A Review of Spatial Variation of Inorganic Nitrogen (N) Wet Deposition in China

Atmospheric nitrogen (N) deposition (Ndep), an important component of the global N cycle, has increased sharply in recent decades in China. Although there were already some studies on Ndep on a national scale, there were some gaps on the magnitude and the spatial patterns of Ndep. In this study, a national-scale Ndep pattern was constructed based on 139 published papers from 2003 to 2014 and the effects of precipitation (P), energy consumption (E) and N fertilizer use (FN) on spatial patterns of Ndep were analyzed. The wet deposition flux of NH4+-N, NO3--N and total Ndep was 6.83, 5.35 and 12.18 kg ha-1 a-1, respectively. Ndep exhibited a decreasing gradient from southeast to northwest of China. Through accuracy assessment of the spatial Ndep distribution and comparisons with other studies, the spatial Ndep distribution by Lu and Tian and this study both gained high accuracy. A strong exponential function was found between P and Ndep, FN and Ndep and E and Ndep, and P and FN had higher contribution than E on the spatial variation of Ndep. Fossil fuel combustion was the main contributor for NO3--N (86.0%) and biomass burning contributed 5.4% on the deposition of NO3--N. The ion of NH4+ was mainly from agricultural activities (85.9%) and fossil fuel combustion (6.0%). Overall, Ndep in China might be considerably affected by the high emissions of NOx and NH3 from fossil fuel combustion and agricultural activities.


Introduction
Atmospheric nitrogen (N) deposition (N dep ) has dramatically increased in the past few decades owing to the rapid increases of industrialization, urbanization and intensified agricultural production in China [1][2][3][4]. Currently, the intensity of N dep is equal or even exceeds that in Europe and America [5], causing general concerns of the governments and the public. Increased N dep in terrestrial or aquatic ecosystems or both degrade human health [6], alter chemical components of soil and water [7], influence greenhouse gas balance [8] and reduce biological diversity [9]. Therefore, it is critical to estimate N dep patterns for quantifying the effects of N amendment and establish control measures to improve environmental quality.
Some studies have reported the observed results of N dep at a local scale in China [10][11][12]. These investigations mainly collected N deposition samples from different sampling sites in some local areas, determined the fluxes of N dep , characterized the seasonal or annual variation, assessed the potential ecological risk and analyzed possible sources of N dep [1,[13][14][15][16][17][18][19]. They have demonstrated that atmospheric N dep in China increased rapidly over recent decades primarily due to increased energy consumption and N fertilizer use, and this increasing trend will continue in the future with the continuing development of China's economy. However, most of these studies did not give the magnitude and spatial pattern of N dep throughout China due to the difficulty of obtaining the N fluxes on a large area of China [20][21][22][23][24].
There have been several studies on N dep throughout China. For example, Lu and Tian [1] reported N dep peaked over central south of China, with an average value of 12.89 kg ha -1 a -1 from site-network observations. Moreover, they [14] resulted in the N dep was 14.05 kg ha -1 a -1 (on the assumption that wet N dep contributes 70% of bulk deposition) in the recent decade, combining site-level monitoring and atmospheric transport model, and they resulted that the most rapid increase centered in southeastern China. Liu et al. [3] believed that N dep increased to 21.1 kg ha -1 a -1 , based on the atmospheric deposition monitoring network and the published papers, and they pointed out that the N dep in the industrialized and agriculturally intensified regions of China as high as the peak levels in northwestern European in 1980s. Jia et al. [25] concluded that N dep was 13.87 kg ha -1 a -1 in the 2000s, using the N fluxes at 41 stations, with an increasing rate of 25% than that in the 1990s and the highest N dep occurred in southern China. Zhu et al. [4] demonstrated that N dep was 13.18 kg ha -1 a -1 , accounting for 73% of total N dep and peaked in central and southern China.
From the above analysis, the magnitude of N dep and the spatial distribution of N dep were not consistent in the mentioned studies. Liu et al. [26] believed that Zhu et al. [4] might underestimate the dissolved inorganic nitrogen (DIN) due to the uncertainty resulting from the sampling, storage and analysis methods in their study [26]. Pan and Li [27] thought that Lu and Tian [14] underestimated N dep based on a ratio of 0.7 and found the ratio was about 0.4 in Northern China [28]. Therefore, it is still an open question on the spatial pattern and magnitude of N dep in China.
On the national scale of N dep , the influencing factors on the spatial variations of N dep were also studied. The spatial variations of N dep had been greatly influenced by factors including N fertilizer use (F N ), energy consumption (E), and precipitation (P). Zhan et al. [29] hold that F N , E, and P jointly explained 84.3% of the spatial pattern of N dep , of which F N (27.2%) was the most important, followed by E (24.8%) and P (9.3%). Zhu and He [4] found P and F N can explain 80-91% of the spatial variation of N dep , but E had little effect on this variation. Jia et al. [25] reported that F N , E and P combined contributed 79% on the spatial variation of N dep , while E contributed 80% of decadal variation followed by F N , but P had little effect. These results obtained different opinions on the influences of F N , E and P on the spatial variations of N dep . The interrelationship between N dep and these factors also should be further studied on a national scale.
The present study aims to (1) identify the magnitude and the spatial pattern of N dep throughout China, (2) summarize how precipitation, N fertilizer use and energy consumption influencing spatial pattern of N dep , quantify the correlation between factors and N dep , and (3) determine the contributions of potential sources to the magnitude of N dep in China.

Materials and Methods
The flowchart of this study is shown in Fig 1. Firstly, the N fluxes from the published papers throughout China were obtained, and then the Kriging interpolation technique is applied to Inorganic Nitrogen (N) Wet Deposition calculate N dep on a national scale and compared the result with other N dep maps in other studies. Then, the influence of P, F N and E on the spatial pattern of N dep is analyzed. Finally, potential sources of N dep are evaluated.

Data collection
To evaluate N dep throughout China, it is critical to systematically collect the relevant published papers. In this study, the data pairs on precipitation sampling in China during 2003-2014 were collected. These studies were located by making a search through ISI Web of Knowledge using keywords "nitrogen deposition", "chemical composition" or "precipitation" and "China", and through CNKI website using the same Chinese keywords. Finally, 139 peer reviewed articles consisting 225 data records (Fig 2) on NH 4 + -N and NO 3 --N in precipitation throughout China were collected (S1 Table). Basic information included the name of the monitoring sites, location, land use, rainfall, monitoring time span, annual precipitation, concentration and depositions of NH 4 + -N and NO 3 --N and literature source from each study. To assure the monitoring quality of rainwater components, the studies based on the technical specifications required for acid deposition monitoring in China (State Environmental Protection Administration of China, 2004) were selected to establish datasets on N dep . The data on the amount of F N and E on provincial scales could be obtained from the China Statistical Yearbook from 2003 to 2014 (http://www.stats.gov.cn/tjsj/). Due to the lack of energy data in Tibet province, we assumed that the per capita energy consumption was similar between the Tibet and Xinjiang provinces, which are both located in western China, and deduced data on energy consumption in Tibet province from the Xinjiang province data.
The data on the annual precipitation were obtained from China Meteorological Administration. The mean annual precipitation in provinces was calculated based on the annual precipitation from 2003 to 2014, respectively, from the weather stations in each province.

Calculation of wet N dep
Wet inorganic N deposition is calculated as the product of the precipitation amount and the concentration of N species in precipitation. The wet N deposition flux was kg N ha -1 and the unit of the precipitation is mm. The units of the concentration of N species in precipitation include mg N L -1 [30] and μeq L -1 [31]. Both of the two units are commonly used. Thus, when the unit of the concentration of N species is mg N L -1 , the calculation formula of nitrogen deposition is: where N dep is the N deposition flux per year (kg ha -1 a -1 ); C i is the concentration of NH 4 + -N or NO 3 --N (mg N L -1 ); P i is the annual precipitation (mm); 100 is the conversion factor.
Otherwise, the formula is: where N dep is the N deposition flux per year (kg ha -1 a -1 ); C i is the concentration of NH 4 + -N or NO 3 --N (μeq L -1 ); P i is the annual precipitation (mm); 14 is the atomic weight of N and 10 5 is the conversion factor.

Geo-statistical method
A geostatistical method was used to produce spatially continuous estimates from discrete data points. National-scale N dep maps were constructed using the Kriging interpolation technique. An unknown value associated with a point can be estimated by Kriging as follows: where λ i is the Kriging weights computed from a normal system of equations using a semivariance function, derived by minimization of the error variance; the unknown value Z(x 0 ) is interpreted as a random variable located in x 0 , as well as the values of neighbor samples Z(x i ), i = 1, . . ., N. Prior to Kriging interpolation, the Explore Data tool of ArcGIS 10.0 software is applied to conduct a data analysis, including data's distributing, outlier identification, and trend analysis; the optimal variogram model and parameters are determined by GS plus.

Source apportionment of ionic species
Positive matrix factorization (PMF) developed by the U.S. Environmental Protection Agency (EPA) is a multivariate factor analysis that utilizes error estimates and produces non-negative results [32]. PMF is used to factorize a given dataset into two matrices, the source profile (F) and source contribution (G), also called factors, which is expressed by the following formula: where x ij is are the elements of the input data matrix, g ik and f kj are the elements of the factor scores and factor loading matrices, respectively; e ij is the residuals (i.e. the difference between input data and predicted values) and p is the number of factors resolved [33]. The resolving algorithm computes G and F elements that minimize the so-called object function Q.
where S ij represents the elements of uncertainty matrix, and each element is the uncertainty of jth species for sample i.

Results and Discussions
Accuracy assessment of the spatial N dep distribution and comparisons with other studies Although there were several studies on the estimation of wet N dep on a national scale in China, most of them showed different spatial patterns. Which map of N dep could reflect the real spatial distribution of N dep in China is still a question.
At a point scale, the 41 sites of N dep in Zhu [4] were used to estimate the accuracy of the spatial distribution of N dep by the method of Kriging. The Q-Q plot of the distribution of sitemonitored N dep versus that of the interpolated N dep in this study is shown in Fig 3. The interpolated N dep were distributed around the 1:1 line. The regression model between the original and interpolated N dep had the regression coefficient (0.96) closer to 1 and a high R 2 value. This indicated that there were close distributions between interpolated N dep values and true N dep values for the 41 testing data. The Q-Q plot of the N dep from Zhu et al. [4] and Lu and Tian [14] versus the 41 testing data were also described in Fig 3. The N dep by Lu and Tian [14] also obtained high accuracy, with low RMSE and high R 2 values.
On a provincial scale, comparison of the results of N dep (kg ha -1 a -1 ) in this study with those by Jia et al. [25] and Lu and Tian [14] is shown in Fig 4. Good agreements were also found for Fig 3. Comparison of N dep (kg ha -1 a -1 ) monitored in 41 sites with the estimation results in this study, by Jia et al. [25] and Lu and Tian [14] (x-axis was the testing data in the work by Zhu et al. [4], y-axis was the results estimated in this study (a), by Jia et al. [25] (b), Lu and Tian [14] (c)). Note: a regression cofficient closer to 1.00, a higher R 2 value indicate more reliable results of interpolation. the comparison of N dep with the results by Lu and Tian [14], giving confidence in the analysis of spatial pattern of N dep in China. This also confirmed that our results were more consistent with that by Lu and Tian [14] than that by Jia et al. [25].
On a national scale, to further explore the accuracy assessment of the spatial N dep distribution, we compared our results with that by Lu and Tian [14] using the data of provided 74 monitored sites by Du and Liu [34] (Fig 5J). There were four hotspots on the N dep map in this study, namely the North China Plain or Jing-jin-ji region, the Yangtze River Delta, Sichuan Basin and the Pearl River Delta. We suspected that Lu and Tian had underestimated slightly in Jing-jin-ji region, which should have the considerable magnitude of N dep with three other hotspots ( Fig 5J). However, Du and Liu [34] could not determine the magnitude of N dep in Middle Yangtze region including Anhui province and in the south of Middle Yellow region including Henan province due to no data monitored. The work by Lu and Tian [14] reported this region also had high N dep and we confirmed this hotspot in our study.
It should be noticed that this study might overestimate the N dep on a national scale, since most of the monitoring sites used in these published papers in China were distributed in developed areas, which would overestimate N dep on a national scale [5]. Also, there are some uncertainties in the estimation of N dep in China, which resulted from different concepts, sampling procedures, analysis methods and scaling-up methods. The effects of scaling-up method on national scale results require further study and the observation network for N dep needs to be strengthened to decrease the uncertainty.

Spatial pattern of N dep in China
The average of wet deposition flux of NH 4 + -N was 6.83 kg ha -1 a -1 with a standard deviation (STDEV) of 5.15, while the NO 3 --N was 5.35 kg ha -1 a -1 with a STDEV of 5.71. The average of ratio of NH 4 + -N/NO 3 --N was 1.28, which was slightly higher than the averaged ratio (1.22) in China, concluded by Zhu et al. [4]. The ratio of NH 4 + -N/NO 3 --N was widely considered a proxy for the sources of atmospheric reactive N [4,35,36]. Agricultural activity is the main source of N dep if the ratio is higher than 1, whereas, industrial activity is the main source if this ratio is lower than 1. The ratio of NH 4 + -N/NO 3 --N in this study indicated both the agricultural and industrial activities collectively influence the deposition of atmosphere N. The N dep was 12.18 kg ha -1 a -1 and the total N deposition in China would be 20.30 kg ha -1 a -1 assuming that the contribution of dry deposition was about 40% in China [4,37]. The magnitude and spatial pattern of N dep differed significantly in different regions in China (Fig 5I). Both NH 4 + -N and NO 3 --N peaked in central southern and southeastern China which are characterized by rapid industrial development and intensive N fertilizer applications [14]. N dep exhibited a decreasing gradient from the southeast to the northwest of China. The red line (Fig 5) indicated the significant heterogeneity in the levels of economic development for different regions, which resulted in a matching spatial heterogeneity in N dep across China. Similar results were also found in the study by Jia and Liu [3,25]. The low N dep were in areas including Qinghai-Tibet Plateau, Inner Mongolia and northwest China, where had not well developed industrial activities [5].
High N dep occurred across the south of Middle Yellow region, the North Coastal region and the middle and lower reaches of Yangtze River Basin (Fig 5G, 5H and 5I), which was in good agreement with the results by Lu and Tian (Fig 5D and 5E), but much different with that by Jia et al. (Fig 5F) [14,25]. Jia et al. [32] did not found the hotspots of N dep in the south of Middle Yellow region including Henan and Shaanxi provinces and in the North Coastal region including Beijing, Tianjin, Hebei and Shandong provinces. Du and Liu [34] also concluded high N dep in the North Coastal region including Beijing, Tianjin, Hebei and Shandong provinces (Fig 5J) [34] in good agreement with our findings. Jia et al. [25] maybe have underestimated N dep in the North Coastal region due to the uncertainty resulting from the limited number of data and analysis method in this area. Liu et al. [26] believed that Zhu et al. [4] (Fig 5A, 5B and 5C) might underestimate the dissolved N deposition throughout China due to the uncertainty from limited number of samples (41 sites), and the storage in their studies [26]. This study also confirmed that Zhu et al. underestimated N dep in the Southwest region including Chongqing and Guizhou provinces and the results by Du and Liu, Lu and Tian confirmed this suspect.
In summary, there were five hotspots of N dep in China, including the North Coastal region, East Coastal region, Southwest region and South Coastal region, and Middle Yangtze. N dep exhibited a decreasing gradient from southern to western and to northern China. N dep was > 35 kg ha -1 a -1 in some provinces of southern China, such as Chongqing, Hunan, Hubei and Henan, whereas N dep in other provinces of southern China was about 20-35 kg ha -1 a -1 . N dep over northern, northeastern and northwestern China was about 10-20, 5-15, 0-10 kg ha -1 a -1 .
The N dep on a national scale ranged from 9.88 to 21.1 kg ha -1 a -1 (Table 1), showing strong spatial variations. The wet deposition flux of N dep (12.18 kg ha -1 a -1 ) in this study was much lower China into a developed area (East) and an undeveloped area (West) in view of the levels of economic development, resource consumption, and population [5]. than that (21.07 kg ha -1 a -1 ) based on the average of those data points to represent N dep status across the whole China [3]. It was a bit higher than that (9.88 kg ha -1 a -1 ) by Lu and Tian (2007) calculated from at 253 sites from 1990 to 2003, and it was close to the results by Jia et al. (13.87 kg ha -1 a -1 ), Lu and Tian (14.05 kg ha -1 a -1 ) and Zhu (13.18 kg ha -1 a -1 ). These similar studies all considered spatial variability and area-weighted contribution from high-and low-N deposited regions, which was critically important to generate estimation of N dep on a national scale [6,14].

Influencing factors of Precipitation (P), N fertilizer use (F N ) and energy consumption (E) on the spatial patterns of N dep
The process of N dep is relatively clear in theory and has been applied in models, however, no agreement was reached upon how P, F N and E inflenced N dep . It is critical to understand the realationship between N dep and P, F N and E, to simulate and predict future trends in N dep assuming that the existing emission factors for F N and E don't change much. Several models have been developed to simulate the correlation of N dep and P, F N , E ( Table 2). Jia et al. found that N dep was linearly related to P and logarithmically to F N and E [25]. They believed that E, F N and P should be considered together when studying the factors that control the spatial pattern of N dep on the regional scale. N dep was calculated using equation N dep = a Ã ln((F N Ã 18.5%+E Ã 0.24%) Ã P)+b. However, Zhu and He reported N dep was exponentially related to P and E and linearly related to F N [4]. They thought that P and F N explain 80%-91% of the spatial variation of N dep , whereas E did not significantly explain the variability. A multiple linear regression model (N dep = a+b Ã F N +c Ã P) was applied without E by Zhu and He.
In this study, a strong exponential correlation was found between P and N dep , F N and N dep , E and N dep (Fig 6), which was in good agreement with that conducted by Zhu and He [4]. The models by Jia et al. (Fig 7A) and Zhu and He ( Fig 7B) were applied to predict N dep in China in Table 2. Comparison of different models used to simulate P, F N and E influencing spatial patterns of N dep N dep and P, F N and E N dep and P N dep and F N N dep and E Reference N dep = a*ln((F N *18.5%+E*0.24%)*P)+b y = a*P+b y = a+b*ln(F N ) y = a+b*ln(E) [25] N dep = a+b*F N +c*P y = a *P b y = a* F N +b y = a*E b [4] y = a *P+b y = a+b*ln(F N ) y = a+b*ln(E) [38] N dep = a+b*F N c +d*P e y = a*P b y = a*F N b y = a*E b This study a, b, c, d, e are regression coefficients; N dep represents N deposition; P represents precipitation (mm); F N represents fertilizer N use (t km -2 a -1 ); E represents energy consumption (t km -2 a -1 ).
doi:10.1371/journal.pone.0146051.t002 Fig 6. The effects of precipitation (mm), N fertilizer (t km -2 a -1 ) and energy consumption (t km -2 a -1 ) on the spatial pattern of N dep (kg ha -1 a -1 ). The mean N dep (kg ha -1 a -1 ) in provinces were obtained from spatial maps of N dep (kg ha -1 a -1 ) in China using Kriging. Inorganic Nitrogen (N) Wet Deposition this study. To improve this estimation of N dep , we established a new model to simulate this correlation based on a strong exponential correlation found (Fig 6). We agreed that E had little effect on the spatial pattern of N dep proposed by Zhu and He [4] through our practice in this study. Thus, we adopted an equation (N dep = a+b Ã F N c +d Ã P e ) to predict N dep and found a higher R 2 ( Fig 7C) compared with the results by Jia et al. (Fig 7A) and Zhu and He (Fig 7B). To confirm the effective of this new model, we used the data published by Jia et al. [25] to test whether this equation can reflect the spatial variation of N dep in China in 2000s and good agreement was found for the comparison of N dep with prediction ( Fig 7D).
It should be noted that we agree with E contributing much to the magnitude of decadal N dep in China [25], but had little effect on the spatial variation of N dep [4]. In summary, P, F N and E were all significantly correlated with the magnitude of N dep , P and F N contributed more than E to the spatial variation of N dep . It was critically essential to reduce E and F N to control reactive N emissions from fossil fuel combustion using maximum fessible reduction [4,22].
Ceratianly, we had to admit that there were some uncertainties in the analysis of how P, F N and E influencing the spatial patterns of N dep , which resulted from the limited statistical data obtained. The constructed analytical relationship was based on a provincial statistical data, and we believe that more data, such as municipal or county-level data, will obtain more reliable statistical models. However, it was too difficult to obtain such municipal or county-level data on both F N and E from the statistical yearbooks in China. The data on energy consumption (expressed as standard coal) on a municipal or county-level scale were not included in municipal or countylevel statistical yearbook and only the total energy consumption on a provincial scale could be obtained. Thus, we have to use the provincial statistical data to explore the correlation.

Anthropogenic sources of N dep in China
Detailed source contributions data are critical for policy makers to develop effective policies to protect Chinese terrestrial ecosystems [3]. Fossil fuel combustion and agricultural activities each sample. Excellent agreement was found, giving confidence that the PMF model captured the major sources and correctly quantified their contributions. The PMF model resolved five distinct sources (Fig 9). The first source had high K + , indicating a biomass burning (Fig 9A). The second source was enriched with SO 4 2and NO 3 - (Fig   9B), indicating a fossil fuel combustion source. The two icons were associated with NO x emitted from coal-fired power plants, residential heating and cooking, and motor vehicles [39]. The third source had a high loading of Ca 2+ and Mg 2+ , representing a crustal or windblown dust source ( Fig 9C). The profile also contained a significant SO 4 2indicating a great effect of neutralizing the acid [39]. The fourth source was dominated by NH 4 + suggesting an agricultural source ( Fig 9D). The fifth source had high loading of Na + and Cl -, a clear signal of sea salt impact ( Fig 9E). However, the profile also contained a significant SO 4 2-, a typical characteristic of aged sea salt.
The percentage contributions of each source to NH 4 + -N and NO 3 --N are shown in Fig 10. Fossil fuel combustion was the main contributor to NO 3 --N (86.0%). Biomass burning also contributed to 5.4% on the deposition of NO 3 --N. NH 4 + -N was mainly from agricultural activities (85.9%), fossil fuel combustion (6.0%) and Crust (7.2%). Overall, N dep in China may be considerably affected by the high emissions of NO x and NH 3 from fossil fuel combustion and agricultural activities and relevant studies will be presented in future papers.

Conclusion
The N dep throughout China was obtained by a method of Kriging, based on the N fluxes from the published papers from 2003 to 2014. The N dep map in our study showed close spatial pattern with that by Lu and Tian (2014). There were five hotspots of N dep across the North Coastal region, East Coastal region, Southwest region and South Coastal region, and Middle Yangtze, and exhibited a decreasing gradient from southeast to northwest of China. The wet deposition flux of NH 4 + -N, NO 3 --N and total N dep was 6.83, 5.35 and 12.18 kg ha -1 a -1 , respectively. A strong exponential correlation was found between P and N dep , F N and N dep and E and N dep , P and F N (80-91%) contributed more than E to the spatial variation of N dep . Fossil fuel combustion was the main contributor to NO 3 --N (86.0%) and biomass burning also contributed to 5.4% on the deposition of NO 3 --N. NH 4 + -N was mainly from agriculture (85.9%), fossil fuel combustion (6.0%). Our findings confirmed that the anthropogenic activities were the main reason for N dep increase and provided a scientific background for studies on ecological effects of N dep in China.
Supporting Information S1 PRISMA Checklist. The PRISMA 2009 Checklist (DOC) S1 Table. The information of the collected data records in this study. (XLSX)