Investigation of Spatiotemporal Variation and Drivers of Aerosol Optical Depth in China from 2010 to 2020

Abstract

China has experienced rapid economic growth and serious control of aerosol emissions in the past decade. Thus, the spatiotemporal variations and driving factors of aerosol optical depth (AOD) are urgently needed to evaluate the effectiveness of aerosol control activities. The innovation of this study is a detailed spatial and temporal analysis of aerosol pollution in eight major regions of China from 2010 to 2020 using the MERRA-2 AOD reanalysis product and the driving mechanism based on the Granger causality test, sensitivity, and contribution analysis. The results show that the spatial distribution of AOD varied across the areas. Divided by the Hu Line, the AOD values of the Eastern areas were significantly higher than those of the Western areas. The temporal trend in the last eleven years was dominated by a continuous decline and moderate fluctuations at both annual and seasonal scales. The relationship between socioeconomic factors and AOD drivers was more significant in economically developed regions, suggesting that China pays more attention to haze control while developing its economy. The driving relationship between AOD and temperature was weak, while wind speed and relative humidity were more influential. For vegetation factors, Granger effects were mainly observed in the Northeast, Beijing-Tianjin-Hebei, Guangdong, Central China, and Southwest regions. In the Guangdong and Southwest regions, vegetation and economic factors were the more influential drivers. This study provides a scientific basis for the detection of aerosol changes, driving mechanisms and pollution management in China.

Full text

Citation: Wang, Y.; Yang, L.; Xie, D.;
Hu, Y.; Cao, D.; Huang, H.; Zhao, D.
Investigation of Spatiotemporal
Variation and Drivers of Aerosol
Optical Depth in China from 2010 to
2020. Atmosphere 2023,14, 477.
https://doi.org/10.3390/
atmos14030477
Academic Editors: Hongmin Zhou,
Tao He, Xiaodan Wu and Ying Qu
Received: 19 January 2023
Revised: 15 February 2023
Accepted: 22 February 2023
Published: 28 February 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
atmosphere
Article
Investigation of Spatiotemporal Variation and Drivers of
Aerosol Optical Depth in China from 2010 to 2020
Yiting Wang 1, 2, * , Lixiang Yang 1, Donghui Xie 2,3 , Yuhao Hu 1, Di Cao 4, Haiyang Huang 1and Dan Zhao 1
1College of Geomatics, Xi’an University of Science and Technology, Xi’an 710054, China
2State Key Laboratory of Remote Sensing Science, Jointly Sponsored by Beijing Normal University and
Aerospace Information Research Institute of Chinese Academy of Sciences, Beijing 100875, China
3Beijing Engineering Research Center for Global Land Remote Sensing Products, Faculty of Geographical
Science, Beijing Normal University, Beijing 100875, China
4China Construction Eighth Engineering Division Corp., Ltd., Shanghai 200131, China
*Correspondence: wyt@xust.edu.cn
Abstract:
China has experienced rapid economic growth and serious control of aerosol emissions in the
past decade. Thus, the spatiotemporal variations and driving factors of aerosol optical depth (AOD) are
urgently needed to evaluate the effectiveness of aerosol control activities. The innovation of this study is
a detailed spatial and temporal analysis of aerosol pollution in eight major regions of China from 2010
to 2020 using the MERRA-2 AOD reanalysis product and the driving mechanism based on the Granger
causality test, sensitivity, and contribution analysis. The results show that the spatial distribution of AOD
varied across the areas. Divided by the Hu Line, the AOD values of the Eastern areas were significantly
higher than those of the Western areas. The temporal trend in the last eleven years was dominated by
a continuous decline and moderate fluctuations at both annual and seasonal scales. The relationship
between socioeconomic factors and AOD drivers was more significant in economically developed regions,
suggesting that China pays more attention to haze control while developing its economy. The driving
relationship between AOD and temperature was weak, while wind speed and relative humidity were
more influential. For vegetation factors, Granger effects were mainly observed in the Northeast, Beijing-
Tianjin-Hebei, Guangdong, Central China, and Southwest regions. In the Guangdong and Southwest
regions, vegetation and economic factors were the more influential drivers. This study provides a scientific
basis for the detection of aerosol changes, driving mechanisms and pollution management in China.
Keywords:
MERRA-2; AOD; AERONET; M–K test; Granger causality test; sensitivity; contribution;
spatiotemporal analysis
1. Introduction
Aerosol refers to a variety of small solid and liquid particles suspended in the at-
mosphere with sizes ranging from 0.1–100 nm [
1
–
3
]. By absorbing and scattering solar
radiation, aerosols affect the radiation budget between Earth and the atmosphere and con-
sequently influence terrestrial carbon transfer [
4
–
6
]. Moreover, aerosols, as cloud condensa-
tion nuclei, directly influence the formation and microphysical characteristics of clouds,
such as cloud albedo, which can further induce changes in weather and climate [
2
,
4
,
7
,
8
].
Heavy aerosol loadings can cause air pollution, reduce visibility, and increase the risk of
respiratory diseases in humans [2,9–13].
The aerosol optical depth (AOD) is computed as the integral of the solar extinction co-
efficient in the vertical direction based on the weakening of incident solar light by aerosols.
Therefore, AOD is a fundamental property indicating the concentration of aerosols and a key
physical quantity characterizing the degree of atmospheric turbidity [
14
–
16
]. Recent develop-
ments in aerosol-related research have shown an increasing need for AOD data with high spatial
and high temporal resolutions [
17
,
18
]. Traditional methods to retrieve AOD include ground
measurements and satellite remote sensing. The former is limited by the sparse distribution
Atmosphere 2023,14, 477. https://doi.org/10.3390/atmos14030477 https://www.mdpi.com/journal/atmosphere

Atmosphere 2023,14, 477 2 of 23
of aerosol observation networks such as AERONET [
19
,
20
], which makes it difficult to obtain
aerosol information on a large scale. The latter is limited by the availability and quality of
the acquired satellite data [
21
,
22
], such as the influence of clouds. In contrast, reanalysis AOD
products are particularly advantageous for providing spatially and temporally consistent AOD
information derived from data assimilation systems that integrate global climate models and
multisource AOD measurements [
17
]. Previous studies have successfully used reanalysis AOD
products to investigate the spatiotemporal changes in AOD [
23
–
25
], such as the Modern-Era
Retrospective Analysis for Research and Applications Aerosol (MERRA) reanalysis product [
25
],
yet the accuracy of the reanalysis product needs to be carefully evaluated.
As a direct indicator of air pollution, the dynamics and drivers of AOD have been the
focus of research [
26
,
27
]. Numerous studies have attempted to investigate the spatial and
temporal changes in AOD, as well as the driving forces, in hazy areas using satellite-derived
or reanalysis AOD products [
28
,
29
]. Existing studies have shown that human activities
such as industrial production, transportation, straw burning, and energy consumption
are important factors influencing AOD [
30
]. In addition, air quality is not only related to
anthropogenic aerosol emissions but is also influenced by meteorological conditions. While
pollution sources are relatively stable, meteorological factors, including temperature (TEM),
relative humidity (RHU), and wind speed (WIN), have important effects on the evolution
of aerosols [
31
]. In addition, vegetation can effectively absorb harmful particles in the
atmosphere, such as sulfur dioxide. The inverse relationship between vegetation cover and
dust emissions is undeniable, with varying levels of influence in different regions [
32
,
33
].
China is located across a wide latitude and under different environmental conditions, which
leads to each parameter having different effect levels under different conditions [
14
,
30
,
34
].
The methods commonly used for driving force analysis mainly include correlation or re-
gression analysis [
6
,
35
,
36
], the Granger causality test, and sensitivity analysis [
37
,
38
]. However,
most regression-based analyses actually reveal a statistical relationship rather than a causal
relationship between AOD and related factors [
39
] since they cannot determine the direction
of causality between two related variables [
40
,
41
]. Granger causality tests reveal causal rela-
tionships between variables [
39
] and identify the direction of causality between two related
variables [
40
,
41
], but they do not provide a sufficiently deep quantitative description to measure
the strength of causality. Sensitivity analysis quantitatively describes the relationship between
the variables and the drivers [
42
], but it does not consider a causal relationship. Therefore, this
paper combines the Granger causality analysis and sensitivity analysis methods together to
investigate AOD driving mechanisms to enhance our understanding of AOD dynamics.
China has experienced rapid economic growth and serious air pollution in the past few
decades [
2
,
28
]. Since the 18th National Congress of the Communist Party of China (CPC)
convened in 2012, China has made the response to climate change a higher priority in state
governance and has adopted a series of policies, measures, and actions to tackle climate
change, and the mitigation of air pollution is one of the specific goals [
43
]. To verify the
effectiveness of air pollution control and provide guidance for further actions, this paper aims
to outline a detailed investigation of the spatiotemporal variations and driving mechanisms
of the AOD in mainland China from 2010 to 2020. The innovations of this study include (1)
using the MERRA-2 AOD reanalysis product with a 3 h temporal resolution to describe highly
dynamic variations in AOD and a detailed evaluation of the trend, robustness, and variability
of the changes in AOD, and (2) detailed analysis of the socioeconomic, meteorological, and
vegetation drivers combining the Granger causality test [
44
], sensitivity, and contribution
rates. This study provides an up-to-date analysis of the long-term trends in aerosols and their
driving factors in major areas in China, which will provide a reference for related studies on
environmental changes and for the ongoing clean air acts in China.
2. Materials and Methods
2.1. Study Area
To adapt to the differences in aerosols across space, regional analysis was performed
to identify long-term trends in representative areas in China. The China Atmospheric

Atmosphere 2023,14, 477 3 of 23
Environment Bulletin [
45
] identifies eight key regions that are close in economic level
and form industrial zones of a certain scale. At the same time, these eight regions have
different degrees of aerosol pollution and are the key areas for the Chinese government
to combat haze. This includes Beijing-Tianjin-Hebei, Fenway Plain, Yangtze River Delta,
Guangdong Province, Northeast China, Central China, Southwest China, and Northwest
China, which we will refer to as BTH, FWP, YRD, GD, NE, CC, SW, and NW hereafter,
as shown in Figure 1and Table 1. In addition, this paper divided mainland China into
two parts using the Hu Line, which divided the country into two parts with significantly
different population densities and socioeconomic development [
46
]. Changes in AOD in
different areas were analyzed using zonal analysis.
Atmosphere 2023, 14, x FOR PEER REVIEW 3 of 25
test [44], sensitivity, and contribution rates. This study provides an up-to-date analysis of
the long-term trends in aerosols and their driving factors in major areas in China, which
will provide a reference for related studies on environmental changes and for the ongoing
clean air acts in China.
2. Materials and Methods
2.1. Study Area
To adapt to the differences in aerosols across space, regional analysis was performed
to identify long-term trends in representative areas in China. The China Atmospheric
Environment Bulletin [45] identifies eight key regions that are close in economic level and
form industrial zones of a certain scale. At the same time, these eight regions have
different degrees of aerosol pollution and are the key areas for the Chinese government
to combat haze. This includes Beijing-Tianjin-Hebei, Fenway Plain, Yangtze River Delta,
Guangdong Province, Northeast China, Central China, Southwest China, and Northwest
China, which we will refer to as BTH, FWP, YRD, GD, NE, CC, SW, and NW hereafter, as
shown in Figure 1 and Table 1. In addition, this paper divided mainland China into two
parts using the Hu Line, which divided the country into two parts with significantly
different population densities and socioeconomic development [46]. Changes in AOD in
different areas were analyzed using zonal analysis.
Figure 1. Map of Study Area.
Figure 1. Map of Study Area.
Table 1. Administrative districts of eight major areas.
Region Administrative District
Beijing-Tianjin-Hebei (BTH) Beijing city, Tianjin city, Hebei Province
Fenway Plain (FWP)
Shanxi Province, Shaanxi Province and Henan Provinces
Yangtze River Delta (YRD) Shanghai city, Jiangsu Province, Zhejiang Province,
Anhui Province
Guangdong province (GD) Guangdong Province
Northeast (NE)
Heilongjiang Province, Jilin Province, Liaoning Province
Central China (CC) Hunan Province, Hubei Province, Jiangxi Province
Southwest (SW) Yunnan Province, Sichuan Province, Guizhou Province,
Chongqing city
Northwest (NW) Shaanxi Province, Gansu Province, Ningxia Province,
Xinjiang Province

Atmosphere 2023,14, 477 4 of 23
2.2. Data Description and Processing
2.2.1. MERRA-2 AOD Data
The MERRA-2 reanalysis product (available at https://disc.gsfc.nasa.gov, accessed on
15 May 2022) is derived from a data assimilation system that integrates the GEOS-5 climate
model and a variety of satellite- and ground-observed AOD data, such as the satellite-
derived AOD data from AVHRR, MISR, and MODIS sensors and ground measurements
from AERONET. It provides 21 categories of data from 1980 to the present, including water
vapor, radiation, AOD, etc. [
25
]. MERRA-2 AOD data with a 0.5
◦×
0.625
◦
spatial resolution
and a 3 h temporal resolution from 2010 to 2020 in mainland China were used in this study.
Daily averages of MERRA-2 AOD data at 3 h intervals were calculated for subsequent
analysis.
2.2.2. AREONET AOD Data
AERONET (available at http://aeronet.gsfc.nasa.gov, accessed on 18 May 2022)) is
a global ground-based network that provides long-term measurements of aerosol optical
properties at wide-range wavelengths from 340 to 1640 nm with a high temporal resolu-
tion [
47
]. AERONET uses the CE-318 sun photometer to measure and provide data on
aerosol optical, microphysical, and radiometric properties for more than 25 years [
7
,
17
,
26
].
AERONET aerosol data have three different levels, including level 1.0 unprocessed raw
data, level 1.5 cloud filtered and quality controlled data, and level 2.0 cloud screened and
quality assured data, corresponding to different sites [
28
]. A total of 87 AERONET sites in
China (Figure 1), measuring instantaneous AOD at a temporal resolution of less than one
hour, were used in this study. Since very few sites can provide complete records of AOD
data for 2010–2020, we included all AERONET data of levels 1.5 and 2.0 at the moments
near the time of MERRA-2 AOD data at 3 h intervals for validation.
To maintain consistency with the MERRA-2 AOD data at 550 nm, we first interpolated
the AERONET AOD measurements at 500 nm and 675 nm to 550 nm using:
AOD550 =β×550−α(1)



α=−ln(AOD500
AOD675 )
ln(500
675 )
β=AOD500
500−α=AOD675
675−α
(2)
where
α
is the Angstrom exponent, estimated from the slope of the spectral AOD plots
in logarithmic scales,
β
is the turbidity coefficient, AOD
500
means AOD at 550 nm, and
AOD675 means AOD at 675 nm.
2.2.3. Socioeconomic Data
As both aerosol emissions and controls are related to socioeconomic development, this
paper selected two socioeconomic indicators, gross domestic product per capita (GDPPC)
and industrial output per square kilometer (IOPSK), to investigate the feedback between
AOD changes and socioeconomic development. Data were collected from the China
Statistical Yearbook (http://www.stats.gov.cn/, accessed on 20 May 2022) from 2010 to 2020,
which provides academic statistics issued by the National Bureau of Statistics. However,
the China Statistical Yearbook has not publicized provincial industrial output data since
2020, thus the IOPSK data from 2010–2019 were included in our analysis.
2.2.4. Meteorological Data
This study collected daily meteorological data from the 699 primary meteorological
stations from the China Meteorological Administration website (https://cmdp.ncc-cma.net,
accessed on 22 May 2022). Wind speed (WIN) affects the velocity of airflow, temperature
(TEM) indicates the intensity of particle motion, and relative humidity (RHU) indicates the
percentage of water vapor pressure and saturated water vapor pressure in the air, with
variables that have a significant impact on AOD evolution. Therefore, we chose WIN, TEM,

Atmosphere 2023,14, 477 5 of 23
and RHU as the weather driving factor, as claimed in previous studies [
30
]. All daily meteo-
rological data were temporally aggregated to different time scales to adapt to the analysis of
the AOD spatiotemporal variations. Figure 2shows the distribution of the meteorological
sites and the seasonal mean values of the three factors. In our analysis, spring refers to
the months from March to May, summer from June to August, autumn from September
to November, and winter from December to February. All the meteorological data were
spatially interpolated to raster images using the Kriging interpolation method
[48–50]
and
then resampled to the raster cells of MERRA-2 AOD images for comparative analysis.
Atmosphere 2023, 14, x FOR PEER REVIEW 6 of 25
Figure 2. Seasonal averages of meteorological data from 2010 to 2020.
2.2.5. Vegetation Continuous Fields Data
The Terra MODIS Vegetation Continuous Fields (VCF) product (MOD44B) is a
subpixel-level representation of surface vegetation cover estimates globally. It is designed
to continuously represent Earth’s terrestrial surface as a proportion of basic vegetation
traits and provides a gradation of three surface cover components: percent tree cover,
percent nontree cover, and percent bare. Considering the large-scale reforestation
Figure 2. Seasonal averages of meteorological data from 2010 to 2020.

Atmosphere 2023,14, 477 6 of 23
2.2.5. Vegetation Continuous Fields Data
The Terra MODIS Vegetation Continuous Fields (VCF) product (MOD44B) is a subpixel-
level representation of surface vegetation cover estimates globally. It is designed to contin-
uously represent Earth’s terrestrial surface as a proportion of basic vegetation traits and
provides a gradation of three surface cover components: percent tree cover, percent nontree
cover, and percent bare. Considering the large-scale reforestation activities in China, this
study used the yearly percent tree cover data of MOD44B with 250 m spatial resolution in
China from 2010 to 2020 to evaluate the effect of increasing trees on AOD changes.
2.3. Methodology
Based on MERRA2-AOD data, this paper investigates the spatial and temporal evo-
lution of AOD in China from 2010 to 2020 at the pixel and regional scales and analyzes
the driving mechanisms of meteorological, economic, and vegetation factors on AOD. The
overall technical process is shown in Figure 3.
Atmosphere 2023, 14, x FOR PEER REVIEW 7 of 25
activities in China, this study used the yearly percent tree cover data of MOD44B with 250
m spatial resolution in China from 2010 to 2020 to evaluate the effect of increasing trees
on AOD changes.
2.3. Methodology
Based on MERRA2-AOD data, this paper investigates the spatial and temporal
evolution of AOD in China from 2010 to 2020 at the pixel and regional scales and analyzes
the driving mechanisms of meteorological, economic, and vegetation factors on AOD. The
overall technical process is shown in Figure 3.
Figure 3. Overall technical flowchart.
2.3.1. Mann–Kendall (M–K) Test
The Mann–Kendall (M–K) test was used to analyze the temporal trend in the seasonal
and annual means of AOD at each pixel. The M–K test is a nonparametric statistical test
that detects long-term trends as well as sudden changes in a meteorological element [51].
The mathematical formulations of the M–K test are as follows.
First, assume a set of time series data m1, …, mn in which each variable is independent
and randomly distributed. If k, j ≤ n, and k ≠ j, the statistics of the test are calculated as
follows:
𝑆=𝑆𝑔𝑛(𝑚−𝑚)



 (3)
Among them,
Figure 3. Overall technical flowchart.
2.3.1. Mann–Kendall (M–K) Test
The Mann–Kendall (M–K) test was used to analyze the temporal trend in the seasonal
and annual means of AOD at each pixel. The M–K test is a nonparametric statistical test
that detects long-term trends as well as sudden changes in a meteorological element [
51
].
The mathematical formulations of the M–K test are as follows.

Atmosphere 2023,14, 477 7 of 23
First, assume a set of time series data m
1
,
. . .
,m
n
in which each variable is independent
and randomly distributed. If k,j
≤
n, and k
6=
j, the statistics of the test are calculated as
follows:
S=
n−1
∑
k=1
n
∑
j=k+1
Sgnmj−mk(3)
Among them,
Sgnmj−mk=


1, mj−mk>0
0, mj−mk=0
−1, mj−mk<0
(4)
Sfollows a normal distribution with a mean of 0 and a variance of Var(S) = n(n
−
1)(2n
+ 5)/18. The original M–K statistic, designated by Z, was computed as follows:
Z=







S−1
√Var(S),S>0
0, S=0
S+1
√Var(S),S<0
(5)
where Z > 0 indicates a growing trend and Z< 0 indicates a decreasing trend. When |Z| >
1.96 and 2.58, the trend identified by the Zvalue is statistically significant with confidence
levels of 95% and 99%, respectively [52].
2.3.2. Trend Sustainability
Based on the M–K test, a rescaled polar difference (R/S) analysis-based approach was
used to calculate the Hurst index to characterize the sustainability of the AOD change
trend. Details can be found in Li, et al. [
53
]. A Hurst index greater than 0.5 means that the
future trend is consistent with the past trend, and a Hurst index less than 0.5 means that
the future trend is opposite to the past trend.
2.3.3. Coefficient of Variation
The coefficient of variation (CV) is used to quantify the variation in the AOD at annual
scales, expressed as:
CVAOD =σAOD
AOD (6)
where CV
AOD
represents the coefficient of variation of the AOD from 2010 to 2020,
σAOD
is
the standard deviation of annual means of AOD, and
AOD
is the multiyear average of the
AOD values from 2010 to 2020. Larger CV
AOD
values indicate more significant fluctuations.
2.3.4. Granger Causality Test
The Granger causality test was the first attempt to quantify the causal relation between
time series, in which causality and feedback are defined in an explicit and testable man-
ner [
40
,
44
]. The Granger causality test method uses different lag orders for two time series
and then evaluates the impact of one time series on the other time series. The regression
between the two series can be expressed as:
yt=
l
∑
i=1
γixt−i+
l
∑
i=1
τiyt−i+εt(7)
where
γ
and
τ
are regression coefficients;
ε
is random error; and lis the lag length (=1 in
this study). In Equation (7), Yis a function of previous values of the dependent variables

Atmosphere 2023,14, 477 8 of 23
(i.e., Y) and independent variables (i.e., X). To test whether Xcauses Y, we eliminate Xby
restricting γin Equation (7) to zero and derive the regression function as:
yt=
l
∑
i=1
τ0
iyt−i+ε0
t(8)
To test whether the restricted estimates are statistically significantly different from the
unrestricted estimates [i.e., Equation (7) or Equation (8)], the following Fratio is calculated:
FX∼Y=(RSR−RSU)/H
RSU/(p−R)∼F(H,p−R)(9)
where RS
R
and RS
U
are the sum of squared errors of restricted and unrestricted versions of
Equations (7) and (8), respectively. The term His the number of coefficients set to zero in
the restricted version, Ris the number of predictors in the unrestricted version, and pis the
number of observations. Taking the confidence level as 0.95, if F
X~Y ≤
F
0.05
(H,p-R), we
accept the null hypothesis that the variable eliminated from Equation (7) is not the cause of
the dependent variable according to the Granger causality test. Otherwise, we reject the
null hypothesis.
The Granger causality between AOD and related factors was investigated on different
temporal and spatial scales. The Granger causalities between AOD and socioeconomic
factors (i.e., GDPPC and IOPSK) were analyzed by the variations in their regional means
and annual means from 2010 to 2020. The Granger causality tests between AOD and
meteorological factors were analyzed by the seasonal means at both the pixel and regional
scales. For VCF, Granger causality tests were conducted using the annual means at both the
pixel and regional scales. This can help reveal the operational scale of the factors driving
AOD changes.
2.3.5. Sensitivity and Contribution Rates
The sensitivity rate (SR) of the AOD to the driving factors was defined as [37,38,42]:
SR =DF
AOD ×∑DFi−DFAODi−AOD
∑DFi−DF2(10)
where DF
i
is the average value of the driving factors in year i, including economic, me-
teorological, and vegetation driving factors.
DF
and
AOD
are the mean values of the
driving factors and AOD from 2010 to 2020, respectively. SR > 0 means AOD increases
with increasing driving factors, while SR < 0 means AOD decreases with increasing driving
factors. A larger |SR| indicates a higher sensitivity of the driving factors to AOD. The
contribution rate (CR) of the driving factors to the AOD is defined as [42]:
CR =L×Senslo pe
DF ×SR ×100% (11)
where Sen
slope
represents the rate of change of the driving factor and Lis the length of the
time series. In this paper, Formulas (10) and (11) are used to calculate the sensitivity rate of
AOD to the drivers and the contribution rate of the drivers to the AOD. CR > 0 indicates
that the driving factors positively contributes to the AOD, i.e., an increase in driving factors
leads to an increase in AOD, while CR < 0 indicates that the driving factors negatively
contributes to the AOD, i.e., an increase in the driving factors leads to a decrease in AOD.
A larger |CR| indicates a higher contribution.

Atmosphere 2023,14, 477 9 of 23
3. Results
3.1. Data Validation
To validate the accuracy of the MERRA-2 product, the AOD values derived from
AERONET measurements and the MERRA-2 product were compared at the site level,
and the AERONET data that are closer to the times of MERRA-2 AOD were selected.
Figure 4compares AERONET AOD data of levels 1.5 and 2.0 and MERRA-2 AOD at close
moments. A good consistency between MERRA-2 AOD and AERONET measurements
was found, with a root mean square error (RMSE) of 0.16 and a correlation coefficient (R
2
)
of 0.73. Overall, the data accuracy is good and consistent with previous studies [
14
,
30
].
A number of data points deviated from the 1:1 line, probably due to numerous reasons,
such as uncertainty in measurement and models and rapidly changing atmospheres. The
large difference in spatial resolutions between the site-level AERONET data and the coarse-
resolution MERRA-2 AOD may be a major reason.
Atmosphere 2023, 14, x FOR PEER REVIEW 10 of 25
as uncertainty in measurement and models and rapidly changing atmospheres. The large
difference in spatial resolutions between the site-level AERONET data and the coarse-
resolution MERRA-2 AOD may be a major reason.
Figure 4. Comparison of AERONET AOD and MERRA-2 AOD from 2010 to 2020.
3.2. Seasonal and Annual Mean AOD
Figure 5 shows the seasonal and annual average AOD values in China for the period
2010 to 2020. Divided by the Hu Line, the regional mean AOD of the Eastern and Western
areas were 0.446 ± 0.162 and 0.188 ± 0.115, respectively, indicating significant differences
in the two areas. For the Eastern and Western regions, the variation in AOD was similar
across seasons. The AOD is higher in spring and summer, and a high AOD area in the
Western region is located in the Taklamakan Desert region, where high wind speed in
spring and summer cause frequent windy and sandy weather in the Taklamakan Desert
[25,54]. The lowest wind speed and slow air movement in winter resulted in the lowest
AOD in the Western region in winter. The overall high AOD in spring in the Eastern
region and the high AOD in summer are concentrated in the BTH, which is related to local
human activities [2]. The standard deviation (STD) values of AOD followed the order of
winter > summer > autumn > spring, which was different from the order of mean AOD
values. The mean value reflects the overall level of AOD, while the standard deviation
indicates the regional variations in AOD. This indicates that in the Eastern region, there
are distinct geographical characteristics of AOD across seasons due to differences in
aerosol sources, topography, and climate. AOD values are also higher in summer in the
BTH region, which may be due to the high aerosol emissions from local industrial
production, straw burning, traffic, and energy consumption. This is the direct cause of the
high standard deviation in the summer in the Eastern region. In contrast, the overall AOD
in spring is high, and the standard deviation is low. This indicates that the AOD values in
winter and summer have large differences in different regions [30].
Figure 4. Comparison of AERONET AOD and MERRA-2 AOD from 2010 to 2020.
3.2. Seasonal and Annual Mean AOD
Figure 5shows the seasonal and annual average AOD values in China for the period
2010 to 2020. Divided by the Hu Line, the regional mean AOD of the Eastern and Western
areas were 0.446
±
0.162 and 0.188
±
0.115, respectively, indicating significant differences in
the two areas. For the Eastern and Western regions, the variation in AOD was similar across
seasons. The AOD is higher in spring and summer, and a high AOD area in the Western
region is located in the Taklamakan Desert region, where high wind speed in spring and
summer cause frequent windy and sandy weather in the Taklamakan Desert [
25
,
54
]. The
lowest wind speed and slow air movement in winter resulted in the lowest AOD in the
Western region in winter. The overall high AOD in spring in the Eastern region and the high
AOD in summer are concentrated in the BTH, which is related to local human activities [
2
].
The standard deviation (STD) values of AOD followed the order of winter > summer >
autumn > spring, which was different from the order of mean AOD values. The mean
value reflects the overall level of AOD, while the standard deviation indicates the regional
variations in AOD. This indicates that in the Eastern region, there are distinct geographical
characteristics of AOD across seasons due to differences in aerosol sources, topography,

Atmosphere 2023,14, 477 10 of 23
and climate. AOD values are also higher in summer in the BTH region, which may be due
to the high aerosol emissions from local industrial production, straw burning, traffic, and
energy consumption. This is the direct cause of the high standard deviation in the summer
in the Eastern region. In contrast, the overall AOD in spring is high, and the standard
deviation is low. This indicates that the AOD values in winter and summer have large
differences in different regions [30].
Atmosphere 2023, 14, x FOR PEER REVIEW 11 of 25
Figure 5. The seasonal and annual average AOD values in China from 2010 to 2020.
To account for the regional differences, Figure 6 shows the statistical distribution of
seasonal and mean AOD values in different major areas. The annual mean AOD values
ranged from 0.271 to 0.587 and followed the order of CC > YRD > GD > FWP > BTH > SW
> NE > NW. The areas of the FWP, NW, and YRD showed very small seasonal variations,
while the other areas showed larger seasonal variations. The AOD in GD is higher in
spring than in other seasons because coarse dust particles transported from the north over
long distances in spring adversely affect the air quality in GD [3]. The spatial variability
of AOD in the SW and BTH regions is large. Seven out of eight major areas had the highest
AOD values in spring, while only the BTH area had the highest AOD values in summer.
Such differences might be attributed to different driving forces.
Figure 6. Statistics of seasonal and annual average AOD values in different areas. Note: the outliers
indicate the extreme great/minor values of the data, which were calculated as the values beyond the
upper and lower bounds of the data.
Figure 5. The seasonal and annual average AOD values in China from 2010 to 2020.
To account for the regional differences, Figure 6shows the statistical distribution of
seasonal and mean AOD values in different major areas. The annual mean AOD values
ranged from 0.271 to 0.587 and followed the order of CC > YRD > GD > FWP > BTH > SW
> NE > NW. The areas of the FWP, NW, and YRD showed very small seasonal variations,
while the other areas showed larger seasonal variations. The AOD in GD is higher in spring
than in other seasons because coarse dust particles transported from the north over long
distances in spring adversely affect the air quality in GD [
3
]. The spatial variability of AOD
in the SW and BTH regions is large. Seven out of eight major areas had the highest AOD
values in spring, while only the BTH area had the highest AOD values in summer. Such
differences might be attributed to different driving forces.

Atmosphere 2023,14, 477 11 of 23
Atmosphere 2023, 14, x FOR PEER REVIEW 11 of 25
Figure 5. The seasonal and annual average AOD values in China from 2010 to 2020.
To account for the regional differences, Figure 6 shows the statistical distribution of
seasonal and mean AOD values in different major areas. The annual mean AOD values
ranged from 0.271 to 0.587 and followed the order of CC > YRD > GD > FWP > BTH > SW
> NE > NW. The areas of the FWP, NW, and YRD showed very small seasonal variations,
while the other areas showed larger seasonal variations. The AOD in GD is higher in
spring than in other seasons because coarse dust particles transported from the north over
long distances in spring adversely affect the air quality in GD [3]. The spatial variability
of AOD in the SW and BTH regions is large. Seven out of eight major areas had the highest
AOD values in spring, while only the BTH area had the highest AOD values in summer.
Such differences might be attributed to different driving forces.
Figure 6. Statistics of seasonal and annual average AOD values in different areas. Note: the outliers
indicate the extreme great/minor values of the data, which were calculated as the values beyond the
upper and lower bounds of the data.
Figure 6.
Statistics of seasonal and annual average AOD values in different areas. Note: the outliers
indicate the extreme great/minor values of the data, which were calculated as the values beyond the
upper and lower bounds of the data.
3.3. Temporal Trend of AOD
Figure 7shows the temporal trend of daily average AOD values in the Eastern and
Western areas of China divided by the Hu Line from 2010 to 2020. It clearly shows that
the AOD has been decreasing from 2010 to 2020 in both the Eastern and Western areas of
China, indicating effective control for aerosol emissions. Such a decreasing trend is more
significant in the Eastern area than in the Western area. This is reasonable because the AOD
in the Eastern areas was higher due to the greater intensity of human activity and might
have experienced a larger reduction after control. Figure 8shows the monthly average and
standard deviations of AOD values across the 11 years from 2010 to 2020 in the Eastern and
Western regions of the Hu Line. The highest AOD in the Western region was in May (0.261
±
0.159), which was lower than the lowest AOD of the Eastern region, which occurred
in December (0.359
±
0.21). Overall, the highest AOD value occurred in April (0.3775
±
0.2305), and the lowest value occurred in December (0.2024 ±0.1908).
Atmosphere 2023, 14, x FOR PEER REVIEW 12 of 25
3.3. Temporal Trend of AOD
Figure 7 shows the temporal trend of daily average AOD values in the Eastern and
Western areas of China divided by the Hu Line from 2010 to 2020. It clearly shows that
the AOD has been decreasing from 2010 to 2020 in both the Eastern and Western areas of
China, indicating effective control for aerosol emissions. Such a decreasing trend is more
significant in the Eastern area than in the Western area. This is reasonable because the
AOD in the Eastern areas was higher due to the greater intensity of human activity and
might have experienced a larger reduction after control. Figure 8 shows the monthly
average and standard deviations of AOD values across the 11 years from 2010 to 2020 in
the Eastern and Western regions of the Hu Line. The highest AOD in the Western region
was in May (0.261 ± 0.159), which was lower than the lowest AOD of the Eastern region,
which occurred in December (0.359 ± 0.21). Overall, the highest AOD value occurred in
April (0.3775 ± 0.2305), and the lowest value occurred in December (0.2024 ± 0.1908).
Figure 7. Daily average and standard deviations of AOD values from 2010 to 2020 in the Eastern
and Western regions of the Hu Line.
Figure 7.
Daily average and standard deviations of AOD values from 2010 to 2020 in the Eastern and
Western regions of the Hu Line.

Atmosphere 2023,14, 477 12 of 23
Atmosphere 2023, 14, x FOR PEER REVIEW 13 of 25
Figure 8. Monthly average and standard deviations of AOD values across 11 years from 2010 to
2020 in the Eastern and Western regions of the Hu Line.
Figure 9a shows the temporal trends detected by the M–K test method from the
annual means and seasonal means from 2010 to 2020. In the Eastern areas, the AOD is
generally dominated by anthropogenic aerosols, and the effective control of aerosol
emissions led to a significant decrease in these areas. In contrast, aerosols in the Western
part of the Hu Line are mainly natural since these areas are less populated and developed.
The increasing trends of AOD in the Xinjiang area in four seasons are related to special
topography, aerosol transfer from other regions, and changes in local climatic factors [54].
For temporal variations in annual mean AOD values, 60.7% of the area in China, especially
34.4% of the Eastern area of the Hu Line, showed a decreasing trend with a 95% confidence
level, confirming the significant reductions in AOD temporally. Across different seasons,
the largest reductions in AOD were found in summer, as 54.6% of the area showed a
decreasing trend. The AOD in spring was also reduced significantly and was mainly
distributed in CC and NE. In autumn, only 29.02% of the area showed a decreasing trend,
while a small portion of the area in Xinjiang even showed an increasing trend. The
reductions in winter were the smallest, and only 20.53% of the area showed a decreasing
trend. Figure 9b shows the spatial distribution of Hurst indices in China from 2010 to 2020.
On the annual scale, 88.6% of the areas derived Hurst indices greater than 0.5. This
indicates that the general decreasing trend in most areas and the local increasing trend in
the Taklamakan Desert will likely continue. On seasonal scales, the current trend in most
areas seemed to continue deriving Hurst indices > 0.5, while the exception is that the
Taklamakan Desert region in autumn deriving the Hurst index < 0.5, indicating that the
increasing trend is not sustainable.
Figure 9c shows the annual and seasonal average CV values of AOD in China from
2010 to 2020. On the annual average scale, the overall CV value was 0.11, which shows
higher fluctuations in FWP, CC, BTH, and the border areas of NE and Inner Mongolia.
Seasonally, the CV in NE and northern Inner Mongolia are higher in spring and summer,
while those in CC, southern FWP, and western YRD are higher in summer, which
indicates a large fluctuation of the time series. The possible reason is that straw burning
in NE and CC affects the air quality in the surrounding areas. With the gradual
improvement of air pollution control measures in China, aerosol pollution has decreased,
thus producing large fluctuations in the time series. The fluctuations were most significant
in the Taklamakan Desert region in autumn, probably due to the windy and sandy
Figure 8.
Monthly average and standard deviations of AOD values across 11 years from 2010 to 2020
in the Eastern and Western regions of the Hu Line.
Figure 9a shows the temporal trends detected by the M–K test method from the annual
means and seasonal means from 2010 to 2020. In the Eastern areas, the AOD is generally
dominated by anthropogenic aerosols, and the effective control of aerosol emissions led
to a significant decrease in these areas. In contrast, aerosols in the Western part of the Hu
Line are mainly natural since these areas are less populated and developed. The increasing
trends of AOD in the Xinjiang area in four seasons are related to special topography,
aerosol transfer from other regions, and changes in local climatic factors [
54
]. For temporal
variations in annual mean AOD values, 60.7% of the area in China, especially 34.4% of
the Eastern area of the Hu Line, showed a decreasing trend with a 95% confidence level,
confirming the significant reductions in AOD temporally. Across different seasons, the
largest reductions in AOD were found in summer, as 54.6% of the area showed a decreasing
trend. The AOD in spring was also reduced significantly and was mainly distributed
in CC and NE. In autumn, only 29.02% of the area showed a decreasing trend, while a
small portion of the area in Xinjiang even showed an increasing trend. The reductions in
winter were the smallest, and only 20.53% of the area showed a decreasing trend. Figure 9b
shows the spatial distribution of Hurst indices in China from 2010 to 2020. On the annual
scale, 88.6% of the areas derived Hurst indices greater than 0.5. This indicates that the
general decreasing trend in most areas and the local increasing trend in the Taklamakan
Desert will likely continue. On seasonal scales, the current trend in most areas seemed to
continue deriving Hurst indices > 0.5, while the exception is that the Taklamakan Desert
region in autumn deriving the Hurst index < 0.5, indicating that the increasing trend is not
sustainable.
Figure 9c shows the annual and seasonal average CV values of AOD in China from
2010 to 2020. On the annual average scale, the overall CV value was 0.11, which shows
higher fluctuations in FWP, CC, BTH, and the border areas of NE and Inner Mongolia.
Seasonally, the CV in NE and northern Inner Mongolia are higher in spring and summer,
while those in CC, southern FWP, and western YRD are higher in summer, which indicates
a large fluctuation of the time series. The possible reason is that straw burning in NE and
CC affects the air quality in the surrounding areas. With the gradual improvement of air
pollution control measures in China, aerosol pollution has decreased, thus producing large
fluctuations in the time series. The fluctuations were most significant in the Taklamakan
Desert region in autumn, probably due to the windy and sandy weather in the Taklamakan
region. In winter, the highest fluctuation was found in Yunnan Province.

Atmosphere 2023,14, 477 13 of 23
Atmosphere 2023, 14, x FOR PEER REVIEW 14 of 25
weather in the Taklamakan region. In winter, the highest fluctuation was found in Yunnan
Province.
Figure 9. Results of the M–K test, Hurst index, and CV values computed from the annual and
seasonal variations in AOD from 2010 to 2020.
Figure 9.
Results of the M–K test, Hurst index, and CV values computed from the annual and
seasonal variations in AOD from 2010 to 2020.
Table 2shows the trends of AOD and sustainability in eight major areas. According
to the Z values, seven out of eight major areas showed decreasing trends in AOD with a
95% confidence level. The NW area showed insignificant changes, possibly due to a lower

Atmosphere 2023,14, 477 14 of 23
intensity of human activities and a low level of aerosol emissions in the past decade. For
each major area, the decreasing trends were different across different seasons. The most
significant decreasing trends were found for the FWP area in spring and summer, CC area
in spring, summer, and winter, BTH area in summer, SW area in summer and autumn, YRD
area in spring, and GD area in winter. Combining the seasonal mean AOD in eight major
areas in Figure 5, the seasons that had the most significant decreasing trends coincided with
the hazy seasons in each major area. This indicates that aerosol control measures might
have been taken effectively for the hazy seasons in each major area in China. According
to the Hurst indices, the AOD trend in the GD region in spring and summer and the NW
region in autumn will not be sustained. On the annual scale, the Hurst indices in all regions
are greater than 0.5, which indicates that aerosol pollution will continue to decrease in
major regions of China.
Table 2.
Temporal trends in the AOD for the eight major regions detected by the M–K test and Hurst
index.
NE FWP CC BTH NW SW YRD GD
Spring −1.713 −3.425 −2.958 −2.335 −0.467 −2.491 −3.270 −1.090 *
Summer −2.491 −3.425 −2.803 −3.270 −0.467 −3.114 −2.335 −2.335 *
Autumn −2.335 −1.713 −2.647 −2.491 1.246 * −2.647 −2.335 −2.491
Winter −2.024 −2.180 −1.401 −1.713 −2.024 −1.401 −2.024 −2.803
Annual −2.491 −3.425 −2.647 −3.581 −0.934 −3.114 −3.270 −2.803
Note: The color of each cell represents different trends. White indicates no significant change (|Z| < 1.96), light
green indicates a decreasing trend (
−
2.58 < Z <
−
1.96), and light blue indicates a significant decreasing trend (Z <
−2.58). * means that the Hurst index is less than 0.5.
3.4. Analysis of Driving Factors
3.4.1. Socioeconomic Factors
Figure 10 shows the temporal changes in the annual values of AOD, IOPSK, and
GDPPC in eight major areas from 2010 to 2020. Generally, the AOD showed decreasing
trends, while the IOPSK and GDPPC values showed increasing trends in all eight major ar-
eas from 2010 to 2020. This indicates that China has effectively controlled aerosol emissions
while maintaining upward economic growth. A positive relationship between the AOD
and socioeconomic factors (i.e., IOPSK and GDPPC) was found in most areas, in which an
area with a high level of economic development normally means a high level of AOD. For
example, the developed YRD and GD areas were hazier, while the less developed NE, NW,
and SW areas were less hazy. However, CC is an exception, which was less developed but
had the highest AOD values from 2010 to 2020 among the eight areas. This also necessitates
further analysis of the factors driving the spatiotemporal changes in AOD.
Atmosphere 2023, 14, x FOR PEER REVIEW 16 of 25
Figure 10. Temporal changes in the annual values of (a) AOD, (b) IOPSK, and (c) GDPPC in eight
major areas from 2010 to 2020.
Table 3 shows the results of the driving mechanism between AOD and
socioeconomic factors. Among the eight regions, the Granger causality between AOD and
GDPPC was confirmed in BTH, while the causality between AOD and IOPSK was
confirmed in SW. According to the results of SR, AOD responded most significantly to
GDPPC in the FWP region and to IOPSK in the BTH. According to the CR results, the
contribution of GDPPC to AOD was higher in the YRD, CC, and GD regions. The
contribution of IOPSK to AOD was higher in the BTH, FWP, and GD regions. Overall,
economic development contributes to AOD in most of the regions, which indicates that
local economic development benefits haze control activities. From 2010 to 2020, while
China’s industrial structure changed, the use of fossil fuels decreased, and new energy
sources were gradually exploited, which are also important for mitigating aerosol
pollution.
Table 3. Driving force analysis performed on a regional scale between AOD and socioeconomic
factors of major regions.
GDPPC IOPSK
GCT SR CR/% GCT SR CR/%
NE / −0.083 −2.0 / −0.202 −3.573
NW / 0.012 0.979 / −0.030 −3.394
SW / −0.269 −26.398 √ −0.223 −20.999
YRD / −0.431 −36.907 / −0.439 −22.623
FWP / −0.591 −24.408 / −0.325 −28.175
CC / −0.358 −34.187 / −0.329 −22.270
BTH √ −0.378 −26.156 / −0.503 −33.358
GD / −0.436 −31.280 / −0.316 −28.924
Note: X is GDPPC/IOPSK, Y is AOD. √ indicates that causality is established (p ≤ 0.05) and / indicates
that causality is not established (p > 0.05).
3.4.2. Meteorological Factors
Figure 11a shows the results of the Granger causality test performed on a pixel basis
between AOD and meteorological factors. The blank areas represent the pixels that failed
the Granger causality test, while the colored areas confirmed a Granger causality between
AOD and related factors. Granger causality has been found between AOD and WIN/RHU
in most areas of northern China, including the NW, NE, BTH, FWP, and Taklamakan
Desert areas. In the SW, YRD, CC, and GD areas, AOD was related to the TEM, WIN, and
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
AOD
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
IOPSK(Thousand)/km2
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
202
0
GDPPC(Thousand)
Figure 10.
Temporal changes in the annual values of (
a
) AOD, (
b
) IOPSK, and (
c
) GDPPC in eight
major areas from 2010 to 2020.

Atmosphere 2023,14, 477 15 of 23
Table 3shows the results of the driving mechanism between AOD and socioeconomic
factors. Among the eight regions, the Granger causality between AOD and GDPPC was
confirmed in BTH, while the causality between AOD and IOPSK was confirmed in SW.
According to the results of SR, AOD responded most significantly to GDPPC in the FWP
region and to IOPSK in the BTH. According to the CR results, the contribution of GDPPC
to AOD was higher in the YRD, CC, and GD regions. The contribution of IOPSK to AOD
was higher in the BTH, FWP, and GD regions. Overall, economic development contributes
to AOD in most of the regions, which indicates that local economic development benefits
haze control activities. From 2010 to 2020, while China’s industrial structure changed, the
use of fossil fuels decreased, and new energy sources were gradually exploited, which are
also important for mitigating aerosol pollution.
Table 3.
Driving force analysis performed on a regional scale between AOD and socioeconomic
factors of major regions.
GDPPC IOPSK
GCT SR CR/% GCT SR CR/%
NE / −0.083 −2.0 / −0.202 −3.573
NW / 0.012 0.979 / −0.030 −3.394
SW / −0.269 −26.398 √−0.223 −20.999
YRD / −0.431 −36.907 / −0.439 −22.623
FWP / −0.591 −24.408 / −0.325 −28.175
CC / −0.358 −34.187 / −0.329 −22.270
BTH √−0.378 −26.156 / −0.503 −33.358
GD / −0.436 −31.280 / −0.316 −28.924
Note: X is GDPPC/IOPSK, Y is AOD.
√
indicates that causality is established (p
≤
0.05) and / indicates that
causality is not established (p> 0.05).
3.4.2. Meteorological Factors
Figure 11a shows the results of the Granger causality test performed on a pixel basis
between AOD and meteorological factors. The blank areas represent the pixels that failed
the Granger causality test, while the colored areas confirmed a Granger causality between
AOD and related factors. Granger causality has been found between AOD and WIN/RHU
in most areas of northern China, including the NW, NE, BTH, FWP, and Taklamakan Desert
areas. In the SW, YRD, CC, and GD areas, AOD was related to the TEM, WIN, and RHU.
Particularly for the CC area, the RHU and WIN were major causes for AOD change, which
might indicate that WIN and RHU can influence AOD at a pixel scale.
Figure 11b shows the SR of AOD to meteorological drivers. The overall SR of AOD
to TEM is small, approximately
−
0.25 in approximately 26.4% of the regions, mainly in
CC, GD, and the Eastern part of the SW region. Approximately 30.8% of the regions in
China have a positive SR of AOD to RHU, mainly in the BTH, GD, and Qinghai-Tibet
Plateau regions. This indicates that AOD increases with increasing RHU in these regions.
The sensitivity of AOD to WIN was significantly different on both sides of the Hu Line.
This phenomenon indicates that the windy and sandy weather to the west of the Hu Line
promotes aerosol pollution. On the east side of the Hu Line, the WIN accelerates airflow
and mitigates aerosol pollution. Figure 11c shows the CR of TEM, RHU, and WIN on
AOD. TEM contributed negatively to AOD in the Southern region, while TEM contributed
positively to AOD in the Qinghai-Tibet Plateau and part of NE. In the SW, CC, and Southern
parts of the NE, RHU reduced AOD pollution. WIN contributed more to AOD in the NE,
Qinghai-Tibet Plateau, and Taklamakan Desert regions, where WIN caused more windy
and dusty weather and intensified aerosol pollution. In the SW, CC, and coastal regions,
WIN had a significant mitigation effect on aerosol pollution.

Atmosphere 2023,14, 477 16 of 23
Atmosphere 2023, 14, x FOR PEER REVIEW 18 of 25
Figure 11. Results of the Granger causality test, SR, and CR analysis between AOD and
meteorological driving factors.
Table 4. Driving force analysis performed on a regional scale between AOD and meteorological
factors of major regions.
CGT SR CR/%
TEM RHU WIN TEM RHU WIN TEM RHU WIN
NE / / / 0.096 0.259 0.361 7.04 −1.7932 6.273
NW / / √ 0.177 −1.372 1.376 6.214 −4.723 7.421
SW / / √ −0.324 −2.681 1.209 −5.434 −12.221 15.593
YRD / √ / −0.09 −0.692 0.795 −2.580 −4.759 −5.659
FWP / √ / 0.092 0.228 −0.307 1.439 0.364 −3.832
CC
√
√
√
−0.371 −1.677 −1.297 −8.378 −10.936 −13.079
BTH / / √ 0.285 1.078 −0.492 4.662 −3.802 0.949
GD √ √ / −0.481 0.532 −0.618 −5.484 3.126 −2.474
Note: √ indicates that causality is established (p ≤ 0.05), / indicates that causality is not established
(p > 0.05).
Figure 11.
Results of the Granger causality test, SR, and CR analysis between AOD and meteorological
driving factors.
Table 4shows the results of the driving mechanism performed on a regional scale
between AOD and meteorological factors. The GCT results clearly show that different areas
have different causality relations. No causality was found in the NE area. In the NW and
SW areas, WIN caused changes in AOD. The changes in AOD in the GD area were only
attributed to TEM, and AOD changes in the YRD and FWP areas were only attributed to
RHU. In the BTH area, WIN caused changes in AOD. In the CC area, TEM, RHU, and WIN
were all causes attributed to changes in AOD. Overall, the SR of AOD to TEM is low. For
TEM, the SR and CR values were negative in the SW, YRD, CC, and GD but were positive in
the NE, NW, FWP, and BTH regions. AOD showed a more significant negative sensitivity
to RHU in the NW, SW, and CC regions, indicating that RHU inhibits AOD. In the NW and
SW regions, AOD showed high positive sensitivity to WIN, and the contribution of WIN to
AOD was high in these two regions, which indicated that WIN was the main driver. In the
CC region, AOD showed a high negative sensitivity to WIN, and the contribution of WIN
to AOD in this region was high at
−
13.079, indicating that WIN accelerated the airflow and
alleviated aerosol pollution.

Atmosphere 2023,14, 477 17 of 23
Table 4.
Driving force analysis performed on a regional scale between AOD and meteorological
factors of major regions.
CGT SR CR/%
TEM RHU WIN TEM RHU WIN TEM RHU WIN
NE / / / 0.096 0.259 0.361 7.04 −1.7932 6.273
NW / / √0.177 −1.372 1.376 6.214 −4.723 7.421
SW / / √−0.324 −2.681 1.209 −5.434 −12.221 15.593
YRD / √/−0.09 −0.692 0.795 −2.580 −4.759 −5.659
FWP / √/ 0.092 0.228 −0.307 1.439 0.364 −3.832
CC √ √ √ −0.371 −1.677 −1.297 −8.378 −10.936 −13.079
BTH / / √0.285 1.078 −0.492 4.662 −3.802 0.949
GD √ √ /−0.481 0.532 −0.618 −5.484 3.126 −2.474
Note: √indicates that causality is established (p≤0.05), / indicates that causality is not established (p> 0.05).
3.4.3. Vegetation Continuous Fields Factor
Figure 12 shows the results of the driving mechanism performed on a pixel scale
between AOD and VCF. The distribution of the Granger effect area is more fragmented,
mainly in the NE, BTH, GD, CC, and SW. Overall, the sensitivity of AOD to VCF is negative
in most areas, which indicates that VCF suppresses aerosol pollution, and the sensitivity
of AOD to VCF was higher in southern areas, such as GD, CC, and Southern YRD, and
weaker in areas west of the Hu Line, probably because the overall level of AOD was
lower in Western areas, and thus, the sensitivity of AOD to VCF was not significant. The
contribution of VCF to AOD was negative in most areas of China, and VCF had a more
significant inhibitory effect on AOD in parts of the Qinghai-Tibet Plateau, SW, GD, and NE,
with CR values exceeding 10%.
Atmosphere 2023, 14, x FOR PEER REVIEW 19 of 25
3.4.3. Vegetation Continuous Fields Factor
Figure 12 shows the results of the driving mechanism performed on a pixel scale
between AOD and VCF. The distribution of the Granger effect area is more fragmented,
mainly in the NE, BTH, GD, CC, and SW. Overall, the sensitivity of AOD to VCF is
negative in most areas, which indicates that VCF suppresses aerosol pollution, and the
sensitivity of AOD to VCF was higher in southern areas, such as GD, CC, and Southern
YRD, and weaker in areas west of the Hu Line, probably because the overall level of AOD
was lower in Western areas, and thus, the sensitivity of AOD to VCF was not significant.
The contribution of VCF to AOD was negative in most areas of China, and VCF had a
more significant inhibitory effect on AOD in parts of the Qinghai-Tibet Plateau, SW, GD,
and NE, with CR values exceeding 10%.
Figure 12. Results of the Granger causality test, SR, and CR between AOD and VCF.
Table 5 shows the results of the driving mechanism performed on a regional scale
between AOD and VCF. It shows that there is Granger causality between VCF and AOD
in SW, BTH, CC, NE, and GD. In the areas where Granger causality exists, the results of
CR and SR indicate that VCF suppresses aerosol pollution. The sensitivity of AOD to VCF
is high in the SW, GD, and CC areas with SR of −1.523, −0.999, and −1.047, respectively. In
the YRD, AOD had a significant positive sensitivity to VCF with an SR of 1.08. The
sensitivity of VCF to VCF was high in the SW, GD, CC, and NE regions with significant
contributions of AOD with CR values of −20.93%, −24.74%, −11.1%, and −10.88%,
respectively.
Table 5. Driving force analysis performed on a regional scale between AOD and VCF in major
regions in China.
GCT SR CR/%
NE √ −0.946 −10.88
NW / 0.134 2.31
SW √ −1.523 −20.93
YRD / 1.080 −4.76
FWP / 0.226 1.63
CC √ −1.047 −11.1
BTH √ 0.171 −0.74
GD √ −0.999 −24.74
Note: √ indicates that causality is established (p ≤ 0.05), / indicates that causality is not established
(p > 0.05).
Figure 12. Results of the Granger causality test, SR, and CR between AOD and VCF.
Table 5shows the results of the driving mechanism performed on a regional scale
between AOD and VCF. It shows that there is Granger causality between VCF and AOD in
SW, BTH, CC, NE, and GD. In the areas where Granger causality exists, the results of CR and
SR indicate that VCF suppresses aerosol pollution. The sensitivity of AOD to VCF is high
in the SW, GD, and CC areas with SR of
−
1.523,
−
0.999, and
−
1.047, respectively. In the
YRD, AOD had a significant positive sensitivity to VCF with an SR of 1.08. The sensitivity
of VCF to VCF was high in the SW, GD, CC, and NE regions with significant contributions
of AOD with CR values of −20.93%, −24.74%, −11.1%, and −10.88%, respectively.

Atmosphere 2023,14, 477 18 of 23
Table 5.
Driving force analysis performed on a regional scale between AOD and VCF in major regions
in China.
GCT SR CR/%
NE √−0.946 −10.88
NW / 0.134 2.31
SW √−1.523 −20.93
YRD / 1.080 −4.76
FWP / 0.226 1.63
CC √−1.047 −11.1
BTH √0.171 −0.74
GD √−0.999 −24.74
Note: √indicates that causality is established (p≤0.05), / indicates that causality is not established (p> 0.05).
4. Discussion
This study investigated the spatiotemporal variations and driving factors of AOD in
mainland China from 2010 to 2020 using the MERRA-2 reanalysis product.
4.1. Environmental Impact of AOD
In addition to examining the causal relationship between drivers and AOD, the
Granger causality test can also explain the feedback of the environment on AOD. There-
fore, this section discusses the effects of AOD on WIN, RHU, TEM, and VCF, as shown in
Figure 13. The variation in TEM due to AOD is small and only concentrated in the coastal
areas and the Northern parts of SW. The Granger effects of AOD on RHU and WIN are
similar and mainly concentrated in the Northern parts of China, CC and the Western part
of SW. The impact of VCF on AOD was mainly concentrated in parts of the FWP, SW, and
NW. However, whether AOD exerts a promoting or inhibiting effect on vegetation growth
is related to the level of AOD itself [
5
]. Therefore, whether the decrease in AOD leads to an
increase or decrease in VCF needs to be further investigated.
Atmosphere 2023, 14, x FOR PEER REVIEW 20 of 25
4. Discussion
This study investigated the spatiotemporal variations and driving factors of AOD in
mainland China from 2010 to 2020 using the MERRA-2 reanalysis product.
4.1. Environmental Impact of AOD
In addition to examining the causal relationship between drivers and AOD, the
Granger causality test can also explain the feedback of the environment on AOD.
Therefore, this section discusses the effects of AOD on WIN, RHU, TEM, and VCF, as
shown in Figure 13. The variation in TEM due to AOD is small and only concentrated in
the coastal areas and the Northern parts of SW. The Granger effects of AOD on RHU and
WIN are similar and mainly concentrated in the Northern parts of China, CC and the
Western part of SW. The impact of VCF on AOD was mainly concentrated in parts of the
FWP, SW, and NW. However, whether AOD exerts a promoting or inhibiting effect on
vegetation growth is related to the level of AOD itself [5]. Therefore, whether the decrease
in AOD leads to an increase or decrease in VCF needs to be further investigated.
Figure 13. Granger effect of AOD on meteorological and vegetation factors.
4.2. Spatial and Temporal Evolutionary Characteristics of AOD
The spatial distribution of AOD varied across the areas. Divided by the Hu Line, the
AOD values of the Eastern areas were significantly higher than those of the Western areas.
The high AOD values of the Western areas were mainly distributed in the Taklamakan
Desert, where sparse vegetation, high dust, perennial drought, and low rainfall were
major causes of the more dusty weather [55]. Both the linear trend analysis on a regional
scale and the M–K trend analysis on pixel and regional scales showed a significant
decreasing trend in most areas of mainland China, especially in all the Eastern areas of
the Hu Line. In autumn, Hurst < 0.5 in the Taklamakan Desert region was mainly due to
the unpredictable local wind and sandy weather, so the AOD trend was not sustainable.
For most of China, especially inland cities with stable pollution sources and a Hurst index
> 0.5, AOD will continue to decrease in the coming period.
Figure 13. Granger effect of AOD on meteorological and vegetation factors.
4.2. Spatial and Temporal Evolutionary Characteristics of AOD
The spatial distribution of AOD varied across the areas. Divided by the Hu Line, the
AOD values of the Eastern areas were significantly higher than those of the Western areas.

Atmosphere 2023,14, 477 19 of 23
The high AOD values of the Western areas were mainly distributed in the Taklamakan
Desert, where sparse vegetation, high dust, perennial drought, and low rainfall were major
causes of the more dusty weather [
55
]. Both the linear trend analysis on a regional scale
and the M–K trend analysis on pixel and regional scales showed a significant decreasing
trend in most areas of mainland China, especially in all the Eastern areas of the Hu Line. In
autumn, Hurst < 0.5 in the Taklamakan Desert region was mainly due to the unpredictable
local wind and sandy weather, so the AOD trend was not sustainable. For most of China,
especially inland cities with stable pollution sources and a Hurst index > 0.5, AOD will
continue to decrease in the coming period.
4.3. Driving Mechanism of AOD by Region
Based on the results of the Granger causality test, a Granger causality relationship
between the changes in AOD and socioeconomic factors was found for the BTH and SW
areas. This indicates that economic growth benefited aerosol control activities. Such a causal
relationship was not found in other major areas because the influence of economic growth
on aerosol control is confounded with many other factors, such as government-initiated
policies.
For meteorological factors, the changes in AOD in different major areas were attributed
to different reasons, which might be related to local geographic factors such as climate and
topography. In the Western area of the Hu Line, the AOD in the Taklamakan Desert was
caused by RHU and WIN, which was consistent with previous findings by Fan, et al. [
54
]
that the coupled effects of meteorological and topographic factors in the Tarim Basin explain
the source and evolution of aerosols in the area. For the three meteorological factors, WIN
caused changes in AOD in the SW, NW, and BTH areas, while RHU caused changes in
AOD in the YRD and FWP areas, and TEM caused changes in AOD in the GD area. In the
CC area, TEM, RHU, and WIN were all causes of the changes in AOD.
Vegetation can prevent wind and sand and regulate climate, which helps to mitigate
aerosol pollution [
6
,
56
–
58
]. This paper reveals the geographic characteristics of the effect of
vegetation on AOD at the pixel scale and regional scale. Overall, VCF plays a positive role
in mitigating aerosol pollution, which is consistent with previous conclusions [
30
,
59
]. The
contribution of VCF to AOD is more significant in NE, NW, and parts of the Qinghai-Tibet
Plateau, which indicates that China has achieved initial results in controlling windy and
sandy weather through afforestation. In addition, the continued implementation of the
policy of returning farmland to forest has both increased the proportion of forestland
and reduced atmospheric pollution activities such as straw burning, which has alleviated
aerosol pollution to some extent [30].
Although meteorological factors were found to cause decreases in AOD in all major
regions, it is worth noting that the overall decreasing trend of AOD is related to aerosol
control activities. Existing studies have confirmed that human activities are the leading
cause of air pollution [
3
,
31
,
60
]. With rapid economic growth in the past decade, the AOD in
mainland China has shown a significant decreasing trend, which was a combined result of
intensified aerosol emissions and enhanced aerosol controls by human activities. Since 2012,
the governments of China have taken a series of measures to reduce the emission of aerosol
gases. These control activities, such as industrial structure adjustment, limiting the number
of oil-burning motor vehicles, environmental protection, vegetation restoration, and manual
spraying operations to reduce dust in the air, were taken at different spatial and temporal
scales and were difficult to quantify. However, these environmental protection activities can
also affect the radiation budget between the biosphere and atmosphere and even influence
the circulation of atmospheric currents [
61
,
62
]. Changes in meteorological factors, such as
RHU and WIN, are consequences of these environmental protection activities. This study
only elucidates the driving relationship between AOD and environmental factors, while a
further study of the effects of environmental protection activities on climate is expected to
enhance our understanding of the feedback between human activities and climate.

Atmosphere 2023,14, 477 20 of 23
4.4. Limitation and Further Efforts
Based on MERRA2-AOD data, this paper analyzes the spatial and temporal evolution
and driving mechanisms of AOD in China from 2010 to 2020. At the same time, the
evolution and driving mechanisms of AOD in China’s major economic regions are studied
at the regional scale. However, the study at the regional scale makes us pay insufficient
attention to the evolution and driving mechanisms of AOD in key cities. In the Eastern
region, human activities are an important factor influencing AOD. Therefore, the study
of AOD in large cities is of great importance. The outbreak of COVID-19 has brought the
world a series of changes. The Chinese government has taken control measures, which
have greatly affected the way of life of people and socioeconomic activities. The spatial and
temporal evolution of AOD during COVID-19 should be discussed separately; otherwise, it
will affect our analysis of AOD for 2010–2020. Therefore, studying the spatial and temporal
evolution and driving mechanisms of AOD at the urban scale and considering the impact
of COVID is the next step of our research.
5. Conclusions
Based on MERRA-2-AOD reanalysis data, this paper investigated the spatial and
temporal variations in AOD in China from 2010 to 2020 and the factors driving these
changes. The main conclusions are as follows:
(1) The spatial distribution of AOD varied across the areas. Divided by the Hu Line, the
AOD values of the Eastern areas were significantly higher than those of the Western areas.
The high AOD values of the Western areas were mainly distributed in the Taklamakan
Desert, which followed the order of spring > summer > autumn > winter. The annual mean
AOD values of the eight major regions ranged from 0.271 to 0.587 in the order of CC > YRD
>GD>FWP>BTH>SW>NE>NW.
(2) AOD in China was dominated by continuous decline and moderate fluctuations
(0.1 < CV < 0.15) on the annual average scale and continuous decline and moderate to high
fluctuations on the seasonal scale. From 2010 to 2020, AOD in Eastern and Western China
showed a sustainable decline in general, while the decline was more pronounced in the
Eastern region of the Hu Line than in the Western region.
(3) The Granger causality between AOD and GDPPC was confirmed in the BTH region,
while the causality between AOD and IOPSK was confirmed in the SW region.
(4) The driving mechanism of meteorological factors on AOD had obvious geographic
characteristics. Overall, the driving relationship between AOD and TEM was weak. The
sensitivity of AOD to RHU was positive in BTH, the Qinghai-Tibet Plateau, the Sichuan
Basin, and GD and negative in the rest of the regions. Divided by the Hu Line, the sensitivity
of AOD to WIN was positive in the Western areas and was negative in most Eastern areas.
The actual contribution of TEM and RHU to AOD was small overall.
(5) Overall, VCF had a certain ameliorating effect on aerosol pollution, which indicated
that the measures of afforestation to control aerosol loadings in China had achieved initial
results. This paper focuses on the spatial and temporal evolution of AOD in China from 2010
to 2020 and analyzes the driving mechanisms of economic, vegetation, and meteorological
factors on AOD. However, the factors affecting AOD are complex, and other influencing
factors are not thoroughly investigated in this paper. In addition, the layout and influencing
factors of AOD during COVID-19 may change, and this paper does not investigate AOD in
depth during this period, which will be the focus of our next study.
Author Contributions:
Y.W.: Conceptualization, methodology, resources, and formal analysis; Y.W.
and L.Y.: writing—original draft preparation; L.Y. and Y.H.: data curation and visualization; D.X.,
D.C., H.H. and D.Z.: writing—review and editing; Y.W.: funding acquisition. All authors have read
and agreed to the published version of the manuscript.
Funding:
This work was supported by the National Natural Science Foundation of China, grant
number 41901301; The Open Research Fund Program of State Key Laboratory of Eco-hydraulics in
Northwest Arid Region, Xi’an University of Technology, grant number 2020KFKT-7; The Science

Atmosphere 2023,14, 477 21 of 23
Fund Program for Distinguished Young Scholars by Xi’an University of Science and Technology,
grant number 2022YQ3; and Open Fund of State Key Laboratory of Remote Sensing Science, grant
number OFSLRSS201922.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement:
The study was conducted based on publicly available data. The URL
for data acquisition was given in the article. Data supporting the results of this study are available on
the corresponding website.
Acknowledgments:
We would like to thank the anonymous reviewers for their insightful comments
on this manuscript. Thanks to the editors for all their work on this manuscript. Thanks to Lulu Chang,
Li Wang and Zhihao Shen of Xi’an University of Science and Technology for their work related to
data collection.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. He, Q.; Huang, B. Satellite-based mapping of daily high-resolution ground PM2.5 in China via space-time regression modeling.
Remote Sens. Environ. 2018,206, 72–83. [CrossRef]
2.
Hu, Z.; Jin, Q.; Ma, Y.; Pu, B.; Ji, Z.; Wang, Y.; Dong, W. Temporal evolution of aerosols and their extreme events in polluted Asian
regions during Terra’s 20-year observations. Remote Sens. Environ. 2021,263, 112541. [CrossRef]
3.
Fan, M.; Zhang, S.; Chen, L.; Tao, J. Analysis of long-term(2000–2013) sptio-temporal aerosol distribution over Pearl River Delta
region in China by using MODIS data. J. Remote Sens. 2016,20, 1413–1423. [CrossRef]
4.
Sockol, A.; Small Griswold, J.D. Intercomparison between CMIP5 model and MODIS satellite-retrieved data of aerosol optical
depth, cloud fraction, and cloud-aerosol interactions. Earth Space Sci. 2017,4, 485–505. [CrossRef]
5.
Yue, X.; Unger, N. Aerosol optical depth thresholds as a tool to assess diffuse radiation fertilization of the land carbon uptake in
China. Atmos. Chem. Phys. 2017,17, 1329–1342. [CrossRef]
6.
Li, L. A robust deep learning approach for spatiotemporal estimation of Satellite AOD and PM2.5. Remote Sens.
2020
,12, 264.
[CrossRef]
7.
Dubovik, O.; Holben, B.; Eck, T.F.; Smirnov, A.; Kaufman, Y.J.; King, M.D.; Tanre, D.; Slutsker, I. Variability of absorption and
optical properties of key aerosol types observed in worldwide locations. J. Atmos. Sci. 2002,59, 590–608. [CrossRef]
8.
Kanniah, K.D.; Lim, H.Q.; Kaskaoutis, D.G.; Cracknell, A.P. Investigating aerosol properties in Peninsular Malaysia via the
synergy of satellite remote sensing and ground-based measurements. Atmos. Res. 2014,138, 223–239. [CrossRef]
9.
Pang, S.; Lin, Y.; Tang, C.; Zhang, L.; Tian, P. Spatial distribution and trend of global AOD and dust AOD based on MERRA2
reanalysis data. J. Lanzhou Univ. (Nat. Sci. Ed.) 2021,57, 54–62. [CrossRef]
10.
Della Ceca, L.S.; Ferreyra, M.F.G.; Lyapustin, A.; Chudnovsky, A.; Otero, L.; Carreras, H.; Barnaba, F. Satellite-based view of the
aerosol spatial and temporal variability in the Cordoba region (Argentina) using over ten years of high-resolution data. ISPRS J.
Photogramm. Remote Sens. 2018,145, 250–267. [CrossRef]
11.
Qin, K.; Wang, L.; Wu, L.; Xu, J.; Rao, L.; Letu, H.; Shi, T.; Wang, R. A campaign for investigating aerosol optical properties during
winter hazes over Shijiazhuang, China. Atmos. Res. 2017,198, 113–122. [CrossRef]
12.
Wu, L.; Lu, X.; Qin, K.; Bai, Y.; Li, J.; Ren, C.; Zhang, Y. Analysis to Xuzhou aerosol optical characteristics with ground-based
measurements by sun photometer. Kexue Tongbao/Chin. Sci. Bull. 2016,61, 2287–2998. [CrossRef]
13.
Fosu-Amankwah, K.; Bessardon, G.E.Q.; Quansah, E.; Amekudzi, L.K.; Brooks, B.J.; Damoah, R. Assessment of aerosol burden
over Ghana. Sci. Afr. 2021,14, e00971. [CrossRef]
14.
Zhao, Q.; Yang, P.; Li, Z.; Yao, W.; Yao, Y. Spatial and Temporal Characteristics of AOD and Meteorological Factors in China
During the Period of COVID-19. Geomat. Inf. Sci. Wuhan Univ. 2021, 1–19. [CrossRef]
15.
Wang, Q.; Li, S.; Zeng, Q.; Sun, L.; Yang, J.; Lin, H. Retrieval and Validation of AOD from Himawari-8 Data over Bohai Rim
Region, China. Remote Sens. 2020,12, 3425. [CrossRef]
16.
Aube, M.; O’Neill, N.T.; Royer, A.; Lavoue, D. A modeling approach for aerosol optical depth analysis during forest fire events.
Conf. Atmos. Environ. Remote Sens. Data Process. Util. 2004,5548, 417–426. [CrossRef]
17.
Zhang, T.; Zang, L.; Mao, F.; Wan, Y.; Zhu, Y. Evaluation of Himawari-8/AHI, MERRA-2, and CAMS Aerosol Products over
China. Remote Sens. 2020,12, 1684. [CrossRef]
18.
Zhang, T.; Zeng, C.; Gong, W.; Wang, L.; Sun, K.; Shen, H.; Zhu, Z.; Zhu, Z. Improving Spatial Coverage for Aqua MODIS AOD
Using NDVI-Based Multi-Temporal Regression Analysis. Remote Sens. 2017,9, 340. [CrossRef]
19.
Anoruo, C.M. Validation of OMI seasonal and spatio-temporal variations in aerosol-cloud interactions over Banizoumbou using
AERONET data. J. Atmos. Sol.-Terr. Phys. 2020,211, 105457. [CrossRef]
20.
Wei, J.; Huang, W.; Li, Z.; Xue, W.; Peng, Y.; Sun, L.; Cribb, M. Estimating 1-km-resolution PM2.5 concentrations across China
using the space-time random forest approach. Remote Sens. Environ. 2019,231, 111221. [CrossRef]

Atmosphere 2023,14, 477 22 of 23
21.
Zhang, M.; Ma, Y.; Shi, Y.; Gong, W.; Chen, S.; Jin, S.; Wang, J. Controlling factors analysis for the Himawari-8 aerosol optical
depth accuracy from the standpoint of size distribution, solar zenith angles and scattering angles. Atmos. Environ.
2020
,233,
117501. [CrossRef]
22. Yang, J.; Hu, M. Filling the missing data gaps of daily MODIS AOD using spatiotemporal interpolation. Sci. Total Environ. 2018,
633, 677–683. [CrossRef]
23.
Rienecker, M.M.; Suarez, M.J.; Gelaro, R.; Todling, R.; Bacmeister, J.; Liu, E.; Bosilovich, M.G.; Schubert, S.D.; Takacs, L.; Kim,
G.-K.; et al. MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications. J. Clim.
2011
,24, 3624–3648.
[CrossRef]
24.
Lynch, P.; Reid, J.S.; Westphal, D.L.; Zhang, J.; Hogan, T.F.; Hyer, E.J.; Curtis, C.A.; Hegg, D.A.; Shi, Y.; Campbell, J.R.; et al. An
11-year global gridded aerosol optical thickness reanalysis (v1.0) for atmospheric and climate sciences. Geosci. Model Dev.
2016
,9,
1489–1522. [CrossRef]
25.
Ningombam, S.S.; Dumka, U.C.; Mugil, S.K.; Kuniyal, J.C.; Hooda, R.K.; Gautam, A.S.; Tiwari, S. Impacts of Aerosol Loading in
the Hindu Kush Himalayan Region Based on MERRA-2 Reanalysis Data. Atmosphere 2021,12, 1290. [CrossRef]
26.
She, L.; Zhang, H.; Wang, W.; Wang, Y.; Shi, Y. Evaluation of the Multi-Angle Implementation of Atmospheric Correction (MAIAC)
Aerosol Algorithm for Himawari-8 Data. Remote Sens. 2019,11, 2771. [CrossRef]
27.
Charabi, Y.; Gastli, A. Spatio-temporal assessment of dust risk maps for solar energy systems using proxy data. Renew. Energy
2012,44, 23–31. [CrossRef]
28.
de Leeuw, G.; Sogacheva, L.; Rodriguez, E.; Kourtidis, K.; Georgoulias, A.K.; Alexandri, G.; Amiridis, V.; Proestakis, E.; Marinou,
E.; Xue, Y.; et al. Two decades of satellite observations of AOD over mainland China using ATSR-2, AATSR and MODIS/Terra:
Data set evaluation and large-scale patterns. Atmos. Chem. Phys. 2018,18, 1573–1592. [CrossRef]
29.
Wu, Z.; Jiang, Q.; Yu, Y.; Xiao, H.; Freese, D. Spatio-Temporal Evolution of a Typical Sandstorm Event in an Arid Area of
Northwest China in April 2018 Based on Remote Sensing Data. Remote Sens. 2022,14, 3065. [CrossRef]
30.
Sun, Z.; Cheng, X.; Xia, X. Spatial-temporal distribution and impact factors of aerosol optical depth over China. Chin. Environ. Sci.
2021,41, 4466–4475. [CrossRef]
31.
Zhao, Q.; Yang, P.; Yao, W.; Yao, Y. Adaptive AOD Forecast Model Based on GNSS-Derived PWV and Meteorological Parameters.
IEEE Trans. Geosci. Remote Sens. 2022,60, 5800610. [CrossRef]
32.
Namdari, S.; Alnasrawi, A.I.Z.; Ghorbanzadeh, O.; Sorooshian, A.; Kamran, K.V.; Ghamisi, P. Time Series of Remote Sensing Data
for Interaction Analysis of the Vegetation Coverage and Dust Activity in the Middle East. Remote Sens.
2022
,14, 2963. [CrossRef]
33.
Zeydan, O.; Tariq, S.; Qayyum, F.; Mehmood, U.; Ul-Haq, Z. Investigating the long-term trends in aerosol optical depth and
its association with meteorological parameters and enhanced vegetation index over Turkey. Environ. Sci. Pollut. Res.
2022
,30,
20337–20356. [CrossRef] [PubMed]
34.
Kumar, K.R.; Attada, R.; Dasari, H.P.; Vellore, R.K.; Langodan, S.; Abualnaja, Y.O.; Hoteit, I. Aerosol Optical Depth variability
over the Arabian Peninsula as inferred from satellite measurements. Atmos. Environ. 2018,187, 346–357. [CrossRef]
35.
Li, X.; Liang, H.; Cheng, W. Spatio-temporal variation in AOD and correlation analysis with PAR and NPP in China from 2001 to
2017. Remote Sens. 2020,12, 976. [CrossRef]
36.
Xu, Q.; Chen, X.; Yang, S.; Tang, L.; Dong, J. Spatiotemporal relationship between Himawari-8 hourly columnar aerosol optical
depth (AOD) and ground-level PM2.5 mass concentration in mainland China. Sci. Total Environ. 2021,765, 144241. [CrossRef]
37.
Sun, H.; Bai, Y.; Lu, M.; Wang, J.; Tuo, Y.; Yan, D.; Zhang, W. Drivers of the water use efficiency changes in China during 1982-2015.
Sci. Total Env. 2021,799, 149145. [CrossRef]
38.
Zheng, H.; Zhang, L.; Zhu, R.; Liu, C.; Sato, Y.; Fukushima, Y. Responses of streamflow to climate and land surface change in the
headwaters of the Yellow River Basin. Water Resour. Res. 2009,45, W00A19. [CrossRef]
39.
Yao, J.; Zhai, H.; Yang, X.; Wen, Z.; Wu, S.; Zhu, H.; Tang, X. Spatiotemporal Variations of Aerosols in China during the COVID-19
Pandemic Lockdown. Remote Sens. 2022,14, 696. [CrossRef]
40.
Mariano, D.A.; dos Santos, C.A.C.; Wardlow, B.D.; Anderson, M.C.; Schiltmeyer, A.V.; Tadesse, T.; Svoboda, M.D. Use of remote
sensing indicators to assess effects of drought and human-induced land degradation on ecosystem health in Northeastern Brazil.
Remote Sens. Environ. 2018,213, 129–143. [CrossRef]
41.
Shen, Y.; Shen, H.; Cheng, Q.; Huang, L.; Zhang, L. Monitoring three-decade expansion of China’s major cities based on satellite
remote sensing images. Remote Sens. 2020,12, 491. [CrossRef]
42.
Guo, W.; Huang, S.; Zhao, J.; Li, Z.; Huang, Q.; Deng, M. Spatio-temporal dynamics and driving forces of potential evapotranspi-
ration in the Wei River Basin. Trans. Chin. Soc. Agric. Eng. (Trans. CSAE) 2021,37, 81–89. [CrossRef]
43.
The State Council Information Office of the People’s Republic of China. Responding to Climate Change: China’s Policies
and Actions. Available online: http://ex.chinadaily.com.cn/exchange/partners/82/rss/channel/language/columns/v0m20b/
stories/WS6179136aa310cdd39bc71a88.html (accessed on 23 October 2022).
44.
Granger, C.W.J. Investigating Causal Relations by Econometric Models and Cross-spectral Methods. Econometrica
1969
,37,
424–438. [CrossRef]
45.
Wang, J.; Zhang, B.; Gui, H.; An, L.; Hua, C.; Liu, C.; You, Y.; Xu, R.; Chi, Q.; Nan, Y.; et al. Atmospheric Environment Bulletin; China
Meteorological Administration: Beijing, China, 2020.
46.
Ding, J.; Cheng, C.; Zhang, W.; Tian, Y. The ideological origins and geographical demarcation significance of Hu Huanyong Line.
Dili Xuebao/Acta Geogr. Sin. 2021,76, 1317–1333. [CrossRef]

Atmosphere 2023,14, 477 23 of 23
47.
Giles, D.M.; Sinyuk, A.; Sorokin, M.G.; Schafer, J.S.; Smirnov, A.; Slutsker, I.; Eck, T.F.; Holben, B.N.; Lewis, J.R.; Campbell, J.R.;
et al. Advancements in the Aerosol Robotic Network (AERONET) Version 3 database—Automated near-real-time quality control
algorithm with improved cloud screening for Sun photometer aerosol optical depth (AOD) measurements. Atmos. Meas. Tech.
2019,12, 169–209. [CrossRef]
48.
Su, H.; Liu, H.; Wu, Q. Prediction of Water Depth From Multispectral Satellite Imagery-The Regression Kriging Alternative. IEEE
Geosci. Remote Sens. Lett. 2015,12, 2511–2515. [CrossRef]
49.
Lee, S.-J.; Serre, M.L.; van Donkelaar, A.; Martin, R.V.; Burnett, R.T.; Jerrett, M. Comparison of geostatistical interpolation and
remote sensing techniques for estimating long-term exposure to ambient PM2.5 concentrations across the continental United
States. Environ. Health Perspect. 2012,120, 1727–1732. [CrossRef]
50.
Kang, J.; Jin, R.; Li, X. Regression Kriging-Based Upscaling of Soil Moisture Measurements From a Wireless Sensor Network
and Multiresource Remote Sensing Information Over Heterogeneous Cropland. IEEE Geosci. Remote Sens. Lett.
2015
,12, 92–96.
[CrossRef]
51.
Mohammad, L.; Mondal, I.; Bandyopadhyay, J.; Pham, Q.B.; Nguyen, X.C.; Dinh, C.D.; Al-Quraishi, A.M.F. Assessment of
spatio-temporal trends of satellite-based aerosol optical depth using Mann–Kendall test and Sen’s slope estimator model. Geomat.
Nat. Hazards Risk 2022,13, 1270–1298. [CrossRef]
52.
Ning, Z.; Zhang, J.; Wang, G. Variation and global pattern of major meteorological elements during 1948~2016. China Environ. Sci.
2021,41, 4085–4095. [CrossRef]
53.
Li, G.; Chen, W.; Zhang, X.; Yang, Z.; Wang, Z.; Bi, P. Spatiotemporal changes and driving factors of vegetation in 14 different
climatic regions in the global from 1981 to 2018. Environ. Sci. Pollut. Res. Int. 2022,29, 75322–75337. [CrossRef] [PubMed]
54.
Fan, J.; Yue, X.; Sun, Q.; Wang, S. Case study of dust event sources from the Gobi and Taklamakan deserts: An investigation of the
horizontal evolution and topographical effect using numerical modeling and remote sensing. J. Environ. Sci.
2017
,56, 62–70.
[CrossRef] [PubMed]
55.
Li, J.; He, Q.; Ge, X.; Abbas, A.; Jin, L. Spatio-temporal changes of AOD in Xinjiang of China from 2000 to 2019: Which factor is
more influential, natural factor or human factor? PLoS ONE 2021,16, e0253942. [CrossRef] [PubMed]
56.
Luo, Y.; Zheng, X.; Zhao, T.; Chen, J. A climatology of aerosol optical depth over China from recent 10 years of MODIS remote
sensing data. Int. J. Climatol. 2014,34, 863–870. [CrossRef]
57.
Deng, X.; Wu, L.; He, C.; Shao, H. Study on Spatiotemporal Variation Pattern of Vegetation Coverage on Qinghai-Tibet Plateau
and the Analysis of Its Climate Driving Factors. Int. J. Environ. Res. Public Health 2022,19, 8836. [CrossRef]
58.
Wang, Y.; Yan, G.; Xie, D.; Hu, R.; Zhang, H. Generating Long Time Series of High Spatiotemporal Resolution FPAR Images in the
Remote Sensing Trend Surface Framework. IEEE Trans. Geosci. Remote Sens. 2022,60, 1–15. [CrossRef]
59.
Tan, S.; Chen, B.; Wang, H.; Che, H.; Yu, H.; Shi, G. Variations in Aerosol Optical Properties over East Asian Dust Storm Source
Regions and Their Climatic Factors during 2000–2021. Atmosphere 2022,13, 992. [CrossRef]
60.
Bai, Y.; Wu, L.; Qin, K.; Zhang, Y.; Shen, Y.; Zhou, Y. A Geographically and Temporally Weighted Regression Model for
Ground-Level PM2.5 Estimation from Satellite-Derived 500 m Resolution AOD. Remote Sens. 2016,8, 262. [CrossRef]
61.
Wang, Y.T.; Xie, D.H.; Liu, S.; Hu, R.H.; Li, Y.H.; Yan, G.J. Scaling of FAPAR from the Field to the Satellite. Remote Sens
2016
,8, 310.
[CrossRef]
62.
Wang, Y.; Xie, D.; Zhan, Y.; Li, H.; Yan, G.; Chen, Y. Assessing the Accuracy of Landsat-MODIS NDVI Fusion with Limited Input
Data: A Strategy for Base Data Selection. Remote Sens. 2021,13, 266. [CrossRef]
Disclaimer/Publisher’s Note:
The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.