Ground-based observations
The ground-based observations used in this study are spatially mapped in
Fig. 1. Detailed information for the observation period, including latitude
(∘ N), longitude (∘ E), elevation (meters above sea level,
a.s.l.), and the classification of each site, is listed in Table 1. The
observation dataset for the chemical composition of precipitation, which was
compiled by the Acid Deposition Monitoring Network in East Asia (EANET)
program (EANET, 2017), was mainly used in this study. In China, EANET
observations were obtained over three areas in southern China (Zhuhai,
Xiamen, and Chongqing) and one area in central China (Xi'an), for a total of
10 sites. In Korea, EANET observations from three sites (Cheju, Imsil, and
Kanghwa) were available. In Japan, data for a total of 11 sites were
available on EANET. Among these 11 sites, the data obtained at Ogasawara,
which is located in the northwest Pacific Ocean (27.09∘ N,
142.22∘ E, 230 m a.s.l.), includes estimates of the chemical
composition of precipitation that are below the detection limit, that is,
11.8 % for NO3- and 7.7 % for nss-SO42-. Therefore,
the data from Ogasawara were excluded. Samples were collected using a
wet-only sampler at a daily interval, except for Banryu (sampled weekly
during 2001–2015) and Ijira (sampled weekly in 2001). Concentrations of
NO3- and SO42- in precipitation were determined by ion
chromatography and qualified via ion balance and conductivity agreement. The
completeness of the data was determined from the precipitation coverage
duration and total precipitation amount (EANET, 2000, 2010). In Japan, two
sites (Ryori and Komae) were included in the study. Under the Global
Atmosphere Watch (GAW) program of the World Meteorological Organization (WMO
GAW, 2004), the Japan Meteorological Agency (JMA) has been conducting
observations of the atmospheric concentration and deposition at the northern,
remote Ryori site since 1976; observations of deposition at Ryori ended in
2011. Samples were collected using a wet-only sampler at a daily interval at
Ryori. The Central Research Institute of Electric Power Industry (CRIEPI)
conducted continuous monitoring of precipitation chemistry. The CRIEPI data
obtained at Komae, which is located near Tokyo, span 1987 to the present
(Fujita et al., 2000). As we have more fully described in our previous
studies (Itahashi et al., 2014a, 2015), CRIEPI monitoring was also conducted
at Goto Island, which is located
at the western edge of Japan. However, the period of coverage was limited to
February 2000 to April 2003, so we excluded the
Goto data from this study.
Samples were collected using a wet-only sampler at 10-day intervals in Komae.
For the EANET, JMA, and CRIEPI monthly mean datasets, the nss-SO42-
concentration (in mol L-1) was calculated from the conservative
assumption that sodium (Na+) is a sea salt tracer, using the following
equation:
nss-SO42-=SO42--0.06028×Na+
To analyze the long-term behavior of the dataset over the 15-year period of
2001–2015, outliers in the observation of EANET, JMA, and CRIEPI were
carefully examined according to method used in Itahashi et al. (2015). The
volume-weighted monthly mean concentrations of NO3-,
nss-SO42-, and Ratio in precipitation were analyzed using
the Smirnov–Grubbs outlier test at each site. In this study, the outliers
for Ratio were directly checked. In this method, outliers were
detected one at a time, assuming that the data likely followed an
approximately normal distribution. The hypothesis of no outliers is rejected
when
G=maxi=1,…,NCi-C‾s>N-1Nt(α/2N,N-2)2N-2+t(α/2N,N-2)2
where N, C‾, and s are the number, mean, and standard
deviation, respectively, of NO3- or nss-SO42- concentrations
in precipitation or Ratio in precipitation (Ci). t(α/2N,N-2)2 denotes the critical value of the t distribution with
N-2 degrees of freedom and a significance level of
(α/2N). Outlier detection and removal were iterated until the
dataset satisfied the specified significance level of 0.05. Following the
above criteria, 4.7, 3.8, and 4.4 % of data points for NO3-,
nss-SO42-, and Ratio in precipitation, respectively, were
discarded from the China dataset; 5.0, 3.0, and 3.0 % of these were
discarded from the Korea dataset; and 2.4, 1.5, and 1.6 % of these were
discarded from the Japan dataset. Finally, the annual mean concentrations of
NO3-, nss-SO42, and the annual mean Ratio in
precipitation were calculated from the monthly mean data when at least 9
months of data were available for a given year at the site. For the treatment
of the monthly accumulated precipitation amount, the Smirnov–Grubbs test for
Ratio was applied to discard outliers. If at least 9 months were
available, the annual accumulated precipitation was calculated.
These observation datasets taken from EANET, JMA, and CRIEPI were essentially
the same as those used in our previous studies (Itahashi et al., 2014a,
2015). A limitation of our previous studies was a lack of spatial coverage
over northern China, especially around the capital of Beijing because EANET
covered only the area from southern to central China (Fig. 1). It has been
recognized that anthropogenic emissions centered over this region (Kurokawa
et al., 2013; Li et al., 2017), the related atmospheric concentration, and
depositions are severe in China. To overcome this limitation and advance our
knowledge of precipitation chemistry over the whole of China, we evaluated
additional sources of data for the chemical concentration of precipitation
over China.
The Integrated Monitoring Program on Acidification of the Chinese Terrestrial
System (IMPACTS) was established through a Chinese–Norwegian cooperative
project (Larssen et al., 2004, 2006) from 2001 to 2003. Under the IMPACTS
program, atmospheric concentration, precipitation composition, and soil,
water, and vegetative effects were studied at five forested sites (LXH, LGC,
LCG, CJT, and TSP; refer to Fig. 1 and Table 1) over southern China. In terms
of deposition, four measurements (wet-only, bulk, throughfall collected below
the tree canopy, and belowground vegetation) were obtained. Wet-only sampling
data were used in this study. Observations of the chemical composition of
precipitation by a wet-only sampler were reported for four sites (LGS, LCG,
CJT, and TSP) from 2001 to 2003 and for one site (LXH) from 2002 to 2003.
Data for LGS in 2001 were not used, owing to insufficient coverage. At TSP,
continuous monitoring results up to 2013 have recently been reported (Yu et
al., 2017).
The precipitation chemistry over northern China has important implications.
The precipitation sample at Beijing Normal University (BNU) in 2003 (Sun et
al., 2004; Tang et al., 2005) and the recent work by the State Key Laboratory
of Atmospheric Boundary Layer Physics and Atmospheric Chemistry/Institute of
Atmospheric Physics/Chinese Academy of Sciences (LAPC/IAP/CAS) from
December 2007 to November 2010 (Pan et al., 2012, 2013; Pan and Wang, 2015;
Wang et al., 2012) were included in this study. Using the observation
framework from BNU, data for a total of 53 rain events were collected at
Beijing. To prevent contamination from dry deposition, the collector surface
was covered with a plastic lid. A detailed description of this collection
method is provided by Tang et al. (2005). In the recent work conducted by
LAPC/IAP/CAS, a 3-year observation from December 2007 to November 2010
was conducted at 10 sites around Beijing. Daily rainwater samples were
collected using a customized wet/dry automatic collector. The precipitation
sensor opened the collection funnel of the cover device when rainfall began.
In this study, December 2007 to November 2008 was regarded as the year 2008,
December 2008 to November 2009 was regarded as the year 2009, and December
2009 to November 2010 was regarded as the year 2010. For the IMPACTS and
LAPC/IAP/CAS datasets, the nss-SO42- concentration was calculated
using Na+ as a sea salt tracer based on Eq. (1). Owing to the lack of
information on Na+ within the samples at BNU, SO42- was used;
however, it was reported that Na+ is not a major ion component of these
samples.
Emission inventories
It is predicted that variation in NO3-, SO42-, and hence
Ratio is directly related to the emission of NOx and SO2. We
used the following emission inventories in this study. The Regional Emission
inventory in ASia (REAS) version 2.1 (Kurokawa et al., 2013), which covers
Asia from 2000 to 2008, was the main dataset used in this study to obtain
NOx and SO2 emissions over China, Korea, and Japan. The recent status
of Asian anthropogenic emissions was assessed by harmonizing different local
emission inventories with a mosaic approach named MIX (Li et al., 2017). MIX
covers the years 2008 and 2010. The data from MIX were also used to acquire
data for NOx and SO2 emissions over China, Korea, and Japan.
The latest country-level status of emissions in China is described by Xia et
al. (2016). Their emissions data cover the period from 2000 to 2014. The
primary case, which analyzed the contributions of advanced combustors with
improved energy efficiency and air pollutant control devices with improved
pollutant removal efficiency, was used in this study. In Korea, the National
Institute of Environmental Research (NIER) provided estimates of the national
emissions via the National Air Pollutants Emission Service, and the latest
data reported cover 1999 to 2013 (NIER, 2017). In Japan, the Japan Auto-Oil
Program (JATOP) provided 5-year interval emission datasets beginning in
1995 (JATOP, 2012a, b). The datasets for 2005 and 2010 were used in this
study.
Using the total amounts of NOx and SO2 emissions estimated via
inventories over China, Korea, and Japan, the NOx / SO2 emission
ratio on a molar basis was calculated by calculating NOx as NO2. The
behavior of the NOx / SO2 emissions ratio is correlated to
Ratio in precipitation.
Satellite observations
The emission inventories had some time lags. Accordingly, satellite
observations of the NO2 and SO2 vertical column density were
combined to capture the recent status of NOx and SO2. The
NOx / SO2 column ratio reflects the NOx / SO2
emission ratio and was effective for characterizing the correspondence with
Ratio in precipitation. Recently, satellite observations have been
widely used as a proxy for emissions data. The NO2 column has been used
to capture NOx emissions (e.g., Miyazaki et al., 2012; Mijling et al.,
2013; Itahashi et al., 2014b; Han et al., 2015; Irie et al., 2016), and the SO2 column has been used for SO2 emissions (e.g., Lee et al., 2011;
Li et al., 2010) and/or volcanic eruptions (e.g., Brenot et al., 2014).
Several studies have indicated the importance of different technologies to
control emissions (e.g., Li et al., 2010; Wang et al., 2015; Krotkov et al.,
2016; van der A et al., 2017). For example, the ratio of Ozone Monitoring Instrument (OMI)-derived SO2 / NO2 was used to determine the
effectiveness of the flue-gas desulfurization devices for power plants in
China (Li et al., 2010; Wang et al., 2015).
The NO2 and SO2 column dataset, which was observed by OMI onboard
the National Aeronautics and Space Administration (NASA) Earth Observing
System Aura satellite, was used in this study (NASA, 2017). The Aura
satellite was launched on 15 July 2004 in a sun-synchronous ascending polar
orbit with a local equator crossing time of 13:45. During the data period, it
measured sunlight backscattered from the Earth over a wide range of
ultraviolet and visible wavelengths to derive abundances of ozone and other
trace gases important for air quality and climate. Science-quality data
operations began on 1 October 2004; hence, the data from 2005 to 2015 were
used to cover our analysis period of 2001–2015. Retrieval algorithms were
based on the products provided by NASA.
In terms of the NO2 column, we used the most recent version of level 3
daily global nitrogen dioxide product (OMNO2d) of version 3.0, which was
released in August 2016 and is gridded at a resolution of
0.25∘ × 0.25∘ (Krotkov, 2013; Krotkov et al., 2017).
This product contains the total and tropospheric column for all atmospheric
conditions and for sky conditions where the cloud fraction is less than
30 %. We analyzed the tropospheric column with clouds screened on the
condition of a cloud fraction of less than 30 %.
In terms of the SO2 column, we used a level 3 daily global sulfur
dioxide product (OMSO2e) of the latest version (3.0), which was released in
February 2015 and is gridded at a resolution of 0.25∘ × 0.25∘ (Krotkov et al., 2015). The dataset contains the total
column of SO2 in the planetary boundary layer (PBL). The algorithm was based on a principal component analysis,
as introduced by Li et al. (2013). Cloud fraction, scene number, solar and
satellite viewing angles, and row anomaly flags were provided as ancillary
parameters. The data filtering of this level 3 dataset included the removal
of rows with any of the following: anomaly flags, radiative cloud fraction
greater than 20 %, solar zenith angle greater than 70.0∘, or
scene number greater than 58 or less than 3. In addition, we adopted the
smoothed method to average out the noise levels of the SO2 column,
following the research of Koukouli et al. (2016), who provided the
anthropogenic loading of SO2 over China as obtained from different
satellite sensors. This method smoothed the SO2 column assigned to each
of the 0.25∘ × 0.25∘ grid cells, which were weighted
by the SO2 column of the surrounding eight cells. In this process, the
negative values were regarded as zero values because our aim was to construct
the NOx / SO2 column ratio from satellite observations. Two data
periods associated with volcanic activities were excluded from our analysis
as follows. The Sarychev Volcano in the Kuril Islands (48.09∘ N,
153.20∘ E) had an explosive eruption that emitted a huge amount of
ash and SO2 at altitudes of 10–16 km (Brenot et al., 2014). The data
from 14 to 22 June 2009 included this large amount of SO2 (over
10 D.U.) in and around the analyzed domain; accordingly, they were discarded
from the calculations of monthly and annual means. An eruption of the Nabro
Volcano in Eritrea (13.37∘ N, 41.07∘ E) from 12 June to
7 July 2011 was also reported. During the night of 12 June 2011, this volcano
started to erupt, and on 14 June 2011, it spewed a volcanic plume across the
route of many flights over east Africa and the Middle East (Brenot et al.,
2014). The data from 15 June to 9 July 2011 were excluded from the
calculations of monthly and annual means according to the approach detailed
in van der A et al. (2017). In Japan, where many active volcanoes are
located, SO2 was continuously emitted at a level that surpassed
anthropogenic emissions (e.g., Itahashi et al., 2017a). Owing to the
difficulties of attempting to separate the effect of volcanic activity, the
SO2 column data for Japan, excluding the two data periods mentioned
above, were used.
Based on the daily gridded data of NO2 and SO2 columns, monthly
averages were calculated first and then the annual averages were calculated.
In the calculation of annual averages, cells with monthly averaged data not
available for at least 9 months were regarded as deficient cells for
consistency with the criteria adopted for the ground-based observations of
NO3-, nss-SO42, and Ratio in precipitation. The
NOx / SO2 column ratio was obtained from the annual averaged
gridded data for NO2 and SO2 columns.
Long-term temporal variation in (a) precipitation,
(b) NOx emission with the NO2 column,
(c) NO3- concentration in precipitation,
(d) SO2 emission with the SO2 column, and
(e) nss-SO42- concentration in precipitation during
2001–2015 over (i) China, (ii) Korea, and
(iii) Japan. Mean and 1 standard deviation across EANET observation
sites (JMA and CRIEPI sites are also included in Japan) are indicated by the
thick lines and shaded areas, respectively, in (a, c, e). For China,
EANET observation sites are denoted with dashed lines for urban sites and
dotted lines for rural and remote sites, and observation campaign data are
indicated by marks corresponding to marks on Fig. 1. The numbers at the
bottom indicate the number of years after 2000 (e.g., 5 indicates 2005 and 10
indicates 2010).
Satellite observations of the (left) NO2 column, (center) SO2 column, and (right)
NO2 / SO2 column ratio averaged over (a) 2006–2007
(the first half of Phase II), (b) 2010–2011 (transition between
Phase II and Phase III), and (c) 2014–2015 (the latter half of
Phase III). Dark-gray cells indicate places where annual mean could not be
calculated. If either the NO2 or SO2 column value is less than
1.0 × 1015 molecules cm-2, this is indicated by a
light-gray cell to clarify the NO2 / SO2 column ratio.
Statistical analysis of average values and linear trends
for precipitation, NO3- concentration
in precipitation, and nss-SO42-
concentration in precipitation over China, Korea, and Japan during phases I,
II, and III.
Phase I (2001–2005)
Phase II (2006–2010)
Phase III (2011–2015)
Mean
Trend
Mean
Trend
Mean
Trend
Precipitation (mm yr-1)
China
1330.2 (N=35)
-4.3±1.9
1390.2 (N=31)
+1.5±5.4
1158.3 (N=28)
+6.6±2.5a
Korea
1114.9 (N=11)
-3.9±15.4
1084.3 (N=15)
+6.9±3.0
1015.1 (N=15)
-11.2±3.3a
Japan
1493.5 (N=44)
-0.5±3.8
1591.7 (N=56)
+0.9±3.2
1555.7 (N=56)
-0.5±3.1
NO3- concentration in precipitation (µeq L-1)
China
51.3 (N=31)
-1.9±9.4
54.3 (N=31)
+0.6±5.5
57.3 (N=34)
-4.7±5.0
Korea
38.0 (N=11)
+3.0±4.1
40.8 (N=15)
-2.4±4.7
42.7 (N=15)
-1.1±2.5
Japan
17.2 (N=44)
-0.9±4.4
18.1 (N=55)
-3.4±1.3a
18.0 (N=55)
-0.6±0.8
nss-SO42- concentration in precipitation (µeq L-1)
China
237.0 (N=32)
+12.7±8.3
230.6 (N=32)
-20.3±8.8
171.5 (N=33)
-13.6±5.2a
Korea
62.2 (N=11)
+10.0±0.8c
66.9 (N=15)
-6.8±3.6
58.6 (N=15)
-4.2±2.6
Japan
29.8 (N=44)
-0.8±5.2
30.4 (N=56)
-5.4±2.9
27.8 (N=55)
-3.3±1.5
Note: total number of
observation sites are shown in parentheses. Linear trends were estimated
using linear regression and are shown as rates (% yr-1).
Significance levels are indicated by a p<0.05, b p<0.01, and c p<0.001, and lack of a mark indicates a lack of
significance. Analysis for Korea during Phase I was for the 2002–2005
period.
Annual changes based on the linear regression results for the (left)
NO2 column, (center) SO2 column, and (right)
NO2 / SO2 column ratio during (a) Phase II and
(b) Phase III. Dark-gray cells indicate areas where annual mean
calculation could not be performed. If either the NO2 or SO2 column
value is less than 1.0 × 1015 molecules cm-2, this is
indicated by a light-gray cell for clarity.
Long-term temporal variation in the
(a) NOx / SO2 emission ratio and NO2 / SO2
column ratio and (b) NO3- / nss-SO42-
concentration in precipitation (Ratio) during 2001–2015 over
(i) China, (ii) Korea, and (iii) Japan. Mean and 1
standard deviation across EANET observation sites (JMA and CRIEPI sites are
also included in Japan) are indicated by the thick lines and shaded areas,
respectively, in (b). For China, EANET observation
sites are denoted with dashed lines for urban sites and dotted lines for
rural and remote sites, and observation campaign data are indicated by marks
corresponding to Fig. 1. The numbers at the bottom indicate the number of
years after 2000 (e.g., 5 indicates 2005 and 10 indicates 2010).