uncertainties in simulating both the NO3- wet and dry deposition. In terms of NHx deposition (Fig. 7c), the differences among models range from 0.04 to 0.09 Tg(N) yr-1 for NH3 dry deposition, and from 0.008 to 0.15 Tg(N) yr-1 and from 0.002 to 0.11 Tg(N) yr-1 for NH4+ dry and wet deposition, respectively. Both EA and SA perturbation cases have relatively high uncertainties in NH4+ dry deposition. Overall, the inter-model variations are considerably high under emission perturbation in Asian regions, especially EA. On one hand, the EA perturbation case assumes the largest amounts of emission reduction among all perturbation cases (Table S3). On the other hand, model evaluation (Tan et al., 2018) reported high model bias in simulating the deposition in this region, and suggest an incomplete knowledge from the combined picture provided by observations and models.