A total of 14 chemical transport models (CTMs) participated in
the first topic of the Model Inter-Comparison Study for Asia (MICS-Asia)
phase III. These model results are compared with each other and an extensive
set of measurements, aiming to evaluate the current CTMs' ability in
simulating aerosol concentrations, to document the similarities and
differences among model performance, and to reveal the characteristics of
aerosol components in large cities over East Asia. In general, these CTMs
can well reproduce the spatial–temporal distributions of aerosols in East
Asia during the year 2010. The multi-model ensemble mean (MMEM) shows
better performance than most single-model predictions, with correlation
coefficients (between MMEM and measurements) ranging from 0.65 (nitrate,
Urbanization and industrialization have stimulated economic growth and
population expansion during the last several decades in East Asia (Spence et
al., 2008; Yan et al., 2016; Chen et al., 2016) but also brought about
noticeable degradation of ecological environment at the same time (Hall,
2002; Han et al., 2014; Yue et al., 2017). Significant increase in
atmospheric aerosol loading, especially from anthropogenic emissions, can
exert adverse effects on weather (Cowan et al., 2013), climate (H. Wang et al.,
2016), air quality (Y. Gao et al., 2016), and human health (Carmichael et
al., 2009). For example, aerosols can modify the thermodynamic structure of
the atmospheric boundary layer by absorbing and scattering solar radiation
(Ding et al., 2016; Petaja et al., 2016), alter cloud properties and
precipitation, by acting as cloud condensation nuclei and ice nuclei (Lohmann
and Diehl, 2006; Wang, 2013), deteriorate visibility, and cause haze events
(Singh and Dey, 2012; Li et al., 2014). In addition, fine particulate matter
with aerodynamic diameters smaller than 2.5
In order to better understand the properties of atmospheric aerosols and their impacts, chemical transport models (CTMs) can be a critical tool, and they have been applied to study various air-pollution issues all over the world. For example, a fully coupled online Weather Research and Forecasting model with chemistry (WRF-Chem) was developed by Grell et al. (2005), and it has been widely used to study the aerosol–radiation–cloud feedbacks on meteorology and air quality (Gao et al., 2014; B. Zhang et al., 2015; Qiu et al., 2017); a Community Multi-scale Air Quality (CMAQ) modeling system was designed by the US Environmental Protection Agency (Byun and Ching, 1999), and it has been applied to address acid deposition, visibility and haze pollution issues (Zhang et al., 2006; Han et al., 2014; Fan et al., 2015); a nested air quality prediction model system (NAQPMS) was developed by the Institute of Atmospheric Physics, Chinese Academy of Science (IAP/CAS) (Wang et al., 2001) to reproduce the mechanism of transport and evolution of atmospheric pollutants in Asia (Li et al., 2012; Z. Wang et al., 2013; J. Li et al., 2017); a global three-dimensional chemical transport model (GEOS-Chem) was first presented by Bey et al. (2001), and researchers use the GEOS-Chem model to study the source sector contribution, long-range transport, and the prediction of future change in ozone and aerosol concentrations (Liao et al., 2006; K. Li et al., 2016b; Zhu et al., 2017).
Although significant advantages can be found in CTMs, how to accurately reproduce or predict the concentrations and the distributions of atmospheric pollutants is still a challenge, with the problems of inaccurate emission inventories, poorly represented initial and boundary conditions, and imperfect physical, dynamical, and chemical parameterizations (Carmichael et al., 2008). Meanwhile, most CTMs are designed to focus on the air quality over developed countries, such as Europe and America, rather than Asia. The assumptions or look-up tables used in CTMs may not be suitable for the simulations of the East Asian environment (Gao et al., 2018). Therefore, before providing meaningful results and answering “what if” questions for policy makers, model performance must be carefully evaluated. Hayami et al. (2008) and Mann et al. (2014) pointed out that different parameterizations used in CTMs can cause large variations in simulation results, and the multi-model ensemble mean (MMEM) tends to show better performance than most single-model predictions when compared with observations (Carmichael et al., 2002; Hayami et al., 2008; Wang et al., 2008; Holloway et al., 2008). In order to develop a better common understanding of the performance and uncertainties of CTMs in East Asian applications, and to acquire a more mature comprehension of the properties of atmospheric aerosols and their impacts, a model intercomparison study should be initiated, and Model Inter-Comparison Study for Asia (MICS-Asia) gives an opportunity to investigate these questions. Meanwhile, model intercomparison study in East Asia is very limited (Phadnis et al., 1998; Kiley et al., 2003; Han et al., 2008), and far more efforts are needed in the future.
The MICS-Asia project was initiated in 1998. In the first phase of
MICS-Asia (MICS-Asia phase I), the primary target was to study the
long-range transport and deposition of
This paper focuses on the first topic of the MICS-Asia phase III and intends to present and summarize the following three objectives, specializing in the topic of aerosols. Firstly, comprehensive evaluations of the strengths and weaknesses of current CTMs for simulating particulate matter (PM) are provided against extensive in situ and satellite measurements, aiming to show the capability of participant models. Secondly, diversities of simulated aerosol concentrations among participant models are analyzed, including possible reasons for the inconsistency. Thirdly, characteristics of aerosol compositions in six metropolitan cities in East Asia are analyzed, which may be helpful to take measures to prevent and control air pollution in the future.
The description of model configurations, model inputs, and observations are presented in Sect. 2. The evaluation of model performance and the intercomparison between participant models are shown in Sect. 3. The conclusions and discussions are presented in Sect. 4.
A total of 14 regional models (M1–M14) participated in MICS-Asia phase III topic 1. All models were required to run for all of the year 2010, and provide gridded monthly simulation results of aerosols in the first model layer. These CTMs include the Weather Research and Forecasting model coupled with Community Multi-scale Air Quality (WRF-CMAQ), WRF-Chem, the nested air quality prediction model system (NAQPMS), the non-hydrostatic mesoscale model coupled with chemistry transport model (NHM-Chem), the Goddard Earth Observing System with chemistry (GEOS-Chem), and the Regional Atmospheric Modeling System coupled with CMAQ (RAMS-CMAQ). Among these models, there are three different versions of WRF-CMAQ (v5.0.2 is used by M1 and M2, v5.0.1 is used by M3, and v4.7.1 is used by M4, M5, and M6), four different versions of WRF-Chem (v3.7.1 is used by M7, v3.6.1 is used by M8, v3.6 is used by M9, and v3.5.1 is used by M10), one version of NAQPMS (M11), NHM-Chem (M12), GEOS-Chem (v9.1.3 is used by M13), and RAMS-CMAQ (v4.6 is used by M14). Basic information about the configurations of each model is summarized in Table 1.
Basic configurations of participant models in MICS-Asia phase III.
A unified simulation domain was designed by MICS-Asia organizers, which
covers the region of (15.4
Simulation domain for each participant model. The final analyzed region is also shown.
Gas-phase chemistry and aerosol chemistry are important parameterizations in CTMs. Luecken et al. (2008) and Balzarini et al. (2014) pointed out that different settings of chemical mechanisms could influence the simulation results significantly.
The gas chemistry of SAPRC99 (Statewide Air Pollution Research Center
99) was used in M1, M2, M4, M5, M6, M12, and M14. It is a detailed mechanism
for the gas-phase atmospheric reactions of VOCs and The Carbon Bond mechanism (CB05) was used in M3. It describes
tropospheric oxidant chemistry and provides a basis for computer modeling
studies of ozone, particulate matter, visibility, acid deposition, and air
toxicity issues, with 51 species and 156 reactions (Yarwood et al., 2005). The second-generation Regional Acid Deposition Model (RADM2) gas-phase
chemical mechanism was used in M9 and M10. The inorganic species considered
in RADM2 include 14 stable species, four reactive intermediates, and three abundant
stable species. The organic chemistry is represented by 26 stable species
and 16 peroxy radicals (Stockwell et al., 1990). This module can simulate
the concentrations of PAN, Based on RADM2, the Regional Atmospheric Chemistry Mechanism (RACM) was
developed with updated reaction rate constants and product yields according
to more recent laboratory measurements. It is capable of simulating the
troposphere from the Earth's surface through the upper troposphere and is
valid for simulating remote to polluted urban conditions (Stockwell et al.,
1997). M7 and M8 selected the RACM module. The rate coefficients were
further updated in M7 (Kim et al., 2009). However, heterogeneous hydrolysis
of The gas chemistry of Carbon Bond mechanism version Z (CBMZ) was used in
M11. This lumped-structure mechanism extends the original framework of
CBM-IV to function properly at larger spatial and longer timescales, with
revised inorganic chemistry, isoprene chemistry, and many other related
parameterizations (Zaveri and Peters, 1999). In M13, the
Jimenez et al. (2003), Luecken et al. (2008), and Yang et al. (2018)
summarized that different gas-phase chemistry mechanisms could predict
large variations in reactive species, such as
AERO with ISORROPIA: aerosol modules (AERO5 and AERO6) with
thermodynamic equilibrium models (ISORROPIA v1.7 and v2) were used in M1,
M2, M3, M4, M5, M6, M11, M12, and M14. Aerosols in AERO were divided into
three modes: Aitken, accumulation, and coarse modes. Gas–liquid–solid
equilibrium in inorganic aerosol was predicted by the ISORROPIA model. The
AERO5 ISORROPIA (v1.7) was mainly used in CMAQ v4, and the updated AERO6
ISORROPIA (v2) has been implemented since CMAQ v5. Overall, nine new PM species (e.g.,
MADE/SORGAM and MADE/VBS: detailed treatments of inorganic aerosol
effects in M7, M8, and M9 were simulated by Modal Aerosol Dynamics Model for
Europe (MADE). Three log-normal modes (Aitken, accumulation, and coarse
modes) were used in this module to present the particle size distribution of
submicrometer aerosol, such as GOCART: the Goddard Chemistry Aerosol Radiation and Transport (GOCART)
model was used in M10 to simulate tropospheric aerosol components, such as
Different chemical species are considered in numerous aerosol equilibrium
models, resulting in different equilibrium partitioning and water uptake
during the simulation processes, which can affect the predicted aerosol
concentrations (Fountoukis and Nenes, 2007). As Moya et al. (2002) and Wang
et al. (2012) classified that the treatment of crustal material in aerosol
chemistry could considerably improve model results in predicting the
partitioning of
Natural emissions of windblown dust have been explicitly parameterized since CMAQ v5 (Foroutan et al., 2017), but all the participating WRF-CMAQ models did not turn this option on, which means dust aerosols were not considered in M1–M6. Meanwhile, the dust scheme in M7 and M8 was also turned off.
Dust particles in M10 and M13 were simulated by the GOCART model (Ginoux et
al., 2001). This model includes eight size groups of mineral dust ranging
from 0.1 to 10
A simplified dust emission parameterization proposed by Shao (2001) was used
in M9 (Shao, 2004). Dust emission in Shao (2004) is
proportional to streamwise saltation flux, and the proportionality depends
on soil texture and soil plastic pressure. The size-resolved dust flux goes
into four size bins, with diameters ranging from 1.95 to 20
A size-segregated dust deflation module proposed by Wang et al. (2000) was
used in M11. It was developed based on three major predictors (friction
velocity, surface humidity, and dominant weather system) and has been
successfully applied in many dust-related simulations (Wang et al., 2002;
Yue et al., 2010). The dust flux
An empirical dust emission mechanism based on the approach of Gillette and
Passi (1988) was used in M12 and M14 (Han et al., 2004). Dust flux can be
calculated through the following formula:
Different dust schemes will produce different dust emission fluxes over arid and semi-arid regions (Zhao et al., 2010; Su and Fung, 2015). Several factors, such as potential source regions, threshold friction velocity, size distribution, and other surface and soil-related parameters used in equations, can be the primary causes for the inconsistency, and the differences in simulated dust emissions will affect the characteristics of spatial–temporal variations of atmospheric aerosol particles.
As one of the major components of primary aerosols, sea-salt aerosols contributes to 20 %–40 % of secondary inorganic aerosols (SIAs) over coastal regions (Liu et al., 2015; Yang et al., 2016). These particles can provide surface areas for condensation and reaction of nitrogen and sulfur, making the simulated concentrations of SIAs more accurate (Kelly et al., 2010; Im, 2013).
In M12, the method of Clarke et al. (2006) was used to simulate the
sea-salt emissions as follows:
In other participating models (sea-salt emission is not considered in M7
and M8), sea-salt emissions were simulated online by using the algorithm
proposed by Gong et al. (2003). The density function
Based on the experience concluded from phase I and phase II, all 14 models in phase III topic 1, in principle, were required to use the “standard” meteorological fields, emission inventories, and boundary conditions in order to reduce the potential diversities caused by model inputs. But different data were selected by participant models. In this section, some basic information about the model inputs are described.
The “standard” hourly meteorological fields were simulated by WRF v3.4.1 with the initial and lateral boundary conditions taken from the National Centers for Environmental Prediction (NCEP) Final Analysis (FNL) data. Four-dimensional data assimilation nudging toward the NCEP FNL data was also adopted to increase the accuracy of simulated meteorological variables. The reference meteorological fields were only used in M1–M6 and M11. For M7, M8, and M9, the standard meteorological simulation was run by the same model (WRF), but feedbacks between meteorological variables and pollutants were also considered in these WRF-Chem models. For M10, the Modern-Era Retrospective analysis for Research and Applications (MERRA) reanalysis was used to drive the WRF (v3.5.1) model. The outputs from the Japan Meteorological Agency (JMA) NHM were used to initialize M12 (Kajino et al., 2012). M13 was driven by assimilated meteorological data from GEOS of NASA's Global Modeling and Assimilation Office (Chen et al., 2009; K. Li et al., 2016b). Although the meteorological initial and lateral boundary conditions were taken from the same NCEP FNL data, three-dimensional meteorological fields used in M14 were simulated by Regional Atmospheric Modeling System (RAMS) (Zhang et al., 2002, 2007; Han et al., 2009, 2013). Consequently, different meteorological fields used in the 14 participant models will cause different atmospheric circulation characteristics, which can further influence the spatial–temporal variation of air pollutants (Gao et al., 2018).
All participant models utilized the “standard” emission inventory,
including anthropogenic, biogenic, biomass burning, air and ship, and
volcano emissions, which was prepared by the emission group in MICS-Asia
phase III. The anthropogenic emission dataset over Asia, named MIX, was
developed by harmonizing five regional and national emission inventories
with a mosaic approach. These five inventories are REAS2 (REAS inventory
version 2.1 for all of Asia; Kurokawa et al., 2013), MEIC (the
Multi-resolution Emission Inventory for China developed by Tsinghua
University), PKU-
Two sets of the chemical initial and boundary conditions (CHASER and
GEOS-Chem) were provided by MICS-Asia phase III. The 3-hourly global CTM
outputs of CHASER (prepared by Nagoya University; Sudo et al., 2002a, b)
were run with 2.8
As is known to all that meteorological fields have significant influences on air quality. Meanwhile, atmospheric compositions can also affect weather and climate. As Gao et al. (2018) pointed out, different coupling methods between aerosols and meteorological variables can cause different simulation results.
In order to simulate the concentrations of air pollutants, meteorological models and chemistry transport models should be implemented either offline or online (Kong et al., 2015). Offline modeling implies that the CTM is run after the meteorological simulation is completed, which means the chemical impacts on meteorology are not considered. Online modeling allows coupling and integration of some of the physical and chemical components (Baklanov et al., 2014). According to the extent of online coupling, there are two ways of coupling: (1) online integrated coupling (meteorology and chemistry are simulated simultaneously in the same grid) and (2) online access coupling (meteorology and chemistry are independent, but information can be exchanged between meteorology and chemistry) (Baklanov et al., 2014). Among these participating models, M4, M5, M6, M12, M13, and M14 are offline models. M1, M2, M3, and M11 are online access models. M7, M8, M9, and M10 are online integrated models.
More details about the model configurations can be found in Table 1 and the other MICS-Asia phase III companion papers (Kong et al., 2019; Li et al., 2019).
Monthly observations of
As is known to all, China has been experiencing heavy air pollution with
high concentrations of fine particles. Recent studies highlighted the
importance of secondary aerosols in the formation of haze episodes (Liu et
al., 2013; Sun et al., 2016a; Chen et al., 2018). However, observations
(e.g.,
The Aerosol Robotic Network (AERONET), a ground-based remote-sensing aerosol network consisting of worldwide automatic Sun- and sky-scanning spectral radiometers (Holben et al., 1998), provides the aerosol optical depth (AOD) products at 440 and 675 nm, which can be used to calculate the AOD at 550 nm according to the Ångström exponent. The AERONET level 2.0 monthly AOD (cloud-screened and quality-assured) data at 33 sites were utilized in this study. Meanwhile, satellite-retrieved 550 nm AOD products from the Moderate Resolution Imaging Spectroradiometer (MODIS) were also used to compare with simulations.
Figures 2 and S3 show the geographical locations of all the
observation sites. Most
The geographical locations of observation stations: EANET (shown in black circles; the number of stations is 39), CNEMC (shown in red triangles; the number of stations is 32), others (observations collected from published literature, shown in purple stars; the number of stations is 32), and AERONET (shown in black boxes; the number of stations is 33). Five defined subregions (Region_1 to Region_5) are also shown.
In general, the wide variety of in situ and satellite measurements used in this paper can allow for a rigorous and comprehensive evaluation of model performance.
According to the objective of MICS-Asia phase III topic 1, comparisons of
aerosol concentrations between observations and simulations are presented to
evaluate the performance of current multi-scale air quality models in East
Asia, including analyzing the similarities and differences between
participant models. Simulation results of BC, OC,
Figure 3 illustrates the observed and simulated ground-level annual mean
concentrations of BC,
Comparison of observed and simulated concentrations of
Analyzing Fig. 3a, we can find that most models show good skills in
simulating the BC concentrations and their spatial distribution
characteristics, with relative high values over large emission areas (e.g.,
north China) (K. Li et al., 2016a). But the NMB for MMEM is
For
For
Simulated
On average, the observed PM
For PM
Time series of the monthly observed and simulated aerosol compositions,
including BC,
Time series of the monthly observed and simulated aerosol
compositions:
The same as Fig. 4 but for
The measured BC concentrations in Region_2 exhibit an obvious
seasonal variation, with the minimum (
For PM
Similar temporal-variation characteristics of PM
The seasonal variation characteristics of observed
As mentioned above, the observed monthly mean concentrations of aerosol
compositions in China are only available at one EANET station (site 17, the
Hongwen station), with missing values in June and October. In order to make
the evaluation more comprehensive, observed seasonal mean concentrations of
Simulated AODs at 550 nm from the nine participant
models (M1, M2, M4, M7, M9, M11, M12, M13, and M14) are compared with the
measurements from AERONET. From Fig. 6, we can find that most models tend to
overpredict AOD values during the whole simulation period in
Region_1, Region_2, and Region_3
with NMBs of 74.0 %, 38.8 %, and 107.0 % for MMEM, respectively. In
Region_4, an obvious seasonality is observed, with the maximum
in spring and the minimum in summer. Models can capture this seasonality
well, although underestimation is found in spring. The
Similar to Fig. 4 but for seasonal cycles of AOD at 550 nm. In this figure, the monthly measurements are taken from AERONET.
Spatial distributions of observed and simulated AODs at 550 nm. The observed AOD values are retrieved from MODIS. Spatial correlation coefficients are given in the bottom left corner of each panel. Observed AODs from AERONET are also shown in circles.
Figure 7 presents the spatial distributions of the observed and simulated
AOD at 550 nm. MODIS AOD is collected from the Terra and Aqua satellites
during the year 2010. The observed AODs from AERONET are also shown. In order
to quantify the ability of each model to simulate the spatial distribution
of aerosol particles, spatial correlation coefficients are also given in the
bottom left corner of each panel. Analyzing the observations from MODIS, we
can conclude that AOD values are higher in central and eastern China,
including the Sichuan province, with the maximum over 1.0. High values can
also be observed in the north of India. Due to dust events happening in arid and
semi-arid regions, AOD values over the Taklimakan are also large
(
Table 2 shows the statistics of correlation coefficient (
Statistics of BC,
It can be found that participant models are able to capture the variability
of BC in China, with
The main purpose of MICS-Asia phase III topic 1 is to assess the ability of
current multi-scale air quality models to reproduce the air-pollutant
concentrations in East Asia. In order to reveal the improvements of the
simulation ability in current CTMs, statistics (e.g., RMSE and
Intercomparison of model performance between MICS-Asia phase II
(blue) and phase III (red) for
The statistics of MICS-Asia phase II are taken from Hayami et al. (2008).
The observed monthly mean concentrations are monitored with high
completeness at the 14 EANET stations in March, July, and December 2001,
and March 2002, and the model-predicted monthly surface concentrations are
from eight regional CTMs. Notably,
Analyzing the RMSEs in Fig. 8, we can conclude that the medians (the
25th percentile, the 75th percentile) for
Although the participating models (8 versus 12 CTMs), observation sites (14 versus 31 EANET stations), and simulation periods (4 months versus 1 year) are different between phase II and phase III, more reasonable statistics are calculated by current CTMs, reflecting better performance in simulating the concentrations of aerosols and their spatial–temporal variations.
Figure 9 shows the spatial distributions of simulated PM
Spatial distributions of simulated PM
Previous studies have revealed that sulfate, nitrate, and ammonium (denoted
as SNA) are the predominant inorganic aerosols in PM, and SNA can contribute
to nearly half of the total PM
SOR and NOR can be used to estimate the degree of secondary formation of
PNR is defined as the mole ratio of ammonium to sulfate and nitrate. When
PNR is larger than unity, sufficient ammonia can be used to neutralize the
acidic sulfate and nitrate; otherwise, there is an incomplete neutralization
of acidic species. Analyzing the calculated PNRs from participant models,
all values are smaller than 1, which means atmospheric conditions are
considered to be ammonia deficient. But the mole ratios of
The same as Fig. 9 but for PM
However, a large CV (> 1.0) is simulated over arid and semi-arid
regions (Fig. 9), such as the Taklimakan Desert and the Gobi Desert, where
dust events are often observed, which means current CTMs have difficulty
processing dust aerosols, especially in producing a similar amount of dust
emissions and in identifying the same potential dust source regions, by
using different dust schemes. Large CVs are also shown in simulated coarse
particles (subtract PM
The CV (standard deviation divided by the
mean) of simulated coarse particles (subtract PM
From Table 3 we can conclude that the low consistency (or the large CV) of simulated coarse particles in each defined subregion is mainly caused by the dust particles. Without the impacts of dust aerosols and sea salts (only simulation results from M7 and M8 are considered), the calculated CVs for Region_1 to Region_5 are 0.29, 0.30, 0.33, 0.19, and 0.10, respectively. Without the impacts of dust aerosols (only simulation results from M1, M2, M4, M5, and M6 are considered), similar spatial distributions are found in Fig. 10, and the CVs averaged over each subregion are 0.37 (Region_1), 0.65 (Region_2), 0.48 (Region_3), 0.59 (Region_4), and 0.65 (Region_5), respectively. But when the influences of dust aerosols and sea salts are both considered (simulation results from M9, M11, M12, and M14 are used), larger CVs are obtained with values of 0.97 for Region_1, 1.04 for Region_2, 1.27 for Region_3, 0.95 for Region_4, and 0.88 for Region_5.
Aerosol chemical compositions simulated by each participant model and the
MMEM in the six metropolitan cities (Beijing, Shanghai, Guangzhou, Delhi, Seoul,
and Tokyo) are shown in Fig. 11. PM
High values of PM
Analyzing the ratios of aerosol compositions to PM
For seasonal variations of PM
This paper mainly focuses on the first topic of the MICS-Asia phase III, and intends to analyze the following objectives: (1) provide a comprehensive evaluation of current air quality models against observations, (2) analyze the diversity of simulated aerosols among participant models, and (3) reveal the characteristics of aerosol components in large cities over East Asia.
Comparisons against monthly observations from EANET and CNEMC demonstrate
that all participant models can well reproduce the spatial–temporal
distributions of aerosols. The MMEM shows
better performance than most single-model predictions, with correlation
coefficients (
In order to reveal the improvements of the simulation ability in current
CTMs, statistics for observed and simulated
Analyzing the ratio of SNA to PM
The coefficient of variation (CV) can be used to quantify the intermodel
deviation, and a large CV is shown in simulated coarse particles (subtract
PM
According to the MMEM simulation results, the highest PM
The MICS-Asia project gives an opportunity to understand the performance of
CTMs in East Asian applications, including the similarities and differences
among air quality models. In order to quantify the impacts of different
model inputs and model configurations, and to reduce the diversities among
simulation results, more detailed sensitivity experiments should be
discussed. For example, simulation results from M1 and M2 can be used to
assess the impacts of boundary conditions, since the configurations in
these two models are similar except the boundary conditions. M1 adopts the downscale results
from GEOS-Chem, while M2 uses the default values from CMAQ. From Fig. S9, we
can find that positive biases are simulated
Monthly pollution concentrations at EANET stations can be collected from
The supplement related to this article is available online at:
LC, YG, and MZ conducted the study design. LC, JZ, HL, JL, KH, BG, XW, YFL, CL, SI, TN, MK, and KY contributed to modeling data. JSF, ZW, and JK provided the emission data and observation data. YG and JZ helped with data processing. MZ, JSF, and JZ were involved in the scientific interpretation and discussion. LC prepared the manuscript with contributions from all co-authors.
The authors declare that they have no conflict of interest.
This article is part of the special issue “Regional assessment of air pollution and climate change over East and Southeast Asia: results from MICS-Asia Phase III”. It is not associated with a conference.
The authors thank the anonymous reviewers for their helpful comments that improved the paper.
This research has been supported by the National Key R&D Programs of China (2017YFB0503901 & 2016YFA0600203), the National Natural Science Foundation of China (41830109, 91544221 & 91644215), the University Natural Science Research Foundation of Jiangsu Province (18KJB170012), the Environment Research and Technology Development Fund (S12-1) of the Ministry of the Environment, Japan, the Startup Foundation for Introducing Talent of NUIST (2018r007), and the Decision-making Consultation Research Foundation of RICEG, NUIST (2018B33).
This paper was edited by Gregory R. Carmichael and reviewed by two anonymous referees.