Atmospheric nitrogen deposition to the northwestern Paciﬁc: seasonal variation and source attribution

Rapid Asian industrialization has led to increased atmospheric nitrogen downwind threatening the marine environment. We present an analysis of the sources and processes controlling atmospheric nitrogen deposition to the northwestern Paciﬁc, using the GEOS-Chem global chemistry model and its adjoint model at 1/2 ◦ × 2/3 ◦ hor- 5 izontal resolution over the East Asia and its adjacent oceans. We focus our analyses on the marginal seas: the Yellow Sea and the South China Sea. Asian nitrogen emissions in the model are 28.6 Tg N a − 1 as NH 3 and 15.7 Tg N a − 1 as NO x . China has the largest sources with 12.8 Tg N a − 1 as NH 3 and 7.9 Tg N a − 1 as NO x ; the high NH 3 emissions reﬂect its intensive agricultural activities. We ﬁnd Asian NH 3 emissions are 10 a factor of 3 higher in summer than winter. The model simulation for 2008–2010 is evaluated with NH 3 and NO 2 column observations from satellite instruments, and wet deposition ﬂux measurements from surface monitoring sites. Simulated atmospheric nitrogen deposition to the northwestern Paciﬁc ranges 0.8–20 kg N ha − 1 a − 1 , decreas-ing rapidly downwind the Asian continent. Deposition ﬂuxes average 11.9 kg N ha − 1 a − 1 15 (5.0 as reduced nitrogen NH x and 6.9 as oxidized nitrogen NO y ) to the Yellow Sea, and 5.6 kg N ha − 1 a − 1 (2.5

izontal resolution over the East Asia and its adjacent oceans. We focus our analyses on the marginal seas: the Yellow Sea and the South China Sea. Asian nitrogen emissions in the model are 28.6 Tg N a −1 as NH 3 and 15.7 Tg N a −1 as NO x . China has the largest sources with 12.8 Tg N a −1 as NH 3 and 7.9 Tg N a −1 as NO x ; the high NH 3 emissions reflect its intensive agricultural activities. We find Asian NH 3 emissions are 10 a factor of 3 higher in summer than winter. The model simulation for 2008-2010 is evaluated with NH 3 and NO 2 column observations from satellite instruments, and wet deposition flux measurements from surface monitoring sites. Simulated atmospheric nitrogen deposition to the northwestern Pacific ranges 0.8-20 kg N ha −1 a −1 , decreasing rapidly downwind the Asian continent. Deposition fluxes average 11.9 kg N ha −1 a −1 15 (5.0 as reduced nitrogen NH x and 6.9 as oxidized nitrogen NO y ) to the Yellow Sea, and 5.6 kg N ha −1 a −1 (2.5 as NH x and 3.1 as NO y ) to the South China Sea. Nitrogen sources over the ocean (ship NO x and oceanic NH 3 ) have little contribution to deposition over the Yellow Sea, about 7 % over the South China Sea, and become important (greater than 30 %) further downwind. We find that the seasonality of nitrogen deposi- 20 tion to the northwestern Pacific is determined by variations in meteorology largely controlled by the East Asian Monsoon and in nitrogen emissions. The model adjoint further estimates that nitrogen deposition to the Yellow Sea originates from sources over China (92 % contribution) and the Korean peninsula (7 %), and by sectors from fertilizer use (24 %), power plants (22 %), and transportation (18 %). Deposition to the South China

Introduction
Anthropogenic emissions of reactive nitrogen (or fixed nitrogen) have led to a rapid growth of nitrogen deposition to both land and marine ecosystems (Galloway et al.,5 2004; Duce et al., 2008;Liu et al., 2013). This additional input of nitrogen nutrient may enhance the primary production and carbon storage of the terrestrial biosphere (Pregitzer et al., 2008;Hyvonen et al., 2008). But excessive nitrogen deposition has been observed over sensitive ecosystems and can cause adverse effects including soil acidification and a reduction in plant biodiversity over land (Bowman et al., 2008;10 Stevens et al., 2004), and eutrophication on lakes and oceans (Bouwman et al., 2002). The northwestern Pacific is a region vulnerable to atmospheric nitrogen deposition as its productivity is generally limited by the low nutrient supply from deep water (Duce et al., 2008;Kim et al., 2011Kim et al., , 2014. Frequent incidences of harmful algal blooms in the marginal seas of the Pacific Ocean such as the Yellow Sea have been of great 15 concern (Hu et al., 2010). This region is subject to significant anthropogenic nitrogen deposition as it is located downwind of the Asian continent with high fixed nitrogen emissions from increasing human activities (Kurokawa et al., 2013;Luo et al., 2014). Increased nitrogen availability in waters of the northwestern Pacific has been observed in the past 30 years, most likely due to increasing deposition from the atmosphere H 2 SO 4 and HNO 3 , NH 3 forms ammonium sulfate and ammonium nitrate particles in the atmosphere. The formation of ammonium particles increase the lifetime of nitrogen in the atmosphere, promoting its long-range transport as dry deposition for particles is slow.
A large fraction (∼ 40 %) of emitted NH 3 and NO x enters the ocean via wet and 10 dry deposition from the atmosphere (Duce et al., 2008). Inputs from rivers provide additional fixed nitrogen to the ocean, but it is estimated that much of the riverine nitrogen is lost by denitrification in continental shelves and has a smaller impact on the open ocean (Seitzinger et al., 2006;Duce et al., 2008). Sanderson et al. (2008) showed using multiple models that about 10-15 % of the emitted NO x is exported out of East 15 Asia as nitrogen oxides (NO y ≡ NO x + HNO 3 + aerosol NO − 3 + PAN + N 2 O 5 + isoprene nitrates) with 34-49 % of them deposited within 1000 km distance. A number of studies have examined the processes of Asian pollution transport to the Pacific (Liu et al., 2003;Liang et al., 2004;Dickerson et al., 2007). Few studies have been conducted to quantify the patterns, processes, and source attribution of atmospheric nitrogen deposition to the northwestern Pacific.
We use the nested version of GEOS-Chem global chemical transport model (CTM) and its adjoint model with horizontal resolution of 1/2 • × 2/3 • Jiang et al., 2015) to investigate the factors controlling atmospheric nitrogen deposition to the northwestern Pacific, particularly over the Yellow Sea and the South China Sea. 25 Three-year (2008-2010) GEOS-Chem model simulations are conducted to quantify the deposition processes and to understand the impact of meteorology on the seasonal variability of atmospheric deposition. We evaluate the model simulation with surface measurements of wet deposition fluxes and satellite observations of NH 3 and NO 2  Zhang et al. (2012) has applied a similar nested model for North America to analyze the sources and processes of nitrogen deposition to the United States. The model includes a fully coupled tropospheric ozone-NO x -hydrocarbon-aerosol chemical mechanism (Bey et al., 2001;Park et al., 2004;Mao et al., 2010). Partitioning of gas and 20 aerosol phase of total NH 3 and HNO 3 is calculated using the ISORROPIA II thermodynamic equilibrium model (Fountoukis and Nenes, 2007). Following Zhang et al. (2012), we assume that isoprene nitrates produced from the oxidation of biogenic isoprene are removed by dry and wet deposition at the same rate as HNO 3 . The reactive uptake coefficients for N 2 O 5 in aerosols are from Evans and Jacob (2005) Model parameterization of wet deposition via both convective updraft and large-scale precipitation scavenging follows the scheme described by Liu et al. (2001) for aerosol, and by Mari et al. (2000) and Amos et al. (2012) for soluble gas. Dry deposition calculation follows a standard big-leaf resistance-in-series model (Wesely, 1989) including the aerodynamic resistance, the boundary layer resistance, and the canopy or surface 5 uptake resistance. Dry deposition velocities are calculated relative to the lowest model layer (∼ 70 m above the surface) as discussed in Zhang et al. (2012). Table 1 summarizes the model calculation of monthly mean daytime (10:00-16:00 LT) dry deposition velocities for different nitrogen species over the northwestern Pacific. Calculated dry deposition velocities are largest for HNO 3 , N 2 O 5 (0.56-10 1.16 cm s −1 ) and NH 3 (0.60-1.10 cm s −1 ), 0.06-0.08 cm s −1 for aerosol NH + 4 and NO − 3 , and near zero for insoluble species such as NO 2 and PAN. The values are generally much smaller than those over land (e.g., Table 1 of Zhang et al., 2012) as the uptake resistance over the smooth ocean surface is high. Deposition velocities are higher in winter than those in summer due to stronger winds near the ocean surface in winter. The 15 model calculated dry deposition velocities for aerosols are consistent with the mean value of 0.1 cm s −1 (with a range of 0.03-0.3 cm s −1 ) estimated by Duce et al. (1991) for aerosol dry deposition over the ocean surface.

Emissions
Global anthropogenic emissions (NO x , SO 2 , CO, and non-methane VOCs) are from 20 the Emission Database for Global Atmospheric Research (EDGAR) inventory (Olivier and Berdowski, 2001) except for global anthropogenic NH 3 emissions that are taken from the Global Emissions InitiAtive (GEIA) inventory (Bouwman et al., 1997). Regional emission inventories are then applied including the European Monitoring and Evaluation Programme (EMEP) inventory (Vestreng and Klein, 2002) over Europe, 25 the EPA 2005 National Emissions Inventory (NEI-2005) over the US, the Canada Criteria Air Contaminants (CAC) inventory (http://www.ec.gc.ca/pollution/default.asp? lang=En&n=E96450C4-1) over Canada, and the Regional Emission inventory in Asia 13662 The model also includes various natural sources of NH 3 and NO x . Lightning NO x emissions are calculated using the cloud top height parameterization of Price and Rind (1992), vertically distributed following Pickering et al. (1998), and further spatially constrained with satellite observations as described by  and Murray et al. (2012). Global lightning source is adjusted to be 6 Tg N a −1 . 10 Soil emissions are computed by the algorithm Yienger and Levy (1995) with canopy reduction factors . Biomass burning emissions of NO x and NH 3 are from the GFED-v2 inventory (van der Werf et al., 2006). Natural NH 3 emissions include both terrestrial and ocean emissions from the GEIA inventory (Bouwman et al., 1997). The REAS-v2 emission inventory is estimated based on activity data and emission 15 factors separated by different source categories (Kurokawa et al., 2013). Major NO x sources include fuel combustion in power plants, industry, transport and domestic sectors, and NH 3 sources are mainly from fertilizer use and manure management of livestock and human waste (Kurokawa et al., 2013). The sectorial information allows us to quantify nitrogen deposition contributions from different source categories in the adjoint 20 analysis as discussed in Sect. 5. The REAS-v2 NH 3 inventory consists of constant annual emissions without any seasonal variation (Kurokawa et al., 2013). Here we keep the annual total NH 3 emissions from REAS-v2 and derive monthly scalars over each model grid cell for NH 3 from different sectors (fertilizer use, livestock and human waste). NH 3 emissions from fertilizer 25 use are controlled by soil properties, meteorology, and the timing of fertilizer application. We follow the method and formula given in Skjøth et al. (2011) andPaulot et al. (2014). We consider nine types of crops (early rice/late rice, winter wheat/spring wheat, maize, cotton, sweet potatoes, potatoes, fruit and vegetables) with the har-Introduction  Monfreda et al. (2008). The growth cycles of those crops and their fertilizer inputs at different application time are based on Liao et al. (1993) and Sacks et al. (2010). For NH 3 emissions from livestock and human waste, we use the temperature-dependent experimental formula from Aneja et al. (2000). For the diurnal variability, the NH 3 agricultural emissions are increased by 90 % during the day and 5 reduced by 90 % at night following Zhu et al. (2013). Figure 1 shows the spatial distribution of annual total NH 3 and NO x emissions over Asia. Monthly NH 3 and NO x emissions from different source types over this region are also shown in Fig. 1 and the annual totals for Asia and China are summarized in Table 2. The largest NH 3 emissions are over the eastern China and India with values over 10 50 kg N ha −1 a −1 . We estimate strong seasonality for the NH 3 emissions from fertilizer use mainly determined by its usage timing, and from livestock and human waste depending on surface temperature. Asian NH 3 emissions are highest in May-August, and a factor of 3 higher than emissions in winter, similar to the seasonality of US NH 3 emissions in Zhang et al. (2012) derived by NH x (NH 3 gas + aerosol NH + 4 ) surface concen- 15 tration measurements and in Zhu et al. (2013) constrained by TES NH 3 observations. Natural NH 3 emissions account for 5 % of the total Asian NH 3 emissions in summer, 11 % in winter, and 7 % annually. Anthropogenic NO x emissions show weak seasonal variation, consistent with other emission estimates (Streets et al., 2003;. Natural NO x emissions (lightning, soil, and biomass burning) account for 23 % 20 of the total Asian NO x emissions in summer, 8 % in winter, and 16 % annually. Annual NH 3 and NO x emissions over China are respectively 12.8 and 7.9 Tg N a −1 (REAS-v2 anthropogenic and natural emissions). Our NH 3 emissions are at the high end of the range of 7.9-13.2 Tg N a −1 in the published Chinese NH 3 emission estimates (Streets et al., 2003;Dong et al., 2010;Paulot et al., 2014, and references therein). 25 This is mainly attributed to a higher estimate of NH 3 from fertilizer use in REAS-v2 (7.8 Tg N a −1 ) than other emission inventories (e.g., 3.2 Tg N a −1 in Huang et al., 2012). The successful simulation of NH 3 column concentrations and ammonium wet deposition fluxes as described below lends support to the high Chinese NH 3 emissions. Introduction as NO x ) , NH 3 emissions in China are a factor of 4 higher, reflecting its high levels of agricultural activities as well as the population.

The adjoint model
The adjoint method provides an efficient way to calculate the sensitivity of model vari-5 ables (e.g., concentrations and deposition fluxes) to model parameters (e.g., emissions). Here we briefly describe the adjoint method, and more details are given in Henze et al. (2007). Mathematically, the GEOS-Chem model can be viewed as a numerical operator F : y n+1 = F (y n , x), where y n is the vector of concentrations at time step n, and x is the vector of model parameters such as emissions. If we define a model = ∇ x J represents the sensitivity of J to model parameters, and λ 0 y = ∇ y0 J represents its sensitivity to the initial conditions. In the adjoint model they are computed simultaneously backwards:  (Henze et al., 2007(Henze et al., , 2009. The adjoint of the ISORROPIA aerosol thermodynamic equilibrium model was constructed by Capps et al. (2012). The GEOS-Chem adjoint model has been evaluated and applied in a number of studies, including optimizing aerosol emission (Henze et al., 2009;Zhu et al., 2013), attributing sources of ozone pollution in the western US , and quan-5 tifying processes affecting nitrogen deposition to biodiversity hotpots worldwide (Paulot et al., 2013(Paulot et al., , 2014. Those studies used the adjoint model at global 4 • × 5 • or 2 • × 2.5 • resolution. The adjoint of the nested-grid GEOS-Chem has been developed by Jiang et al. (2015), and applied to constrain black carbon emissions  and assess human exposure to Equatorial Asian fires (Kim et al., 2015). Here we apply it to quantify sources contributing to atmospheric nitrogen deposition over the northwestern Pacific.

Column concentrations and wet deposition fluxes over Asia
We compare model simulation of NH 3 tropospheric columns with satellite measurements from the Tropospheric Emissions Spectrometer (TES) (Beer, 2006), and NO 2 15 tropospheric columns with those from the Ozone Monitoring Instrument (OMI) (Levelt et al., 2006). Both are aboard the NASA Aura satellite in a sun-synchronous orbit with an ascending equator crossing time of 13:45 (Beer, 2006). These comparisons provide valuable tests of the nitrogen emissions in the model. We evaluate model simulated wet deposition fluxes of ammonium and nitrate with observational 20 data from the Acid Deposition Monitoring Network in East Asia (EANET; data available at http://www.eanet.asia/index.html) and ten sites monitored by the Chinese Academy of Science (CAS) located in North China (Pan et al., 2012). Measurements of nitrogen dry deposition fluxes are rather limited over the northwestern Pacific. Figure 2 compares TES measured and GEOS-Chem simulated NH 3 tropospheric 25 columns in summer (June-August). TES is an infrared Fourier transform spectrometer with high spectral resolution of 0.06 cm −1 (Beer, 2006). The observations have ACPD 15,2015 Atmospheric nitrogen deposition to the northwestern Pacific Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | a spatial resolution of 5 km × 8 km with global coverage achieved in 16 days. NH 3 retrievals from TES are based on the optimal estimation method of Rodgers (2000), as described by Shephard et al. (2011). Following Zhu et al. (2013) that used TES NH 3 observations to optimize the US NH 3 emissions, we filter the TES observations based on the retrieval quality control flags, and only use the daytime observations. We use 5 TES observations in summer as they generally have highest sensitivities during the year (Shephard et al., 2011), and use observations collected from 2005 to 2010 to increase the number of observations for comparison. The GEOS-Chem model results for 2009 are sampled along the TES orbit tracks at the overpass time, and then applied with the TES retrieval operator following Zhang et al. (2006) and Zhu et al. (2013). As 10 shown in Fig. 2, the model generally captures the observed high NH 3 columns over the North and Northeast China, and India (correlation coefficient r = 0.53). The model largely underestimates NH 3 columns over India by 28 %, which suggests NH 3 emissions over India are too low. For observations over China, the model only has a small negative bias of 3 %. 15 Figure 2 also compares OMI measured and GEOS-Chem simulated NO 2 tropospheric columns averaged over March-November 2009. OMI measures backscattered solar radiation over the 270-500 nm wavelength range, and has a spatial resolution of 13 km × 24 km and daily global coverage (Levelt et al., 2006;Boersma et al., 2011). We use the monthly OMI NO 2 data (DOMINO v2.0) from KNMI (http://www.temis.nl/). 20 The wintertime measurements are excluded due to large retrieval errors over snow (O'Byrne et al., 2010). The model generally captures the observed distribution of NO 2 tropospheric columns over Asia (r = 0.93), but it is biased low by 15 % over North China on average. A recent study by Lin et al. (2014) suggested that DOMINO NO 2 columns might be biased high due to overestimates of surface pressure and exclusion 25 of aerosols in the retrieval.
We compare in Fig. 3 the observed vs. simulated seasonal mean ammonium and nitrate wet deposition fluxes at the EANET and CAS monitoring sites. The EANET data and model results are averaged for January 2008-December 2010, and the CAS ACPD 15,2015 Atmospheric nitrogen deposition to the northwestern Pacific  Previous studies have shown that model simulation of wet deposition flux is highly sensitive to the model precipitation (Pinder et al., 2006;Paulot et al., 2014). We evaluate the GEOS-5 precipitation data over the northwestern Pacific with data from the CPC Merged Analysis of Precipitation (CMAP). The CMAP data are based on several satellite measurements as described in Xie and Arkin (1997), and have a spatial resolution 15 of 2.5 • × 2.5 • and monthly variation (data available at http://www.cpc.ncep.noaa.gov/ products/global_precip/html/wpage.cmap.html). Figure 4 compares the monthly averaged GEOS-5 precipitation data with CMAP in January, April, July and October 2009. Both CMAP and GEOS-5 show maximum precipitation over the northwestern Pacific Ocean in July and minimum in January. The GEOS-5 precipitation data generally agree 20 well with the CMAP data (r = 0.83-0.92), with only small negative biases of 2-5 % over the ocean.
To focus on the northwestern Pacific, we further examine the measured and simulated nitrogen wet deposition fluxes at nine coastal EANET sites. Figure 5 shows locations of these monitoring sites and the focused region of this study. Figure 6a

Seasonal variation and deposition process
We now examine the deposition processes, patterns, and seasonal variation of atmospheric nitrogen deposition to the northwestern Pacific. Figure 7 shows the spatial distribution of total nitrogen deposition (ammonium and nitrate, dry and wet) to the northwestern Pacific in January, April, July and October, and

20
(0.8-4 kg N ha −1 a −1 ). We selected two regions as shown in Fig. 7 representing the Yellow Sea and the South China Sea. Table 3 summarizes the monthly and annual nitrogen deposition fluxes over the two regions for 2008-2010. Nitrogen deposition averages 11.9 kg N ha −1 a −1 over the Yellow Sea (5.0 kg N ha −1 a −1 as reduced nitrogen NH x and Introduction Yellow Sea is weak with fluxes in October and January about 10 % higher than April and July. Nitrogen deposition to the South China Sea averages 5.6 kg N ha −1 a −1 with deposition in January nearly a factor of 3 higher than deposition in July. This reflects seasonal variations in both meteorology and nitrogen emissions as will be discussed below.

5
Wet deposition accounts for 67 % of the total nitrogen deposition to the Yellow Sea (82 % for NH x and 57 % for NO y ) and the South China Sea (84 % for NH x and 55 % for NO y ). The ratio of wet vs. dry deposition over the ocean is generally higher than that over the land, because of slow dry deposition velocities (Table 1)

Contribution from the oceanic emissions
It is important to separate the contributions of ocean vs. land emissions to the nitro- 15 gen deposition over the northwestern Pacific. Sources of fixed nitrogen from the ocean include both anthropogenic ship NO x emissions and natural oceanic NH 3 emissions. Those emissions are small compared with land sources, but their contributions to the nitrogen deposition over the open ocean cannot be neglected due to the short lifetimes of nitrogen species. We have conducted two sensitivity simulations respectively 20 with ship NO x emissions or oceanic NH 3 emissions shut off. The differences with the standard simulation represent contributions of each source to the nitrogen deposition.
We separate in Fig. 8 the annual contributions of nitrogen sources over land, ship NO x emissions, and oceanic NH 3 emissions to total nitrogen deposition over the northwestern Pacific. We can see nitrogen deposition to the marginal seas of the north-

Outflow from mainland China
We have demonstrated above that nitrogen deposition to the marginal seas of the 5 northwestern Pacific such as the Yellow Sea and the South China Sea mainly originates from nitrogen sources over the land. We now focus on the outflow fluxes from Mainland China where the largest nitrogen emissions are located. Figure 9 shows the outflow fluxes of fixed nitrogen transported across the coastline of Mainland China (as defined by the grid cells in Fig. 5) in different seasons. Fluxes of NH 3 , NH + 4 , HNO 3 , isoprene 10 nitrates, and NO − 3 are included. Other fixed nitrogen species such as PAN, although important in outflow fluxes, account for less than 1 % of the nitrogen deposition to the northwestern Pacific.
We can see that the spatial and seasonal variation of atmospheric nitrogen deposition over the marginal seas of the northwestern Pacific as shown in Fig. 7 can be mainly  Fig. 5) to the South China Sea, nitrogen fluxes are largest within the boundary layer in January and October. The fluxes turn to inflow in April and July, minimizing deposition to the South China Sea during these months. 25 The seasonal variation of pollution transport over the eastern Asia is largely controlled by the East Asian monsoon system (Liu et al., 2003;Liang et al., 2004;. We show in Fig. 10  ary layer (generally below 950 hPa) and in the free troposphere at 700 hPa plotted over the monthly emissions of fixed nitrogen. In January the northwesterly monsoon prevails at middle latitudes (> 30 • ) in the boundary layer and gradually turns to the northeasterly at lower latitudes (< 30 • ). Asian pollution is generally trapped in the boundary layer by the large-scale subsidence over the continent and transported southward as shown 5 in Fig. 9. In July, the summer southerly monsoon winds bring clean ocean air to the southern China, but at latitudes north of 30 • N the southwesterly winds combined with the high nitrogen emissions over the eastern China lead to large fluxes to the Yellow Sea. Spring and fall represent the transitional periods, and frequent cold fronts are the primary driver lifting anthropogenic pollution to the free troposphere followed by 10 westerly transport (Liu et al., 2003;Liang et al., 2004). Thus the strong seasonal variation in nitrogen deposition to the South China Sea is mainly attributed to the monsoonal Asian outflow. Over the Yellow Sea, we find the weaker winds in July can be compensated by higher nitrogen emissions over the land, leading to the weak seasonality of nitrogen deposition. We find in a sensitivity simula-15 tion without seasonal variations of Asian NH 3 emissions that nitrogen deposition to the Yellow Sea would have been 64 % higher in January than July.

Source attribution using the adjoint method
The adjoint model allows us to further quantify the sources contributing to atmospheric nitrogen deposition over the receptors at the model underlying grid scale. Here we 20 calculate the sensitivities of nitrogen deposition (reduced and oxidized nitrogen, wet and dry) over the Yellow Sea and the South China Sea to grid-resolved NH 3 and NO x emissions for January, April, July and October 2009. For each month, we calculate sensitivity of the monthly mean nitrogen deposition to emissions in that month and a week in the preceding month (accounting for the lifetimes of nitrogen species). We Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | and industry and power plants for NO x ) based on their relative contributions to the total anthropogenic emissions. The top panels of Fig. 11 show the adjoint sensitivities for the monthly total nitrogen deposition to the Yellow Sea. The magnitude of adjoint sensitivity reflects deposition amount contributed by the nitrogen emissions in each grid cell. The sum of sensitivities 5 integrated geographically matches the monthly deposition flux to the Yellow Sea within 5 %. From winter to summer the source regions move southward from North China and Northeast China to East China and Central China, consistent with the seasonal variation of the monsoonal flow. Nitrogen sources over China are the main contributor to the nitrogen deposition to the Yellow Sea (93 in January, 88 in July, and 92 % annually).
Sources over the Korean peninsula contribute 7 % of annual nitrogen deposition to the Yellow Sea.
The bottoms panels of Fig. 11 separate the sensitivities of nitrogen deposition components (reduced and oxidized nitrogen, wet and dry) to emissions from different source types. The total sensitivity of each deposition component also approximately 15 matches the simulated deposition flux (Table 3), with small discrepancies of 0.01-0.06 kg N ha −1 month −1 that can be attributed to nonlinearity between nitrogen deposition and emissions (including nitrogen, SO 2 , and VOC emissions) as discussed in Paulot et al. (2013). Figure 11 shows that NO x emissions from power plants (37 %), followed by emissions from transport (26 %) and industry (22 %) contribute most of 20 the nitrogen deposition in January. In other months, NH 3 emissions from fertilizer use (25-32 %) are the largest source of nitrogen deposition to the Yellow Sea. Annually the major sources contributing to nitrogen deposition to the Yellow Sea are fertilizer use (24 %), power plants (22 %), and transportation (18 %). Figure 12 shows source attribution of atmospheric nitrogen deposition to the South 25 China Sea. Unlike that to the Yellow Sea, nitrogen deposition to the South China Sea shows a distinct winter peak as reflected by the largest source contributing areas in January spreading over the Asian continent. Kim et al. (2014) (Table 1). NH 3 emissions would lead to formation of aerosol NO − 3 from HNO 3 , increasing the lifetime of NO y and allowing them transport to a longer distance. It would thus decrease the dry deposition of NO y (mainly via reduction of HNO 3 ) near the source region (e.g., the Yellow Sea), and enhance its dry deposition further down-25 wind (e.g., the South China Sea). The same response applies to NH x dry deposition and NO x emissions (the April panel of Fig. 11), but it is much weaker because NH x dry deposition fluxes to the ocean are small and mainly from dry deposition of aerosol NH + 4 . This can have important implications on the effectiveness of the emission con-13674 ACPD 15,2015 Atmospheric nitrogen deposition to the northwestern Pacific trol strategy for reducing nitrogen deposition to the Yellow Sea. As shown in Fig. 11, NH 3 emissions from fertilizer use are identified as the largest contributor to nitrogen deposition to the Yellow Sea except in winter. However, we estimate annually 28 % (negative sensitivity of NO y dry deposition vs. sensitivity of NH x total deposition to NH 3 emissions, averaged over the four months in Fig. 11) of the expected benefits of reduc-5 tion of nitrogen deposition to the Yellow Sea via controlling NH 3 would be offset by an increase in NO y dry deposition.

Conclusions
Increasing atmospheric nitrogen deposition to the northwestern Pacific has likely been altering the marine environment. The purpose of this study is to quantify the sources, 10 processes, and seasonal variation of atmospheric nitrogen deposition to the northwestern Pacific. We have used a nested-grid version of the GEOS-Chem global chemistry model and its adjoint model to address the issue. The model has a horizontal resolution of 1/2 • latitude × 2/3 • longitude over the East Asia and its adjacent oceans (70-150 • E, 11 • S-55 • N), and 4 • × 5 • over the rest of the world. It includes a detailed 15 tropospheric chemistry to simulate the sources, transformation, and deposition of fixed nitrogen (NH x and NO y ) in the atmosphere. The model uses the anthropogenic emissions of fixed nitrogen (via NH 3 and NO x ) from the REAS-v2 emission inventory for Asia (Kurokawa et al., 2013). The original NH 3 emissions had no seasonal variation, inconsistent with recent Asian NH 3 emis-20 sion estimates. We calculate the seasonal variations for NH 3 emissions from fertilizer use based on soil properties, meteorology, and the timing of fertilizer application (Skjøth et al., 2011;Paulot et al., 2014), and for NH 3 from livestock and human waste using surface temperature (Aneja et al., 2000). The resulting Asian NH 3 emissions are highest in May-August, with emissions in summer a factor of 3 higher than winter. Total 25 Asian NH 3 and NO x emissions are 28.6 and 16.2 Tg N a −1 , respectively. China has the largest nitrogen sources with 12.8 Tg N a −1 as NH 3 and 7.9 Tg N a −1 as NO x . Both NH 3 ACPD 15,2015 Atmospheric nitrogen deposition to the northwestern Pacific and NO x emissions are dominated by anthropogenic sources. Natural sources account for 7 % for NH 3 , and 16 % for NO x . We evaluate the model simulation of NH 3 and NO 2 tropospheric columns with satellite observations from TES and OMI over Asia. The model generally captures the observed distribution of NH 3 and NO 2 tropospheric columns with only small negative 5 biases for both species (−3 % for NH 3 over China and up to −15 % for NO 2 over the North China), providing support to the model emissions. The model further closely reproduces the magnitudes and variability of ammonium and nitrate wet deposition fluxes at the EANET sites and additional monitoring sites over the North China. Wet deposition fluxes measured over the continental sites show strong seasonality with summer 10 maximum and winter minimum, while for the island sites in the open ocean, deposition fluxes are much smaller with weak seasonal variations.
We analyze three-year (2008-2010) model simulation of atmospheric nitrogen deposition to the northwestern Pacific, particularly over the marginal seas such as the Yellow Sea and the South China Sea. Atmospheric nitrogen deposition reaches as 15 high as 20-55 kg N ha −1 a −1 in the eastern China, and decreases rapidly downwind the Asian continent (0.8-20 kg N ha −1 a −1 over the northwestern Pacific). Nitrogen deposition averages 11.9 kg N ha −1 a −1 over the Yellow Sea (5.0 kg N ha −1 a −1 as NH x and 6.9 kg N ha −1 a −1 as NO y ), and 5.6 kg N ha −1 a −1 to the South China Sea (2.5 as NH x and 3.1 as NO y ). Although Asian NH 3 emissions are much higher than NO x emissions, 20 less NH x is exported and deposited over the open ocean due to its shorter lifetime. We find contributions of nitrogen sources over the ocean, including ship NO x emissions and oceanic NH 3 emissions, are negligible for nitrogen deposition to the Yellow Sea, and about 7 % over the South China Sea. Further downwind in the ocean ship NO x emissions contribute 10-25 % of total nitrogen deposition along the ship tracks, and Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | January (0.62 kg N ha −1 month −1 ) nearly a factor of 3 higher than deposition in July (0.23 kg N ha −1 month −1 ). This is consistent with the nitrogen outflow fluxes from Asia (mainly Mainland China), which are controlled by the East Asian monsoon system as discussed in previous studies (Liu et al., 2003;. In winter the northerly monsoon favors transport of Asian pollution to the open ocean in the bound-5 ary layer, while the summer southerly monsoon winds bring clean ocean air to the southern China. Nitrogen deposition to the Yellow Sea has weak seasonality (0.85-1.12 kg N ha −1 month −1 ). We find the weaker winds in summer over the Yellow Sea suppress dry deposition of nitrogen, but are compensated by higher nitrogen emissions in summer. 10 We have further applied the adjoint of GEOS-Chem to estimate the contributions of nitrogen sources from different sectors and at the model underlying resolution to nitrogen deposition over the Yellow Sea and the South China Sea. This detailed source information can be crucial to design an effective strategy for reducing nitrogen deposition to these areas. Nitrogen deposition to the Yellow Sea mainly originates from 15 nitrogen sources over China (92 % contribution) and the Korean peninsula (7 %) categorized by regions, and is contributed from fertilizer use (24 %), power plants (22 %), and transportation (18 %) categorized by emission sectors. For deposition to the South China Sea, nitrogen sources over Mainland China and Taiwan contribute 66 and 20 % of the annual total deposition, with the rest 14 % from sources over the Southeast Asian 20 countries as well as oceanic NH 3 emissions. Natural sources are particularly important in April, accounting for 17 % of the nitrogen deposition to the South China Sea (7 % from the oceanic NH 3 emissions, 4 % from lightning, and 6 % from biomass burning emissions over Southeast Asia).
The adjoint analyses also indicate that dry deposition of oxidized nitrogen to the 25 Yellow Sea shows negative sensitivity to Asian NH 3 emissions, ie., reducing Asian NH 3 emissions would increase the NO y dry deposition to the Yellow Sea. This response mainly reflects conversion of gaseous NH 3 and HNO 3 to ammonium nitrate aerosol and their different deposition efficiencies. Annually 28 % of the reduction of nitrogen Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Kim, T. W., Lee, K., Duce, R., and Liss, P.: Impact of atmospheric nitrogen deposition on phytoplankton productivity in the South China Sea, Geophys. Res. Lett., 41, 3156-3162, 2014. Kopacz, M., Jacob, D. J., Henze, D. K., Heald, C. L., Streets, D. G., and Zhang, Q Van Roozendael, M., Clémer, K., and Irie, H.: Retrieving tropospheric nitrogen dioxide from the Ozone Monitoring Instrument: effects of aerosols, surface reflectance anisotropy, and vertical profile of nitrogen dioxide, Atmos. Chem. Phys., 14, 1441-1461, doi:10.5194/acp-14-1441 Constraints from Pb-210 and Be-7 on 25 wet deposition and transport in a global three-dimensional chemical tracer model driven by assimilated meteorological fields, J. Geophys. Res., 106, 12109-12128, 2001. Liu, H. Y., Jacob, D. J., Bey, I., Yantosca, R. M., Duncan, B. N., and Sachse, G. W Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | van der Werf, G. R., Randerson, J. T., Giglio, L., Collatz, G. J., Kasibhatla, P. S., and Arellano Jr., A.  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Zhang, L., Jacob, D. J., Kopacz, M., Henze, D. K., Singh, K., and Jaffe, D. A.: Intercontinental source attribution of ozone pollution at western US sites using an adjoint method, Geophys.  2008-2010. b Isoprene nitrates represent the organic nitrates produced from the oxidation of isoprene by OH in the presence of NO x . c Peroxyacetyl nitrate (PAN) and higher peroxyacyl nitrates. 15,2015 Atmospheric nitrogen deposition to the northwestern Pacific  ACPD 15,2015 Atmospheric nitrogen deposition to the northwestern Pacific ACPD 15,2015 Atmospheric nitrogen deposition to the northwestern Pacific   15,2015 Atmospheric nitrogen deposition to the northwestern Pacific  Figure 5. Map of the focused domain. The black dots are the locations of nine EANET sites that used for model evaluation of nitrogen deposition near the coast ( Fig. 6a and b): Mt. Sto. Tomas, Hedo, Cheju, Imsil, Kanghwa, Xiamen (Hongwen and Xiaoping sites), and Zhuhai (Xiang Zhou and Zhuxiandong sites). The red dots represent the grid cells covering the coastline of Mainland China that used for determining the outflow fluxes as indicated by the orange arrows. 15,2015 Atmospheric nitrogen deposition to the northwestern Pacific  15,2015 Atmospheric nitrogen deposition to the northwestern Pacific