To evaluate temporal variations in nitrogen wet deposition across Japan during 1989–2008, we analyzed results of a chemical transport model (the Models-3 Community Multiscale Air Quality) and observational data. The model successfully reproduced the general patterns of spatial and temporal variations of observed NO3− wet deposition rates. Wet deposition rates of NO3− across Japan increased during 1989–2008, with rates of increase of 2–5%/yr. Sensitivity simulations indicated that the increase of NO3− wet deposition rates was mostly (61%–94%) explained by the increased emissions of atmospheric pollutants in China. Contributions of China's emissions increased from 29%–35% during 1989–1993 to 43%–61% during 2004–2008, suggesting that transboundary pollution had a large impact on NO3− wet deposition in Japan. The contribution of observed NO3− to total nitrogen wet deposition (i.e., NO3− + NH4+) increased in southwestern Japan, and currently, NO3− and NH4+ make similar contributions to nitrogen wet deposition across Japan. Interannual variation of NO3− wet deposition was further evaluated using a meteorological index, area-weighted surface pressure anomaly (ASPA). When ASPA was negative, air masses from the Asian continent were more directly transported to Japan, and NO3− concentrations across Japan became high. Thus, anomalies of NO3− concentrations were negatively correlated with ASPA. Anomalies of NO3− wet deposition rates, however, showed a weak positive correlation with ASPA, reflecting a positive correlation between anomalies in precipitation rates and ASPA. This result strongly suggests that precipitation patterns have a large impact on the interannual variation of NO3− wet deposition across Japan.
 The deposition of atmospheric nitrogen has large influences on terrestrial and aquatic ecosystems. For example, an increase of nitrogen deposition has been reported to reduce plant species richness significantly [Stevens et al., 2004] and to increase carbon sinks in terrestrial ecosystems [Reay et al., 2008]. In addition, when atmospheric nitrogen deposition exceeds biological demand (nitrogen saturation), then incoming nitrogen cannot be retained in an ecosystem and leaches into water runoff [Aber et al., 1989; Fisher et al., 2007; Wright et al., 1995]. Nitrogen outputs from terrestrial ecosystems may have negative effects on aquatic ecosystems downstream, including acidification of surface waters and eutrophication of nitrogen-limited streams, lakes, and coastal marine waters [Hinga et al., 1991].
 Because of the importance of atmospheric nitrogen deposition, its deposition rates have been observed by national and international networks across Europe (European Monitoring and Evaluation Programme), North America (North American Deposition Program), and Asia (the Acid Deposition Monitoring Network in East Asia, EANET). These monitoring data have indicated that atmospheric nitrogen deposition decreased in Europe and North America during the 1980s and 1990s because of decreased emission rates of nitrogen oxides (NOx) [Fowler et al., 2007; Sickles and Shadwick, 2007]. In Asia, however, no monitoring program existed for the entire continent during the 1980s and 1990s; the EANET program started monitoring atmospheric deposition in 2000. In 1983, Japan's Environment Agency started a nationwide monitoring network, the Japanese Acid Deposition Survey (JADS), which revealed increasing trends of nitrogen wet deposition from 1989 to 1998 [Seto et al., 2004]. Because the emissions of NOx and SO2 across eastern Asia have increased markedly since 2000 [Ohara et al., 2007] (Figure 1), the trends in atmospheric deposition after 2000 should be evaluated.
 Global and regional chemical transport models (CTMs) are useful tools for evaluating the factors controlling nitrogen deposition. Using 23 global CTMs, Dentener et al.  showed that the United States, Europe, and Asia are the regions most affected by nitrogen deposition. Lamarque et al.  conducted 29 simulations using 6 global CTMs and showed that 60%–70% of emitted NOx is eventually deposited over land, regardless of the amount of nitrogen emissions used in the simulations, and that increased nitrogen emissions proportionally enhance nitrogen deposition. According to an analysis by Sanderson et al.  using 15 global CTMs, 12%–24% of NOx is transported outside of each region (Europe, North America, southern Asia, and eastern Asia), meaning that total oxidized reaction nitrogen (NOy) emitted in each region remains in the region by 76% (Europe) to 88% (eastern Asia). Regional CTMs also have been used to quantify atmospheric nitrogen deposition in Europe [Fagerli and Aas, 2008; Simpson et al., 2006], North America [Zhang et al., 2009], and Asia [Lin et al., 2008a; Wang et al., 2008]. However, to our knowledge, few modeling studies have evaluated the factors controlling interannual trends in nitrogen wet deposition rates [Fagerli and Aas, 2008]. Asia is one of the largest emission sources of reactive nitrogen and the only region experiencing rapid increases in nitrogen emissions [Akimoto, 2003; Ohara et al., 2007]. It is important to evaluate how these increases affect nitrogen deposition rates.
 In this study, temporal variations of nitrogen wet deposition during 1989–2008 were assessed using observational data and a CTM. First, the simulation results were verified by comparison with observational data (section 3.1). The factors controlling temporal and spatial variations of nitrogen deposition were then evaluated (section 3.2). The effect of increasing nitrogen emissions in China on nitrogen deposition fields was also quantified. Because China is the largest emerging country in East Asia, this effect is important in the evaluation of air quality in East Asia. In addition, China's contribution to the deposition field was evaluated by previous studies [Holloway et al., 2002; Lin et al., 2008a], and comparison of these values obtained in this study with those reported in previous studies was helpful in the model validation. Finally, interannual variation of nitrogen deposition in spring was analyzed from a meteorological viewpoint (section 3.3). Dry deposition of nitrogen is not the focus of this study because long-term observational data are not available for nitrogen dry deposition rates. However, the dry deposition of nitrogen appears to be comparable to wet deposition [Ministry of Environment, 2009; Wang et al., 2008], so the contribution of dry deposition should be evaluated in future studies.
2.1. Model Description
 We calculated atmospheric concentrations of trace gaseous and aerosol species, as well as their deposition rates, using the Models-3 Community Multiscale Air Quality (CMAQ) version 4.4 modeling system developed by the U.S. Environmental Protection Agency [Byun and Schere, 2006]. Meteorological fields were calculated using the Regional Atmospheric Modeling System (RAMS) version 4.4 [Pielke et al., 1992]. The RAMS hourly output files were processed for the CMAQ input files using the Meteorology-Chemistry Interface Processor [Byun and Schere, 2006]. In the RAMS simulation, the analysis nudging was conducted using the three-dimensional meteorological fields from the National Centers for Environmental Prediction Reanalysis 1 data sets available with 2.5° × 2.5° horizontal resolution for every 6 h.
 The CMAQ was configured with the chemical mechanism of the Statewide Air Pollution Research Center gas phase mechanism (SAPRC-99) of Carter . Aerosol modules follow Binkowski and Roselle , whereas inorganic thermodynamics modules are based on ISORROPIA [Nenes et al., 1998]. The monthly averaged lateral boundary conditions for most chemical tracers were obtained from the global chemical transport model CHASER (Chemical AGCM for Study of Atmospheric Environment and Radiative Forcing; [Sudo et al., 2002]) simulations for the 1990s. Emission data from anthropogenic and natural sources were those reported by Ohara et al. . This emission inventory, Regional Emission Inventory in Asia (REAS) version 1.1, covers NOx, SO2, CO, NH3, volatile organic compounds, and aerosol species over a large part of Asia with about 0.5° × 0.5° horizontal resolution. Seasonal and diurnal variations are not considered in the REAS database. The model domain (Figure 2) covers a 6240 km × 5440 km area on a rotated polar stereographic map projection centered at 25°N, 115°E, with a 80 km × 80 km grid resolution and a 14-layer vertical structure (set on a terrain following the σz coordinate). This model system was used previously for the analysis of long-term trends in O3 concentrations [Kurokawa et al., 2009; Ohara et al., 2008; Tanimoto et al., 2009], fine particle concentrations [Osada et al., 2009], NO3− deposition rates [Uno et al., 2007], and SO42− deposition rates [Katayama et al., 2008].
 We performed four sets of 20 year simulations for the period 1989–2008 as shown in Table 1. The EyyMyy simulation is the baseline simulation, and we estimated the effect of interannual variations of emissions on variations of nitrogen deposition rates based on the difference between EyyMyy and E00Myy simulations. In addition, the effects of meteorological conditions on interannual variations of nitrogen deposition rates were estimated from the difference between EyyMyy and EyyM00 simulations. In the E*yyMyy simulation, emissions of pollutants from China were excluded from the EyyMyy simulation. Thus, the difference between EyyMyy and E*yyMyy simulations provided information on the effect of China's emissions on nitrogen deposition rates across Japan. The results of the sensitivity analyses are given in section 3.2. The inventory for 2006 was tentatively used for the simulations in 2007 and 2008 owing to the unavailability of statistical data. This assumption might have caused an error of a few percent in the analysis of the interannual trend. Simulation of RAMS was conducted continuously for the whole study period with a 1 month spin-up period. The CMAQ simulation was performed from 1 January to 31 December in each year, and 1–5 January of each year was treated as the spin-up period and not included in this analysis.
Table 1. Settings of the Four Model Simulations Conducted in This Study
Year of 2000
Year of 2000
Each year with China emissions off
 For the evaluation of long-term trends in nitrogen wet deposition across Japan, we used a nationwide monitoring database (JADS) that was started by Japan's Environment Agency in 1983 and is currently maintained by Japan's Ministry of the Environment. Although several monitoring techniques have been modified during its history, the measurements offer a reasonable database for trend analysis of wet deposition on a national scale [Seto et al., 2004]. The monitoring techniques of JADS are reported in the survey manuals [Ministry of Environment, 2001, 2009]. Samples were collected with automated wet-only samplers on different temporal bases (i.e., daily, weekly, and biweekly), depending upon the operational phases, and precipitation amounts were measured at the same sites with approved rain gauges. From 1998, however, samples were collected daily, although weekly sampling continued at a few sites. The samples were subjected to chemical analysis, including SO42−, NO3−, and NH4+ ions, and the quality of measurements was controlled with ion balance and conductivity checks. In the present study, all data were compiled into monthly deposition data sets for further analysis.
 The measurement sites were classified into three categories, urban, rural and remote, based on the site and its surroundings at different spatial scales in consideration of local sources of atmospheric pollutants, potential obstacles and contaminations. Observational data obtained at urban sites were not used in this study, because the horizontal distribution of the model is 80 km and the model cannot resolve urban-scale air pollution. All monitoring sites were also grouped into three regions of the country: northeastern Japan (NEJ), approximately coinciding with the Hokkaido and Tohoku regions; central Japan (CJ), representing the Kanto, Chubu, and Kansai regions; and southwestern Japan (SWJ), corresponding to the Chugoku, Shikoku, and Kyushu regions (Figure 2). For the analysis, 20 monitoring sites on the four main islands of Japan were selected based on data quality, particularly data completeness (Table 2).
Site numbers correspond to the monitoring sites shown in Figure 2 (right).
NEJ, northeastern Japan; CJ, central Japan; SWJ, southwestern Japan.
 We also used data gathered at 28 monitoring sites in eight countries in northern Asia (Mongolia and Russia), eastern Asia (China, Japan, and South Korea), and Southeast Asia (the Philippines, Thailand, and Vietnam) by the EANET program [Acid Deposition Monitoring Network in East Asia, 2007] for model validation of wet deposition rates and precipitation rates across Asia (Figure 2 and Table 3). Measurement details are available in a technical report on the program [Acid Deposition and Oxidants Research Center, 2001]. Samples were collected by wet-only samplers on a daily or weekly basis, but all data were compiled into monthly deposition data sets for this analysis. For most sites, concentrations of SO42−, NO3−, and NH4+ were determined by ion chromatography, but the NH4+ concentration was obtained by spectrophotometry at some sites in China, the Philippines, Russia, and Vietnam. The EANET sites are also classified as urban, rural, or remote. In the present analysis, however, only data from rural and remote sites were employed because the research focuses on nitrogen deposition at the regional scale.
Site numbers correspond to the monitoring sites shown in Figure 2 (left).
3. Results and Discussion
3.1. Evaluation of Model Performance Using Observational Data
3.1.1. Comparison With EANET Data Across Asia
 To evaluate the accuracy of model simulations of nitrogen wet deposition and precipitation rates across Asia, results for the baseline simulation (i.e., EyyMyy) were compared with observational data gathered during 2000–2007 from 28 EANET monitoring sites in northeastern, eastern, and Southeast Asia. The model calculated annual wet deposition and precipitation rates and reproduced the observations well: 55% (NO3−), 61% (NH4+), 77% (SO42−), and 76% (precipitation) of observed data points were reproduced within a factor of two by the model. However, the model largely underestimated NO3− and SO42− wet deposition rates at some sites in inland China close to Chongqing (Figure 3). A similar underestimation by CTMs for monitoring sites in inland China was previously reported [Dentener et al., 2006; Lin et al., 2008b; Wang et al., 2008]. Lin et al. [2008b] suggested that underestimation of NOx and SO2 emission rates around these areas could be one reason for the underestimation of NO3− and SO42− wet deposition rates. This explanation might be applicable to the present case. By contrast, simulated NO3− and SO42− wet deposition rates in coastal China, South Korea, and Japan were mostly within a factor of two of the observed values. Ammonium wet deposition rates were reproduced well in coastal China and overestimated in South Korea and Japan, which is discussed in detail in section 3.1.2.
 Wet deposition rates of NO3−, NH4+, and SO42− in eastern Asian countries (i.e., Japan, China, and South Korea) were obviously higher than those in northeastern Asia and were comparable to those in Southeast Asia. The model simulated well these spatial patterns in the wet deposition rates. In northeastern Asia, NO3−, NH4+, and SO42− wet deposition rates were generally underestimated, and similar model behavior was reported by Wang et al. . Uncertainties in emissions and boundary conditions were possible reasons for this underestimation. In addition, emissions in Russia, which are not considered in REAS version 1.1, were not included in the simulations of this study, which might be another reason for this underestimation. Considering that NOx and SO2 emissions in Russia have a negligible impact on sulfur and nitrate depositions in Japan [Lin et al., 2008a], however, neglecting Russia's emissions is not expected to change the conclusions of this study.
 The observed precipitation rates were highest in Southeast Asia and lowest in northwestern Asia, and this behavior was reproduced well by the model. Overall, the model captured well the precipitation and deposition patterns in northern, eastern, and Southeast Asia.
3.1.2. Comparison With Monitoring Data in Japan
 Model validation for Japan was also conducted in more detail. The model results of 5 year averages of wet deposition rates and precipitation rates during 1989–2008 were compared with data gathered at the 20 monitoring sites in the JADS network (Figure 4). The model reproduced 85%–95% of the observed NO3− wet deposition rates at the 20 sites within a factor of two of the observed values. Sulfate wet deposition and precipitation rates were also underestimated at the same sites in SWJ, suggesting that the underestimation of precipitation rates caused the underestimation of NO3− and SO42− wet deposition rates. Precipitation rates were more accurately reproduced by the model in NEJ and CJ, and wet deposition rates of NO3− and SO42− were relatively accurately reproduced as well. Validity of the predicted meteorological parameters in the same model system were evaluated by Yoshida et al. , who showed that the spatial variation of precipitation rates predicted by RAMS agreed well with the analysis data of the precipitation field (CPC Merged Analysis of Precipitation [Xie and Arkin, 1997]), although observed precipitation rates across Japan were underestimated by 20%–40%.
 Ammonium wet deposition rates were generally overestimated, particularly in NEJ and CJ, whereas precipitation rates were underestimated. One explanation for the NH4+ overestimation is the differences in ion balance between the observed and simulated results. According to observational data, the NH4+/[NO3− + SO42−] equivalent ratio of wet deposition rates was 0.2–0.6 across the four seasons (Figure A1). The model simulated a similar NH4+/[NO3− + SO42−] ratio as was observed in summer and autumn. However, the simulated NH4+/[NO3− + SO42−] ratio was higher than the observations in winter and spring, indicating that the observed ion balance was not captured by the model. Possible explanations for this discrepancy were the lack of mineral cations in the model and uncertainties in NH3 emission rates, as discussed in Appendix A. Simulated NH4+ wet deposition rates, which were modified using the [NH4+]/[NO3−+ SO42−] ratio in each season, are added in Figures 4–6 for reference. The modified NH4+ wet deposition rates showed consistent behavior with wet deposition rates of NO3− and SO42− and precipitation rates, which were reproduced relatively well in NEJ and CJ and underestimated in SWJ. The model performance for NH4+ wet deposition rates will be improved if the model overestimation of the NH4+/[NO3− + SO42−] ratio can be resolved, and this problem should be investigated in future studies.
3.2. Trends in Nitrogen Wet Deposition Rates During 1989–2008
 We compared the temporal trends in nitrogen wet deposition rates between observational data and model results for the baseline and sensitivity simulations.
3.2.1. NO3− Wet Deposition
Figure 5 shows the wet deposition rates of NO3− and NH4+ and precipitation rates averaged over the three regions of NEJ, CJ, and SWJ. Data obtained at 8, 14, and 20 stations were available for the periods 1989–2008, 1994–2008, and 1999–2008, respectively (Table 2). Both observed and simulated NO3− wet deposition rates clearly showed a continuous increase between 1989 and 2008 (Figure 5).
 To evaluate the factors controlling the interannual trends in nitrogen wet deposition rates, we conducted sensitivity simulations as summarized in Table 1. We assessed the rates of increase of NO3− wet deposition rates using linear regression analysis (Figure 6). Because data from only 8 stations were available for the 20 year period (1989–2008), we analyzed the 15 year trend (1994–2008) by employing the largest and longest data sets for 14 stations. Linear fitting revealed significant increases (p < 10−5) in SO2, NOx, and NH3 emissions from China from 1989 to 2005 (Figure 1). Linear regression analysis can also be used to detect whether the emission increase in China resulted in large contributions to the deposition field in Japan. The observed increasing rates of NO3− wet deposition over the 15 year period agreed with those over the 10 year and 20 year periods within 30%–60%, except for the 20 year data in NEJ (where data from only one site were available).
 In spring, when transboundary pollution across Japan is the most pronounced [Tanimoto et al., 2005], both observational data and simulation results showed similar rates of increase of NO3− wet deposition (∼4%/yr). The simulated increasing trend of NO3− wet deposition rates was mostly explained by the emission increase as shown by the comparison between EyyMyy and E00Myy results (Figure 6). Also, comparison between the EyyMyy and E*yyMyy simulations indicated that the increase in NO3− wet deposition rates was largely explained (by 80%–85%) by the increase of emissions in China.
 The observed rates of increase of annual average NO3− wet deposition rates ranged from 2%/yr to 5%/yr in NEJ, CJ, and SWJ, and were reproduced by the model within a factor of two (∼4%/yr) (Figure 6). The simulated rates of increase of NO3− wet deposition between 1994 and 2008 were similar to those between 1989 and 2008, suggesting that the differences in the analytical period did not yield large errors. The simulated increasing trend of annual average NO3− wet deposition rates was largely explained (by 61%–94%) by the emission increase in China.
 Emissions from China accounted for 29%–35% of NO3− wet deposition across Japan during 1989–1993 and for 43%–61% during 2004–2008 (Table 4). These results indicate that emissions from China made large contributions to the increase in NO3− wet deposition rates in Japan. Lin et al. [2008a] estimated that China's emissions contributed 21% of NO3− deposition in Japan in 2001, which is lower than the estimate in this study (31%–49% in 1999–2003). There are three possible reasons for this discrepancy.
Table 4. Simulated and Observed NO3− Wet Deposition Rates in Southwestern Japan, Central Japan, and Northeastern Japan Averaged Over 5 Year Intervalsa
Annual NO3− Wet Deposition Rate (g N m−2)
China's Contribution (Model)
NO3−/[NO3− + NH4+] Ratio
The contribution of China's emissions to the NO3− wet deposition rates and ratios of NO3− to total nitrogen (i.e., NO3− + NH4+) wet deposition rates are also shown. Observed data during 1994–1998, 1999–2003, and 2004–2008 were calculated using data of monitoring sites continuously operated during 1994–2008. SWJ, southwestern Japan; CJ, central Japan; NEJ, northeastern Japan.
 First, the two studies used different emission inventories. However, the NOx emission rates for 2001 used in this study differed from those of Lin et al. [2008a] only by +6%, −10%, and +15% in China, Japan, and other parts of eastern Asia, respectively. Therefore, differences in the emission inventories cannot explain the discrepancy in the contribution of China's emissions to NO3− wet deposition rates in Japan.
 Second, the methodology of the sensitivity analyses differed between the two studies. Lin et al. [2008a] conducted 25% emission reduction, whereas this study conducted 100% emissions reduction. However, Lin et al. [2008a] showed that the nonlinear effect was small in sensitivity analyses with 100% emission reduction for NO3− wet deposition in winter. Thus, it appears that the differences in methodology were not the main reason for the discrepancy, although this possibility cannot be excluded.
 Third, Lin et al. [2008a] evaluated total (dry plus wet) deposition rates, whereas we analyzed only wet deposition rates. Generally, dry deposition occurs in areas closer to emission sources as compared to wet deposition [Holloway et al., 2002]. Holloway et al.  estimated that the contribution of China's emissions to total (wet) deposition in Japan was 19% (28%). Thus, much of the discrepancy between Lin et al. [2008a] and this study likely reflects the differences in total and wet deposition rates.
 Comparison between the EyyMyy and EyyM00 simulations indicated that the contribution of meteorological fields to the increase of the NO3− wet deposition rate was less than 1.3%/yr (Figure 6). This value is similar to the simulated rates of increase of precipitation. The rates of increase of annual average precipitation were underestimated by the model. Still, the observed rates of increase of precipitation were less than 1.2%/yr, 1.8%/yr, and 0.9%/yr in NEJ, CJ, and SWJ, respectively. In addition, other meteorological parameters (e.g., wind field) might have made nonnegligible contributions to variations in the deposition fields. However, it is unlikely that the variations in meteorological conditions contributed to the increase of NO3− wet deposition rates by more than 1–2%/yr.
 As shown in Figure 7, NO3− wet deposition rates were particularly high along the coastal area of the Japan Sea (northwestern Japan). This horizontal distribution roughly reflected the precipitation fields. Nitrate wet deposition rates were simulated to have increased from 1989–1993 to 2004–2008 across all of Japan and the surrounding areas, suggesting that increased emissions from the Asian continent caused the increase of NO3− wet deposition rates across Japan.
3.2.2. NH4+ Wet Deposition
 Both observational data and simulations showed increasing trends in NH4+ wet deposition rates in NEJ and CJ in spring. The observational data indicated that rates of increase of NH4+ wet deposition were comparable to those of NO3− (3–4%/yr) in NEJ and CJ, whereas the model simulated a smaller rate of increase (1–2%/yr). The model also underestimated the rates of increase of annual average NH4+ wet deposition. According to the model, changes in emissions and meteorology made small, comparable contributions (<1.2%/yr) to the simulated rates of increase in NH4+ wet deposition (Figure 6). There are three possible reasons for the discrepancy between the observational data and the simulations.
 First, rates of increase of annual average precipitation were underestimated by the model by 0.5%–1%, which could cause the underestimation of rates of increase in NH4+ wet deposition. However, it is unlikely that the model errors related to meteorology caused discrepancies larger than 1–2%/yr.
 A second possible reason is uncertainty in NH3 emission rates. Ammonia emission rates are generally more uncertain than those of NOx and SO2 [e.g., Streets et al., 2003]. The REAS inventory estimated that NH3 emission rates increased during 1989–2005 by 1.3%/yr, although the NH3 emission rates in REAS have not been updated since 2000. We estimated the interannual trend of NH3 emission rates in China from 2000 to 2008 based on nitrogenous fertilizer consumption and number of livestock compiled in the China Statistical Yearbook [National Bureau of Statistics, 2009]. In the REAS inventory [Ohara et al., 2007], agricultural soil and livestock were estimated to account for 56% and 21% of NH3 emission rates in China in 2000, respectively. The results indicate that NH3 emission rates increased only by 0.6%/yr (+0.8%/yr from fertilizer and −0.2%/yr from livestock). Although the spatial patterns of NH3 emission increases may be heterogeneous [Zhang et al., 2010], it is unlikely that the rates of increase of NH3 emission are greatly underestimated in REAS.
 The third possibility is uncertainty in the model performance of predicting thermodynamic regimes. In the ammonium-limited regime where the [NH3 + NH4+]/[SO42−] ratio is less than unity on an equivalent basis [Seinfeld and Pandis, 2006], NH4+ concentration is strongly dependent on concentrations of total NH4 (i.e., NH3 + NH4+). By contrast, in the ammonium-rich regime ([NH3 + NH4+]/[SO42−] > 1), the NH4+ concentration depends on concentrations of SO42− and total NO3− (i.e., HNO3 + NO3−). Both observed and simulated results indicated that thermodynamic regimes ranged between ammonium-rich and ammonium-limited conditions, as shown in Appendix A (Figure A1). In this case, the sensitivity of NH4+ concentration (and deposition rates) to NOx, SO2, and NH3 emission rates is complex. The simulated rates of increase in NH4+ deposition (0.6–1%/yr) were lower than those of NO3− (∼4%/yr) and SO42− (3–5%/yr) and similar to the rates of increase in NH3 emissions in China (1.6%/yr). By contrast, the observed rates of increase in NH4+ deposition (2–3%/yr) were close to those of NO3− (2–5%/yr) and SO42− (3–5%/yr). Thus, it is likely that the model did not correctly reproduce the sensitivity of NH4+ deposition rates to its precursors. Detailed sensitivity analyses should be conducted to evaluate the factors controlling NH4+ concentrations and deposition rates.
3.2.3. Contributions of NO3− and NH4+ to Nitrogen Wet Deposition
 The observed NO3−/[NO3− + NH4+] equivalent ratios of wet deposition rates were about 50% during 1994–2008 in NEJ and CJ and showed small temporal variations (Table 4). In SWJ, however, the NO3−/[NO3− + NH4+] ratio of wet deposition rates increased from 39% during 1994–1998 to 49% during 2004–2008, and this pattern was consistent with the higher (lower) increasing rate of NO3− (NH4+) wet deposition in SWJ than in NEJ and CJ. Both observational and simulation results showed that NO3− and NH4+ made similar contributions to nitrogen wet deposition in Japan during 2004–2008. This result is consistent with the estimate of Dentener et al. , who showed using global CTMs that the NO3−/[NO3− + NH4+] ratio of wet deposition rate in Japan is higher than rates in other Asian countries. Dentener et al.  also estimated that the NO3−/[NO3− + NH4+] ratio in Japan will increase between 2000 and 2030 in a IPCC–SRES A2 scenario (the most pessimistic scenario) and will decrease in the current legislation scenario developed at IIASA. Nitrate wet deposition rates in Japan showed higher variability among scenarios than did the NH4+ wet deposition rates, suggesting that reduction of NO3− wet deposition rates by legislation regulating emissions is more feasible than reducing NH4+ deposition. Future changes in NOx emission rates and NO3− wet deposition rates should be carefully monitored.
3.3. Interannual Variation in NO3− Wet Deposition Rates in Spring
 We evaluated the factors controlling the interannual variation in NO3− wet deposition rates in spring by analyzing the area-weighted surface pressure anomaly (ASPA) meteorological index, a surface pressure anomaly in the western Pacific. On the basis of a model simulation of conditions in 1981–2005, Kurokawa et al.  showed that springtime O3 concentrations over western-central Japan correlated well with a surface pressure anomaly in the western Pacific (within the region bounded by 29.31°N, 141.62°E; 27.49°N, 149.85°E; 44.36°N, 146.82°E; and 41.76°N, 156.02°E) calculated using RAMS. This correlation was explained by the interannual variation in outflow patterns from the Asian continent.
 In this study, we conducted an analysis similar to that of Kurokawa et al.  for the NO3− wet deposition rates, precipitation rates, and particulate NO3− concentrations in SWJ and CJ during 1989–2008. Our analysis focused on spring, when transboundary pollution is the most pronounced. An anomaly was defined as a deviation from the values averaged over 1989–2008, and results of the E00Myy simulation were used for the calculation of anomalies in NO3− wet deposition rates, particulate NO3− concentrations, and precipitation rates. We also used ASPA as a reference parameter.
 During 1989–2008, simulated anomalies in NO3− wet deposition rates (r = 0.55 and 0.50 for SWJ and CJ, respectively) and precipitation rates (r = 0.59 and 0.71) correlated well with the observed results, suggesting that the model reproduced well the interannual variations of these parameters.
 Simulated anomalies in O3 concentration showed an opposite pattern of interannual variation to that of ASPA in SWJ and CJ (Figure 8). This behavior can be explained by the differences in pressure field and consequent differences in wind field [Kurokawa et al., 2009, Figure 6]. By contrast, NO3− wet deposition rates showed behavior similar to that of ASPA. As shown in Figure 9, anomalies of NO3− wet deposition rates showed a weak positive correlation with ASPA. This trend is roughly explained by the combination of interannual variations of precipitation rates and particulate NO3− concentrations. Precipitation rates showed positive correlations with ASPA in SWJ and CJ, whereas particulate NO3− concentrations showed weak negative correlations.
Figure 10a shows the horizontal distributions of NO3− concentrations, precipitation rates, and NO3− wet deposition rates in low-ASPA years (1993, 1996, 2001, 2007) and high-ASPA years (1991, 1995, 1998, 2003). In the low-ASPA years, the NO3− concentration across Japan was higher than in the high-ASPA years, likely caused by more direct transport of air masses from the Asian continent to Japan during low-ASPA years. Anomalies in the wind field showed that westerly wind components prevailed in the low-ASPA years, when anomalies in NO3− concentration were positive across Japan. This behavior is similar to that of O3 [Kurokawa et al., 2009], although negative anomalies in precipitation might have contributed to the positive anomalies in NO3− concentration.
 Precipitation rates across Japan were higher during high-ASPA years than during low-ASPA years. In particular, anomalies in the precipitation rate were strongly positive in southwestern to central Japan during the high-ASPA period (Figure 10b), when cyclones frequently pass over Japan [Kurokawa et al., 2009]. During the high-ASPA period, anomalies in the NO3− wet deposition rates showed positive values in western-central Japan, and this behavior can be explained by the combination of NO3− concentrations (weakly negative anomaly) and precipitation rate (positive anomaly).
 During the low-ASPA period, by contrast, cyclones frequently pass south of Japan, and anomalies in precipitation rates were generally negative across Japan. Jin et al.  reported similar behavior of precipitation in Japan. As a result of the combination of NO3− concentrations (positive anomaly) and precipitation rates (negative anomaly), anomalies of NO3− wet deposition rates were neutral to weakly positive in western-central Japan.
 As shown above, precipitation patterns greatly affect the distributions of nitrogen deposition in spring in eastern Asia. Based on global CTM results, Dentener et al.  reported that future climate changes would alter the regional amounts of nitrogen deposition by up to 20%, although climate changes would not greatly affect global average nitrogen deposition (∼0.6% change in NO3− deposition rates from 2000 to 2030). Our findings also indicate that the meteorological field is an important factor in regional nitrogen deposition. Because the conditions in ecosystems are highly variable from region to region, future changes in precipitation patterns and consequent deposition patterns in each region should be evaluated more carefully in future studies.
 To evaluate the interannual variations in nitrogen wet deposition across Japan during 1989–2008, we analyzed observational data and model results. In general, the baseline simulation (i.e., EyyMyy) reproduced well the spatial and temporal variations of wet deposition rates of NO3− across Japan. However, NH4+ wet deposition rates were overestimated, particularly in NEJ and CJ. The model also overestimated the [NH4+]/[NO3− + SO42−] ratio for wet deposition rates and atmospheric concentrations, particularly in winter and spring. The treatment of mineral cations (e.g., Na+ and Ca2+) in the model and uncertainties in NH3 emission estimates are possible reasons for the overestimation.
 Factors controlling the temporal trend in nitrogen wet deposition were evaluated by comparing the baseline simulation with sensitivity simulations (E00Myy, EyyM00, and E*yyMyy). Both observational data and simulation results showed that NO3− wet deposition rates increased between 1989 and 2008 in SWJ, CJ, and NEJ, with rates of increase of 2–5%/yr. Sensitivity simulations indicated that the simulated increasing trend of NO3− wet deposition was mostly explained (71%–100%) by the emission increase, as shown by the comparison between EyyMyy and E00Myy results. Also, comparison of EyyMyy and E*yyMyy simulations indicated that the increase in NO3− wet deposition rates was largely explained (61%–94%) by the increase of emissions in China. In addition, the contribution of China's emissions to NO3− wet deposition across Japan increased from 29%–35% during 1989–1993 to 43%–61% during 2004–2008. These results indicate that emissions from China are a major factor in the increase of NO3− wet deposition in Japan.
 Both observational data and model simulations indicated that NO3− and NH4+ made comparable contributions to nitrogen wet deposition in Japan during 2004–2008. Because it is more feasible to reduce NO3− wet deposition rates than those of NH4+ through legislation regulating emissions, NOx emission legislation is at least necessary to reduce nitrogen wet deposition rates in Japan. Future changes in NOx emission rates and NO3− wet deposition rates should be carefully monitored.
 Interannual variation was also evaluated by using the meteorological index ASPA. When ASPA was negative, air masses from the Asian continent were more directly transported to Japan, and NO3− concentrations became high across Japan. Thus, anomalies in NO3− concentrations showed a negative correlation with ASPA. However, anomalies in NO3− wet deposition rates showed a weak positive correlation with ASPA, reflecting the positive correlation between anomalies in precipitation rates and ASPA. This result suggests that precipitation patterns have a large impact on the interannual variation of NO3− wet deposition.
 Our findings indicated that NO3− wet deposition rates across Japan increased by ∼0.2 g N m−2 yr−1 between 1989 and 2008, mostly because of an increase of NOx emissions in China. Considering that the critical annual load of nitrogen deposition is 1–2.5 g N m−2 yr−1 [Fisher et al., 2007], the increase of NO3− wet deposition likely had some effects on ecosystems. Based on analysis of 21 global-scale models, Christensen et al.  predicted that precipitation rates in 2080–2099 will be higher than those in 1980–1999 in eastern Asia during all seasons. These higher precipitation rates could lead to greater nitrogen wet deposition. For a more comprehensive assessment of the effect of atmospheric nitrogen deposition on ecosystems, it is necessary to conduct more site-specific analyses using models with finer horizontal resolution and to evaluate the relative contribution of dry deposition.
 As discussed in section 3.1.2, wet deposition rates of NH4+ were generally overestimated by the model, particularly in NEJ and CJ, while those of its counter ions, NO3− and SO42−, were reproduced relatively well. In Appendix A, the observed and simulated ion balance is compared.
 Observed wet deposition rates showed that the ion balance between NO3−, non-sea-salt- (nss-) SO42−, NH4+, H+, and nss-Ca+ was well achieved (within 10%). Values of nss-SO42− and nss-Ca+ were calculated by subtracting the sea-salt contribution using the average sea-salt composition [Ministry of Environment, 2009]. Among the observed cations, the contribution of nss-Ca+ to the wet deposition rate was less important, and NH4+ and H+ had dominant and comparable contributions throughout the year. The model showed that the equivalent NH4+/[NO3−+ SO42−] ratios of wet deposition rates were 0.4–0.6 in summer (June–August) and autumn (September–November), suggesting that simulated NH4+ and H+ had comparable contributions (Figure A1a). This model result was consistent with observational data. However, the simulated [NH4+]/[NO3−+ SO42−] ratios in wet deposition rates were 0.6–1.0 in winter (December–February) and 0.4–0.8 in spring (March–May), and they were higher than the observed results.
 A possible reason for the discrepancies between the observed and simulated [NH4+]/[NO3−+ SO42−] ratios is that the balance of inorganic ions in atmospheric aerosols was not well captured by the model (Figure A1b). We analyzed the aerosol concentrations observed at 11 EANET stations across Japan from 2000 to 2007. Because aerosol concentrations were measured by the filter pack method on a biweekly basis, the observed aerosol concentrations include several biases [Acid Deposition and Oxidants Research Center, 2003]. Despite some measurement uncertainties, the observed [NH4+]/[NO3− + SO42−] ratios were clearly lower than the model results, particularly in winter. The observed [NH4+]/[NO3− + SO42−] ratio was significantly lower than unity (0.2–0.8) even in the ammonium-rich regime ([total NH4+]/[SO42−] > 1) (Figure A1c). This result indicates that cations other than NH4+, including Na+, Ca2+, and K+, in atmospheric aerosols contributed to the ion balance with NO3− and SO42−. By contrast, the model showed that the [NH4+]/[NO3− + SO42−] ratio approached unity in the ammonium-rich regime (Figure A1d). Thus, inclusion of other cations in the model is necessary for the accurate simulation of NH4+.
 In addition, the [total NH4+]/[SO42−] ratio differed between observational data and model simulation, particularly in winter (see Figures A1c and A1d). The observed [total NH4+]/[SO42−] ratio was lower than unity (i.e., ammonium-poor regime) in winter, and this is one reason for the low [NH4+]/[NO3− + SO42−] ratio. By contrast, the simulated [total NH4+]/[SO42−] ratio was higher than unity (i.e., ammonium-rich regime) in winter, and thus, the simulated [NH4+]/[NO3− + SO42−] ratio approached unity. Previous studies [Gilliland et al., 2003; Pinder et al., 2006] estimated seasonal variation of NH3 emission rates to be large, although these estimates had large uncertainties. Thus, seasonal variation of NH3 emission rates was not considered in REAS version 1.1, which is another possible reason for this discrepancy. Validation of NH3 emission rates and their seasonal variation is necessary in future studies.
 We thank K. Sato (Asia Center for Air Pollution Research) and Y. Hara (National Institute for Environmental Studies) for providing useful information about the EANET program and the RAMS simulation, respectively. We also thank the entire staff of the JADS and EANET for carrying out measurements and providing observation data sets.