Assessment of MERRA-2 Surface PM2.5 Over the Yangtze River Basin: Ground-Based Veriﬁcation, Spatiotemporal Distribution and Meteorological Dependence

remote sensing

Article Assessment of MERRA-2 Surface PM2.5 over the Yangtze River Basin: Ground-based Veriﬁcation, Spatiotemporal Distribution and Meteorological Dependence

Lijie He 1 , Aiwen Lin 1,*, Xinxin Chen 2, Hao Zhou 3 , Zhigao Zhou 1 and Peipei He 4 1 School of Resource and Environmental Science, Wuhan University, Wuhan 430079, China; [email protected] (L.H.); [email protected] (Z.Z.) 2 Hubei Key Laboratory of Critical Zone Evolution, School of Earth Sciences, China University of Geosciences, Wuhan 430074, China; [email protected] 3 MOE Key Laboratory of Fundamental Physical Quantities Measurement, School of Physics, Huazhong University of Science and Technology, Wuhan 430074, China; [email protected] 4 College of Surveying and Geo-Informatics, North China University of Water Resources and Electric Power, Zhengzhou 450045, China; [email protected] * Correspondence: [email protected]; Tel.: +86-137-0719-1591

Received: 8 January 2019; Accepted: 20 February 2019; Published: 23 February 2019

Abstract: A good understanding of how meteorological conditions exacerbate or mitigate air pollution is critical for developing robust emission reduction policies. Thus, based on a multiple linear regression (MLR) model in this study, the quantiﬁed impacts of six meteorological variables on PM2.5 (i.e., particle matter with diameter of 2.5 µm or less) and its major components were estimated over the Yangtze River Basin (YRB). The 38-year (1980–2017) daily PM2.5 and meteorological data were derived from the newly-released Modern-Era Retrospective Analysis and Research and Application, version 2 (MERRA-2) products. The MERRA-2 PM2.5 was underestimated compared with ground measurements, partly due to the bias in the MERRA-2 Aerosol Optical Depth (AOD) assimilation. An over-increasing trend in each PM2.5 component occurred for the whole study period; however, this has been curbed since 2007. The MLR model suggested that meteorological variability could explain up to 67% of the PM2.5 changes. PM2.5 was robustly anti-correlated with surface wind speed, precipitation and boundary layer height (BLH), but was positively correlated with temperature throughout the YRB. The relationship of relative humidity (RH) and total cloud cover with PM2.5 showed regional dependencies, with negative correlation in the Yangtze River Delta (YRD) and positive correlation in the other areas. In particular, PM2.5 was most sensitive to surface wind speed, and the sensitivity was approximately −2.42 µg m−3 m−1 s. This study highlighted the impact of meteorological conditions on PM2.5 growth, although it was much smaller than the anthropogenic emissions impact.

Keywords: MERRA-2; PM2.5; meteorological variables; Yangtze River basin

1. Introduction Since its reform and opening up (1978 onward), with the rapid development of urbanization and industrialization, China has become one of the most polluted areas [1–6]. PM2.5 (i.e., particle matter with diameter of 2.5 µm or less) is an important metric for air pollution, which has attracted considerable attention. When these particles are breathed in, they can lead to premature deaths [7–10]. In addition, they can effectively scatter and absorb solar radiation, thereby reducing atmospheric visibility and producing signiﬁcant climate impacts [11–19].

Remote Sens. 2019, 11, 460; doi:10.3390/rs11040460 www.mdpi.com/journal/remotesensing Remote Sens. 2019, 11, 460 2 of 25

A large number of studies have been performed to analyze potential causes of recent heavy air pollution in China and its effective solutions. Most of them have suggested that the excessive increase in anthropogenic emissions was the most important reason for PM2.5 growth in China [20–25]. For example, Guo et al. [21] found that during Asia-Pacific Economic Cooperation (APEC) China 2014, the air quality in Beijing was effectively improved mainly due to the emission control measures. Furthermore, Ren et al. [22] confirmed that Beijing’ air pollution during the 2015 China Victory Day Parade mainly came from traffic and cooking emissions. Additionally, Larkin [23] reported that rapid urban sprawl led to a significant growth in NO2 and PM2.5 air pollution at 830 cities in East Asia from 2000 to 2010. All of these previous studies indicated that anthropogenic emissions were main sources of PM2.5 growth in China. However, air pollution was observed to have seasonal and interannual variations, although anthropogenic emissions were continuously increasing. Obviously, these variations were caused by local meteorological changes [3,26–34]. Empirical Studies on the sensitivities of PM2.5 concentrations to meteorological conditions were exhibited in Table1. Tai et al. [ 26] revealed that based on a multiple linear regression (MLR) model, local meteorological factors could explain up to 50% of daily PM2.5 variability in the USA from 1998 to 2008. Westervelt et al. [27] further emphasized the maximum impact of temperature, wind speed and precipitation on PM2.5 concentrations at global scale. Additionally, Harrison et al. [28] revealed that PM2.5 concentrations were more strongly correlated with easterly winds from the European mainland than with the local traffic sources. Similarly, in three European cities (Athens, London and Madrid), local meteorological conditions also had significant impacts on particulate matters (PM2.5 and PM10)[29]. In Saudi Arabia, both PM10 and PM2.5 were negatively related to relative humidity; however, they were positively correlated with wind speed and temperature [30]. Recently, He et al [3] discovered that meteorological conditions were the primary factors of daily variations in five major air pollutants (PM2.5,O3, NO2, SO2 and CO) during 2014–2015 in China. Wang et al. [31] further confirmed that an air stagnation event, characterized by weak wind speed, no precipitation and high temperature, easily led to haze in China. Furthermore, in the Beijing urban area, wind rose plots showed that low wind speeds of northeastern or southwestern winds usually led to high black carbon (BC) concentrations [32]. In 68 major cities of China, Yang et al. [33] reported that ground PM2.5 concentrations were associated with local meteorological conditions at seasonal, yearly and regional scales. The impacts of PM2.5 pollution on atmospheric visibility, occupational health and occupants’ behaviors were also estimated in China [34]. Even though extensive research has been carried out to explore the relationship between meteorological conditions and air pollution, most of it has focused on the impact in daily PM2.5 variability for a short term, such as during a haze event. However, studies on the sensitivity of air pollution to long-term meteorological changes are still limited, mainly due to a lack of data [35,36]. The newly-released Modern-Era Retrospective Analysis and Research and Application, version 2 (MERRA-2) product could provide daily PM2.5 and meteorological data after 1980 for free, giving us an opportunity to quantify the long-term effect of meteorological conditions on PM2.5 [37–40]. An important advancement of the MERRA-2 aerosol reanalysis product was that it not only assimilated the bias-corrected 550nm-AODs from the space-based AVHRR (Advanced Very High-Resolution Radiometer) and MODIS (Moderate Resolution Imaging Spectroradiometer) instruments, but also assimilated the non-bias-corrected 550nm-AODs from the space-based MISR (Multiangle Imaging Spectroradiometer) and the ground-based AERONET (Aerosol Robotic Network) stations. Furthermore, five types of aerosols, including dust, sea salt, sulfate (SO4), black carbon (BC) and organic carbon (OC), were also simulated based on the NASA Global Modeling and Assimilation Office (GMAO) Earth system model (GEOS-5) and the Goddard Chemistry, Aerosol, Radiation and Transport (GOCART) aerosol module. Notably, the MERRA-2 AOD assimilation in GEOS-5 involved very careful cloud screening and homogenization of the observing system through neural net scheme, thus providing high-quality AOD observations [37,38]. In general, the newly-released MERRA-2 atmospheric reanalysis was widely used for studying global climate change and regional environment due to its high data quality, continuous spatial and temporal resolutions and diverse aerosol species. Remote Sens. 2019, 11, 460 3 of 25

The Yangtze River Basin (YRB) was divided into four regions due to their diverse underlying surfaces and aerosol sources, from east to west, followed by the Yangtze River Delta (YRD), Central China (CC), Sichuan Basin (SB) and the source of the YRB (SYR) (Figure1)[ 8,9,21]. Both YRD and CC had an average elevation of no more than 50m and had frequent urbanization and industrialization activities, resulting in high aerosol loadings [8,9]. However, the YRD might be similarly affected by sea salt aerosols. In the SB, the unique basin topography would lead to a low wind speed, high relative humidity and cloudy meteorological condition, which was different from its surrounding areas. It was not conducive to the diffusion and dilution of aerosol particles [8]. By contrast, most of the SYR were located in the Qinghai-Tibet Plateau, which had unique plateau weather and little anthropogenic activities. Coarse particles were the major aerosol type; for example, occasional large areas of sand and dust were transported here via surface winds in spring [31]. Overall, such diverse underlying surfaces, aerosol sources and meteorological conditions were conducive to studying the sensitivities of different PM2.5 components to local meteorological conditions. Therefore, the YRB was selected as the region of interest in this study. This study aimed to (1) verify the performance of MERRA-2 PM2.5 reanalysis; (2) make a good knowledge of spatiotemporal distributions of each PM2.5 component over the YRB from 1980 to 2017; and (3) reveal the quantiﬁed impacts of major meteorological factors (including surface wind speed, surface temperature, precipitation, relative humidity, total cloud cover and boundary layer height) on each PM2.5 component. The structure was organized as follows. Section2 introduced data and analytical methods. Section3 analyzed MERRA-2 PM 2.5 veriﬁcation and spatiotemporal distributions. The sensitivities of different PM2.5 components to meteorological variables were further discussed in Section4. Finally, conclusions were provided in Section5.

Table 1. Empirical Studies of the impacts of meteorological conditions on PM2.5 (i.e., particle matter with diameter of 2.5 µm or less) concentrations.

Meteorological Pollution Study Areas Methods Main Conclusions Variables Indicators Surface temperature, precipitation, total Daily meteorological United states Multiple linear cloud cover, wind PM2.5, SO4, variables could explain (1998–2008) [26] regression (MLR) speed, wind OC, EC * up to 50% changes of direction and relative PM2.5 in the USA. humidity Surface temperature, PM concentrations precipitation, total 2.5 were most sensitive to Global scale Chemistry-climate cloud cover, wind PM surface temperature, (2005–2016) [27] model speed, relative 2.5 wind speed and humidity and precipitation. pressure

PM2.5 concentrations exhibited a stronger correlation with United Kingdom Regression Wind speed and PM , PM , 10 2.5 easterly winds from (2009) [28] analyses direction NOx * the European mainland than with the local trafﬁc sources. Wind velocity, Principal temperature, relative Local meteorological Three European component and humidity, conditions had a cities (2005–2015) PM , PM regression precipitation, solar 10 2.5 signiﬁcant impact on [29] analyses radiation and air pollutions. atmospheric pressure Remote Sens. 2019, 11, 460 4 of 25

Table 1. Cont.

Meteorological Pollution Study Areas Methods Main Conclusions Variables Indicators Compared to no-stagnation Air stagnation event conditions, air (i.e., daily 10-m wind stagnation events in Europe, US, speed < 3.2 m s−1, winter could increase China MLR 500-hPa tropospheric PM 2.5 PM concentrations (2013–2016) [30] wind speed < 13m 2.5 in the United States, s−1 and no Europe and China by precipitation) 46%, 68% and 60%, respectively.

Both PM10 and PM2.5 were anti-correlated Relative humidity, Saudi Arabia with relative humidity, MLR wind speed and PM , PM (2014–2015) [31] 2.5 10 but positively related temperature to wind speed and temperature. High BC concentrations were Beijing China Relative humidity Black carbon MLR usually related to poor (2005–2013) [32] and wind speed (BC) visibility and low wind speed.

Relative humidity, PM2.5 was signiﬁcantly China Multivariate wind speed, surface correlated with PM (2013–2014) [33] analysis pressure and 2.5 meteorological temperature conditions in China.

* SO4, OC, EC, NOx refer to sulfate, organic carbon, element carbon and nitrogen oxides, respectively.

2. Data and Methods

2.1. Data The MERRA-2 atmospheric reanalysis product was newly released by the NASA Global Modeling and Assimilation Ofﬁce (GMAO) in 2017. Based on the NASA GMAO Earth system model version 5 (GEOS 5) and the Goddard Chemistry, Aerosol, Radiation and Transport (GOCART) aerosol module, the gridded aerosol data of MERRA-2 were assimilated with satellite and ground observations. More details on MERRA-2 PM2.5 treatment could be found in [37–40], and were brieﬂy described here. MERRA-2 PM2.5 concentrations were calculated by Equation (1):

PM2.5 = 1.375 × SO4 + 1.6 × OC + BC + Dust2.5 + SS2.5 (1)

Here, SO4, OC, BC, Dust2.5 and SS2.5 represented sulfate, organic carbon, black carbon, dust and sea-salt particulate matter with a diameter of less than 2.5 µm from the GOCART aerosol module, respectively. Notably, the nitrate particulate matter primarily emitted by vehicle exhaust and industrial production was lacking in the MERRA-2 PM2.5 reanalysis, probably leading to biases when compared with ground measurements. In order to verify the performance of the MERRA-2 PM2.5 product, we compared MERRA-2 PM2.5 with two-year (2015–2016) daily mean PM2.5 ground observations from 476 state-controlled air quality monitoring stations over the YRB (Figure1). By using the tapered element oscillating microbalance method and/or the beta absorption method, these state-controlled air quality monitoring stations provided hourly and 24-hour average concentrations of six major air pollutants (including PM2.5, PM10, SO2, NO2,O3 and CO) after 2013 [40]. The systematic uncertainty Remote Sens. 2019, 11, 460 5 of 25

of ground PM2.5 from these air quality monitoring stations was controlled within 15% in accordance with theRemote latest Sens. version2019, 11, x FOR of China’s PEER REVIEW environmental protection standards [40]. 5 of 27 In addition, the spatial and temporal distributions of MERRA-2 PM2.5 were further compared concentrations of six major air pollutants (including PM2.5, PM10, SO2, NO2, O3 and CO) after 2013 with another PM2.5 reanalysis product released by the Atmospheric Composition Analysis Group of [40]. The systematic uncertainty of ground PM2.5 from these air quality monitoring stations was NASA. Similarly, based on the GEOS-Chem chemical transport model, the satellite-retrieved PM2.5 controlled within 15% in accordance with the latest version of China's environmental protection concentrations (0.01◦ × 0.01◦) were estimated by combining column integrated AODs from MODIS standards [40]. (ModerateIn Resolution addition, the Imaging spatial and Spectroradiometer), temporal distributions MISR of MERRA-2 (Multiangle PM2.5Imaging were further Spectroradiometer) compared and SeaWIFSwith another (Sea-Viewing PM2.5 reanalysis Wide product Field released of View by the Sensor) Atmospheric instruments Composition [41, 42Analysis]. Then, Group in of order to adjustNASA. the residual Similarly, bias, based the on satellite-derived the GEOS-Chem PMchemical2.5 concentrations transport model, were the calibratedsatellite-retrieved by ground-based PM2.5 concentrations (0.01° × 0.01°) were estimated by combining column integrated AODs from MODIS observations using a geographically weighted regression (GWR). Hereafter, the PM2.5 reanalysis (Moderate Resolution Imaging Spectroradiometer), MISR (Multiangle Imaging Spectroradiometer) product was abbreviated as GWR PM2.5 in this study. As shown in Figure A1, the annual mean GWR and SeaWIFS (Sea-Viewing Wide Field of View Sensor) instruments [41,42]. Then, in order to adjust PM2.5 concentrations were compared with ground PM2.5 measurements from 476 state-controlled air the residual bias, the satellite-derived PM2.5 concentrations were calibrated by ground-based quality monitoring stations over the YRB during 2015–2016. A high correlation coefficient (R = 0.76) observations using a geographically weighted regression (GWR). Hereafter, the PM2.5 reanalysis demonstratedproduct was a good abbreviated consistency as GWR between PM2.5 GWRin this PMstudy.2.5 andAs shown ground in PMFigure2.5, whichA1, the provided annual mean usenough confidenceGWR to PM use2.5 the concentrations GWR PM2.5 product were compared to verify with the spatiotemporal ground PM2.5 distribution measurements of MERRA-2 from 476 PM2.5. Sincestate-controlled both the non-bias-correctedair quality monitoring AODs stations (MISR over the and YRB AERONET) during 2015–2016. and the A high bias-corrected correlation AODs (MODIScoefficient and AVHRR) (R = 0.76) were demonstrated assimilated a in good the MERRA-2 consistency aerosol between reanalysis, GWR PM2.5 these and above ground AOD PM2.5 datasets, which provided us enough confidence to use the GWR PM2.5 product to verify the spatiotemporal were all not suitable for verifying the MERRA-2 AODs. Therefore, the daily ground AOD observations distribution of MERRA-2 PM2.5. from five ChinaSince both Aerosol the non-bias-corrected Remote Sensing NetworkAODs (MISR (CARSNET) and AERONET) sites in and the the YRB bias-corrected during 2013–2016, AODs were selected(MODIS to be compared and AVHRR) with were their assimilated corresponding in the MERRA-2 MERRA-2 aerosol AODs reanalysis, at 550 nm these (Figure above1 and AOD Table 2). The CARSNETdatasets were was all a groundnot suitable aerosol for verifying network the with MERRA-2 approximately AODs. Therefore, 50 sites the across daily China, ground established AOD by the Chinaobservations meteorological from five administrationsChina Aerosol Remote and Sensing university Network research (CARSNET) institutes sites [9 in,12 the]. YRB It used during Cimel sun photometers2013 – 2016, to retrieve were selected the Level-1.0 to be compared (unscreened) with their and corresponding Level-1.5 (cloud-screened) MERRA-2 AODs AOD at 550 data nm at eight wavelengths(Figure (340,1 and 380,Table 440, 2). 500,The CARSNET 675, 870, 1020was a and ground 1640 aerosol nm) every network 15minutes with approximately [9]. The total 50 sites uncertainty across China, established by the China meteorological administrations and university research in the CARSNET AODs was about 0.01 to 0.02 [12]. The CARSNET AODs used in this study were all institutes [9,12]. It used Cimel sun photometers to retrieve the Level-1.0 (unscreened) and Level-1.5 cloud-screened(cloud-screened) data. AOD As shown data atin eight Figure wavelengths1, from east(340, to 380, west, 440, the 500, five 675, CARSNET 870, 1020 and sites 1640 were nm) Lin’an (rural site,every 39 15 mminutes above [9]. sea The level), total Wuhanuncertainty (urban in the site, CARSNET 15 m above AODs sea was level), about 0.01 Changde to 0.02 (rural[12]. The site, 38 m aboveCARSNET sea level), AODs Chengdu used in (urban this study site, were 485 all m cloud-screened above sea level) data. and As shown Shangri-La in Figure (remote 1, from site, east 3280 m aboveto sea west, level) the [five12]. CARSNET sites were Lin’an (rural site, 39 m above sea level), Wuhan (urban site, The15 m MERRA-2 above sea level), atmospheric Changde reanalysis(rural site, 38 product m above also sea level), provided Chengdu grid-level (urban site, (0.05 485◦ ×m 0.625above◦ ) daily sea level) and Shangri-La (remote site, 3280 m above sea level) [12]. mean meteorological variables data during 1980–2017 (Table2). They were surface wind speed (m s −1), The MERRA-2 atmospheric reanalysis product also provided grid-level (0.05° × 0.625°) daily surfacemean temperature meteorological (K), surfacevariables precipitationdata during 1980–2017 (mm d-1), (Table relative 2). They humidity were surface (%), wind total cloudspeed (m cover (%) and boundarys-1), surface layer temperature height (m).(K), surface precipitation (mm d-1), relative humidity (%), total cloud cover (%) and boundary layer height (m).

FigureFigure 1. Distribution 1. Distribution of ground of ground air monitoring air monitoring stations stations and and China China Aerosol Aerosol Remote Remote Sensing Sensing Network (CARSNET)Network sites (CARSNET) over the sitesYangtze over River the Yangtze Basin (YRB). River Basin The YRB(YRB). was The divided YRB was into divided four regions,into four labeled from 1 to 4, which are the Yangtze River Delta (YRD), Central China (CC), Sichuan Basin (SB) and source of the YRB (SYR). Remote Sens. 2019, 11, 460 6 of 25

Table 2. Summary of data used in this study.

Temporal Spatial Resolution Data Source Coverage MERRA-2 * 0.05◦ × 0.625◦ daily\1980–2017 http://disc.sci.gsfc.nasa.gov/mdisc/ PM http://ﬁzz.phys.dal.ca/~{}atmos/ 2.5 GWR * 0.01◦ × 0.01◦ daily\1998–2017 martin/?page_id=140 http://113.108.142.147: Ground \ daily\2015–2016 20035/emcpublish/ MERRA-2 0.05◦ × 0.625◦ daily\2013–2016 http://disc.sci.gsfc.nasa.gov/mdisc/ AOD * CARSNET * \ daily\2013–2016 \ https://ladsweb.modaps.eosdis. Terra, Aqua 0.01◦ × 0.01◦ daily\2016 nasa.gov/search/ Meteorological MERRA-2 0.05◦ × 0.625◦ daily\1980–2017 http://disc.sci.gsfc.nasa.gov/mdisc/ factors * AOD, MERRA-2, GWR, CARSNET refer to Aerosol Optical Depth, Modern-Era Retrospective Analysis and Research and Application, version 2 product, Geographically Weighted Regression product and China Aerosol Remote Sensing Network, respectively.

2.2. Methods

2.2.1. Comparison Analysis

According to [40], grid-level daily mean MERRA-2 PM2.5 concentrations were compared with the corresponding ground measurements during 2015–2016 over the YRB by using a series of linear regressions. The regression parameters include slope, y–intercept, correlation coefﬁcient (R), root mean square error (RMSE, Equation (2)), mean absolute error (MAE, Equation (3)) and the relative mean bias (RMB, Equation (4)): s 1 n 2 RMSE = ∑ PM2.5(MERRA−2)i − PM2.5(ground)i (2) n i=1

1 1 MAE = ∑ PM2.5(MERRA−2)i − PM2.5(ground)i (3) n n RMB = PM2.5MERRA−2/PM2.5ground (4)

The MERRA-2 PM2.5 reanalysis was thought to be reasonably good if RMB ≥ 0.5 [27].

2.2.2. Trend Analysis

Multi-year average trends (1980–2017) in PM2.5 and its major components (including SO4, BC, OC, Dust2.5 and SS2.5) were assessed based on a linear regression. Furthermore, their corresponding signiﬁcant levels were calculated using the Mann-Kendall (MK) nonparametric method. They were described by Equations (5)–(8): n−1 n S = ∑ ∑ sgn Xj − Xi (5) i=1 j=n+1   +1, i f Xj − Xi > 0  sgn Xj − Xi = 0, i f Xj − Xi = 0 (6)   −1, i f Xj − Xi < 0 q n(n − 1)(2n + 5) − ∑p=1 tp tp − 1 2tp + 5 Var(s) = (7) 18 Remote Sens. 2019, 11, 460 7 of 25

 − √S 1 , i f S > 0  Var(S)  Z = 0, i f S = 0 (8)   √S+1 , i f S < 0  Var(S) where S is the statistical value, Var (S) is the variance of the statistical value S and Xj and Xi represent the annual mean concentrations of the total PM2.5 or its major components in the jth and ith years, respectively. The trend was signiﬁcant (i.e., p-value < 0.05) only when the standardized test statistics |Z| > |Z(1 − α/2)|.

2.2.3. Multiple Linear Regression

In order to assess the comprehensive impact of various meteorological factors on PM2.5, the following multiple linear regression (MLR) model was used:

6 y = β0 + ∑ βixi + εi (9) i=1 where the dependent variable y refers to daily mean PM2.5 concentrations (including PM2.5, SO4, ◦ ◦ OC, BC, Dust2.5 and SS2.5) in each 0.05 × 0.625 grid, β0 is a constant term, xi is the daily mean meteorological factor described in Section 2.1 and εi is the interaction term. Moreover, the slope coefﬁcient βi can be interpreted as the sensitivity of PM2.5 to changes in one meteorological factor, assuming that the other meteorological variables remain ﬁxed. This was determined via a least-square adjustment [26,27].

3. Results

3.1. Ground-Based Veriﬁcation of MERRA-2 PM2.5

3.1.1. Regional and Seasonal Comparisons between MERRA-2 and Ground PM2.5

Comparisons of daily mean PM2.5 concentrations derived from MERRA-2 and 476 ground sites over the YRB are presented in Figure2. The comparison over the YRD (Figure2b) was superior to other regions of the YRB, with higher R (0.58) and RMB (0.90) values. By contrast, the SB (Figure2d) produced the worst performance, with the lowest R (0.31) and RMB (0.55) values. The unique basin topography was prone to cloudy weather, which may restrict the assimilation of MERRA-2 PM2.5 products [37]. −3 Generally, MERRA-2 produced lower daily mean PM2.5 concentrations (4.77–23.26 µg m ) and the −3 bias was more outstanding for high ground PM2.5 (>75 µg m ). For instance, MERRA-2 PM2.5 −3 concentrations were entirely less than ground observations when PM2.5 > 75 µg m . The lack of nitrate in the MERRA-2 PM2.5 reanalysis would explain the underestimation [37,40]. In addition, to a certain extent, the bias caused by MERRA-2 AOD assimilation might be another reason for the MERRA-2 PM2.5 underestimation, which will be veriﬁed in detail in the next section. Notably, in Figure2d, no matter how large the ground PM 2.5 concentrations were, MERRA-2 PM2.5 remained small. The reason might be that the unique basin terrain was prone to cloudy weather in the SB, and thus probably led to cloud screening bias for MERRA-2 aerosol reanalysis [8,9,40]. Remote Sens. 2019, 11, x FOR PEER REVIEW 8 of 27

other regions of the YRB, with higher R (0.58) and RMB (0.90) values. By contrast, the SB (Figure 2(d)) produced the worst performance, with the lowest R (0.31) and RMB (0.55) values. The unique basin topography was prone to cloudy weather, which may restrict the assimilation of MERRA-2 PM2.5 products [37]. Generally, MERRA-2 produced lower daily mean PM2.5 concentrations (4.77 - 23.26 µg m-3) and the bias was more outstanding for high ground PM2.5 (> 75 µg m-3). For instance, MERRA-2 PM2.5 concentrations were entirely less than ground observations when PM2.5 > 75 µg m-3. The lack of nitrate in the MERRA-2 PM2.5 reanalysis would explain the underestimation [37,40]. In addition, to a certain extent, the bias caused by MERRA-2 AOD assimilation might be another reason for the MERRA-2 PM2.5 underestimation, which will be verified in detail in the next section. Notably, in Figure 2(d), no matter how large the ground PM2.5 concentrations were, MERRA-2 PM2.5 remained small. The reason might be that the unique basin terrain was prone to cloudy Remoteweather Sens. 2019 in ,the11, 460SB, and thus probably led to cloud screening bias for MERRA-2 aerosol reanalysis8 of 25 [8,9,40].

Figure 2. Scatter density maps between MERRA-2 and ground-based daily mean PM2.5 concentrations Figure 2. Scatter density maps between MERRA-2 and ground-based daily mean PM2.5 over the (a) YRB, (b) YRD, (c) CC, (d) SB and (e) SYR for the period of Jan. 2015–Dec. 2016. concentrations over the (a) YRB, (b) YRD, (c) CC, (d) SB and (e) SYR for the period of Jan. 2015–Dec. The parameter N represents the number of matchups, R is the correlation coefﬁcient, RMSE is the root 2016. The parameter N represents the number of matchups, R is the correlation coefficient, RMSE is mean square error, RMB is relative mean bias and MAE is the mean absolute error. The color bars the root mean square error, RMB is relative mean bias and MAE is the mean absolute error. The representcolor bars the represent density ofthe the density matchups. of the matchups. Figure3 depicts the seasonal cycles of PM concentrations from MERRA-2 reanalysis data Figure 3 depicts the seasonal cycles of PM2.5 2.5concentrations from MERRA-2 reanalysis data and andground ground measurements. measurements. The The seasonal seasonal comparison comparison between between the MERRA-2 the MERRA-2 and ground and ground PM2.5 was PM 2.5 wascalculated calculated as as follows. follows. First, First, according according to toFigure Figure 1, 1 the, the YRB YRB was was divided divided into into four four sub-regions sub-regions includingincluding the the YRD, YRD, CC, CC, SB SB and and SYR, SYR, withwith 87,87, 233,233, 95 and and 60 60 state-controlled state-controlled air air quality quality monitoring monitoring stations, respectively. Then, based on the arithmetic averaging method, the seasonal average ground-based PM2.5 concentrations in each sub-region were calculated by the daily average PM2.5 from state-controlled air quality monitoring stations. Finally, the daily mean grid values of MERRA-2 PM2.5 where the ground site was located were extracted from the MERRA-2 satellite images. Similarly, the seasonal average MERRA-2 PM2.5 concentrations in each sub-region were calculated by the arithmetic averaging method. Results showed that both PM2.5 products had similar seasonal variations in all regions, with high values in winter and low values in summer. Additionally, the bias also exhibited seasonal dependence. MERRA-2 PM2.5 matched better with ground observations in summer, when the biases were only 4.16 µg m−3 (YRD), −1.07 µg m−3 (CC), −9.55 µg m−3 (SB), −5.39 µg m−3 (SYR) and −2.31 µg m−3 (YRB). The largest biases were observed in winter, reaching −15.83 µg m−3 (YRD), −27.45 µg m−3 (CC), −43.73 µg m−3 (SB), −19.77 µg m−3 (SYR) and −27.67 µg m−3 (YRB). Similar comparison results were also reported over America [37] and the North China plain [40]. Remote Sens. 2019, 11, x FOR PEER REVIEW 9 of 27

stations, respectively. Then, based on the arithmetic averaging method, the seasonal average ground-based PM2.5 concentrations in each sub-region were calculated by the daily average PM2.5 from state-controlled air quality monitoring stations. Finally, the daily mean grid values of MERRA-2 PM2.5 where the ground site was located were extracted from the MERRA-2 satellite images. Similarly, the seasonal average MERRA-2 PM2.5 concentrations in each sub-region were calculated by the arithmetic averaging method. Results showed that both PM2.5 products had similar seasonal variations in all regions, with high values in winter and low values in summer. Additionally, the bias also exhibited seasonal dependence. MERRA-2 PM2.5 matched better with ground observations in summer, when the biases were only 4.16 µg m-3 (YRD), -1.07 µg m-3 (CC), -9.55 µg m-3 (SB), -5.39 µg m-3 (SYR) and -2.31 µg m-3 (YRB). The largest biases were observed in winter, reaching -15.83 µg m-3 (YRD), -27.45 µg m-3 (CC), -43.73 µg m-3 (SB), -19.77 µg m-3 (SYR) and Remote Sens. 2019, 11, 460 9 of 25 -27.67 µg m-3 (YRB). Similar comparison results were also reported over America [37] and the North China plain [40].

FigureFigure 3. Monthly 3. Monthly variations variations in in PM PM2.52.5concentrations concentrations derived derived from MERRA-2 (red (red dashed dashed lines) lines) reanalysisreanalysis and and ground ground sites sites (blue (blue dashed dashed lines) lines) over over the the ( a(a)) YRB, YRB, ((bb)) YRD,YRD, ( c)) CC, CC, ( (dd)) SB SB and and (e ()e SYR) SYR for thefor periodthe period of Jan. of 2015–Dec.Jan. 2015–Dec. 2016. 2016. The The shading shading represents represents the standardthe standard deviation deviation of MERRA-2 of MERRA-2 (red)

and(red) ground-based and ground-based PM2.5 (gray). PM2.5 (gray).

3.1.2.3.1.2 Impact Impact of AODof AOD Assimilation assimilation

Here, the effect of MERRA-2 AOD assimilation on the PM2.5 simulation was estimated. Since there were no CARSNET AOD observations in the MERRA-2 aerosol assimilation, we compared daily mean MERRA-2 AOD with those ground measurements from ﬁve CARSNET sites of the YRB during 2013-2016 (Figure4). As shown in Figure4a, although MERRA-2 AOD was in good agreement with CARSNET AOD (R = 0.82), approximately 19% of MERRA-2 AOD were still underestimated, especially when CARSNET AOD > 1.0. This was further conﬁrmed by Figure4b. The high MERRA-2 AOD values in Wuhan and Chengdu (>0.8) were all below their corresponding CARSNET observations. By contrast, Shangri-La with low mean AOD (<0.2) exhibited a good performance. Compared with the remote site of Shangri-La, Wuhan and Chengdu were two urban CARSNET sites with more complex aerosol sources and more diverse underlying surfaces, probably restraining the assimilation of the MERRA-2 AOD reanalysis [37,40]. Overall, the underestimation of MERRA-2 PM2.5 was partly attributed to the negative bias of MERRA-2 AOD used in the GOCART model. Remote Sens. 2019, 11, x FOR PEER REVIEW 10 of 27

Here, the effect of MERRA-2 AOD assimilation on the PM2.5 simulation was estimated. Since there were no CARSNET AOD observations in the MERRA-2 aerosol assimilation, we compared daily mean MERRA-2 AOD with those ground measurements from five CARSNET sites of the YRB during 2013-2016 (Figure 4). As shown in Figure 4(a), although MERRA-2 AOD was in good agreement with CARSNET AOD (R = 0.82), approximately 19% of MERRA-2 AOD were still underestimated, especially when CARSNET AOD > 1.0. This was further confirmed by Figure 4(b). The high MERRA-2 AOD values in Wuhan and Chengdu (> 0.8) were all below their corresponding CARSNET observations. By contrast, Shangri-La with low mean AOD (< 0.2) exhibited a good performance. Compared with the remote site of Shangri-La, Wuhan and Chengdu were two urban CARSNET sites with more complex aerosol sources and more diverse underlying surfaces, probably restraining the assimilation of the MERRA-2 AOD reanalysis [37,40]. Overall, the Remote Sens. 2019, 11, 460 10 of 25 underestimation of MERRA-2 PM2.5 was partly attributed to the negative bias of MERRA-2 AOD used in the GOCART model.

Figure 4. Scatter plot of daily averaged (a) and annual mean AOD (b) between MERRA-2 and CARSNETFigure 4. AOD Scatter at ﬁve plot ground of daily sites averaged of the (YRBa) and during annual 2013–2016. mean AOD In (b (b),) the between vertical MERRA-2 and horizontal and barsCARSNET refer to the AOD standard at five deviationsground sites of theof the annual YRB meanduring MERRA-2 2013–2016. and In Figure CARSNET 4(b), AOD,the vertical respectively. and horizontal bars refer to the standard deviations of the annual mean MERRA-2 and CARSNET AOD, Figurerespectively.5 depicts AOD daily variations from Aqua, Terra, CARSNET and MERRA-2 products at Wuhan. As shown in Figure5a, SO 4 dominated the aerosol types of Wuhan for the whole year. Figure 5 depicts AOD daily variations from Aqua, Terra, CARSNET and MERRA-2 products at All AOD products showed similar daily variations, i.e., high values approximately occurred between Wuhan. As shown in Figure 5(a), SO4 dominated the aerosol types of Wuhan for the whole year. All the 50th and the 150th days. In addition, both satellite (Aqua and Terra) and ground (CARSNET) AOD AOD products showed similar daily variations, i.e., high values approximately occurred between observations were higher than MERRA-2 AODs. Additionally, is worth noting that neither the satellites the 50th and the 150th days. In addition, both satellite (Aqua and Terra) and ground (CARSNET) norAOD sun-photometer observations providedwere higher continuous than MERRA-2 AOD observations,AODs. Additionally, especially is worth during noting a heavy that neither pollution episode (Figure5b). This would make the assimilation of MERRA-2 AOD impossible, implying that Remotethe satellites Sens. 2019 , nor11, x FOR sun-photometer PEER REVIEW provided continuous AOD observations, especially during 11 of 27 a theheavy GOCART pollution model episode for simulating (Figure aerosols 5(b)). This in MERRA-2 would make needed the further assimilation improvement. of MERRA-2 AOD impossible, implying that the GOCART model for simulating aerosols in MERRA-2 needed further improvement.

Figure 5. (a) Daily variations in AOD from satellite retrievals (Aqua and Terra), ground observations Figure 5(a). Daily variations in AOD from satellite retrievals (Aqua and Terra), ground observations (CARSNET) and MERRA-2 reanalysis at Wuhan in 2016. The contributions of each aerosol species (CARSNET) and MERRA-2 reanalysis at Wuhan in 2016. The contributions of each aerosol species derived from MERRA-2 reanalysis are also shown. (b) Daily variations in AOD during a heavy derived from MERRA-2 reanalysis are also shown. Figure 5(b). Daily variations in AOD during a pollution episode. heavy pollution episode.

3.2. Spatiotemporal Distribution of MERRA-2 PM2.5 Components 3.2 Spatiotemporal Distribution of MERRA-2 PM2.5 Components

3.2.1. Annual Mean Distribution in MERRA-2 PM2.5 Components 3.2.1 Annual mean distribution in MERRA-2 PM2.5 components PM2.5 components varied by region due to differences in anthropogenic emissions and PM2.5 components varied by region due to differences in anthropogenic emissions and meteorological conditions. The spatial distributions of annual mean MERRA-2 PM2.5 components meteorological conditions. The spatial distributions of annual mean MERRA-2 PM2.5 components were illustrated in Figure6. In Figure6a, high PM 2.5 concentrations were observed over the YRD, were illustrated in Figure 6. In Figure 6(a), high PM2.5 concentrations were observed over the YRD, CC and SB, while low values were concentrated in the SYR. A similar spatial pattern of PM2.5 was CC and SB, while low values were concentrated in the SYR. A similar spatial pattern of PM2.5 was also conﬁrmed by the GWR PM (Figure A2). Previous studies reported that GWR PM reanalysis also confirmed by the GWR PM2.52.5 (Figure A2). Previous studies reported that GWR PM2.5 2.5reanalysis yieldedyielded good good performance performance relative relative to ground to ground observations observations [41 [41,42].,42]. Therefore, Therefore, we we used used it hereit here to verifyto verify the spatial distribution of MERRA-2 PM2.5 reanalysis. According to the ambient air quality standard established by the World Health Organization (WHO), the annual mean PM2.5 concentration should not exceed 10 µg/m3 [23]. Approximately 67.49% and 78.27% of the pixels in the YRB exceeded the threshold for MERRA-2 and GWR, respectively (Figure A2). This fact suggested that it was necessary to clarify the composition of PM2.5 and its influencing factors over the YRB. In terms of each PM2.5 component, sulfate PM2.5 (SO4) occupied an average of 39.87% of the PM2.5 concentration, and its high values were concentrated in YRD, CC and SB. Similarly, high carbonaceous PM2.5 concentrations (BC and OC) were also observed in those regions, accounting for 8.02% and 28.32% of the PM2.5 concentrations, respectively. Emissions from power plants, vehicle exhaust and biomass burning could certainly explain the high PM2.5 concentrations in those areas. For dust PM2.5 (Dust2.5), it accounted for 22.75% of the PM2.5 on average. In particular, high Dust2.5 concentrations appeared in the northern part of the YRB, which primarily came from local arid sources. Another region with high Dust2.5 concentrations was the SYR, possibly due to dust transportation from the Taklimakan desert in summer [7,8,43]. MERRA-2 sea salt PM2.5 (SS2.5) constituted the lowest proportion of the total (approximately 1.02%), and rarely penetrated into land.

Remote Sens. 2019, 11, 460 11 of 25

the spatial distribution of MERRA-2 PM2.5 reanalysis. According to the ambient air quality standard established by the World Health Organization (WHO), the annual mean PM2.5 concentration should not exceed 10 µg/m3 [23]. Approximately 67.49% and 78.27% of the pixels in the YRB exceeded the threshold for MERRA-2 and GWR, respectively (Figure A2). This fact suggested that it was necessary to clarify the composition of PM2.5 and its inﬂuencing factors over the YRB. In terms of each PM2.5 component, sulfate PM2.5 (SO4) occupied an average of 39.87% of the PM2.5 concentration, and its high values were concentrated in YRD, CC and SB. Similarly, high carbonaceous PM2.5 concentrations (BC and OC) were also observed in those regions, accounting for 8.02% and 28.32% of the PM2.5 concentrations, respectively. Emissions from power plants, vehicle exhaust and biomass burning could certainly explain the high PM2.5 concentrations in those areas. For dust PM2.5 (Dust2.5), it accounted for 22.75% of the PM2.5 on average. In particular, high Dust2.5 concentrations appeared in the northern part of the YRB, which primarily came from local arid sources. Another region with high Dust2.5 concentrations was the SYR, possibly due to dust transportation from the Taklimakan desert in summer [7,8,43]. MERRA-2 sea salt PM2.5 (SS2.5) constituted the lowest proportion of the totalRemote (approximately Sens. 2019, 11, x FOR 1.02%), PEER and REVIEW rarely penetrated into land. 12 of 27

Figure 6. Annual mean distributions of each MERRA-2 PM2.5 component, including (a) the PM2.5, Figure 6 Annual mean distributions of each MERRA-2 PM2.5 component, including (a) the PM2.5, (b) (b) Sulfate PM2.5,(c) Black Carbon PM2.5,(d) Organic Carbon PM2.5,(e) Dust PM2.5 and (f) Sea Salt Sulfate PM2.5, (c) Black Carbon PM2.5, (d) Organic Carbon PM2.5, (e) Dust PM2.5 and (f) Sea Salt PM2.5 PM over the YRB during 1980–2017. over2.5 the YRB during 1980–2017.

3.2.2. Seasonal Variation in MERRA-2 PM2.5 Components 3.2.2 Seasonal variation in MERRA-2 PM2.5 components Figures7 and8 showed the spatial and seasonal differences of each PM 2.5 component. A signiﬁcant Figure 7 and Figure 8 showed the spatial and seasonal differences of each PM2.5 component. A seasonal variation of PM2.5 was observed over the YRB, especially over the YRD, CC and SB. significant seasonal variation of PM2.5 was observed over the YRB, especially over the YRD, CC and On average, its highest concentration of 26.63 µg m−3 occurred in winter and the lowest, 18.47 SB. On average, its highest concentration of 26.63 µg m−3 occurred in winter and the lowest, 18.47 µ −3 gµg m m,−3 appeared, appeared in in summer summer (Table (Table3 ). 3). This This was was associated associated with with the the prevailing prevailing local local monsoon monsoon climate [31]. In summer, windy and rainy weather was conducive to scavenging PM particles from climate [31]. In summer, windy and rainy weather was conducive to scavenging PM2.5 2.5 particles thefrom atmosphere, the atmosphere, whereas whereas the dry weather the dry weather affected affectedby the Mongolian by the Mongolian anticyclone anticyclone in winter in prevented winter theprevented diffusion ofthe aerosol diffusion particles. of aerosol The particles. increase inThe emissions increase causedin emissions by heating caused in by winter heating could in winter certainly be anothercould certainly reason forbe another the high reason PM2.5 forconcentration the high PM [312.5]. concentration For each PM 2.5[31].component, For each PM seasonal2.5 component, variations in SOseasonal4, BC and variations OC were in consistent SO4, BC and with OC those were of consistent PM2.5. Nevertheless, with those of SS 2.5 PMwitnessed2.5. Nevertheless, little seasonal SS2.5 −3 changeswitnessed over little the YRB.seasonal Dust changes2.5 concentration over the YRB. was Dust highest2.5 concentration in spring (8.63 wasµ highestg m ), in and spring most (8.63 occurred µg in northernm−3), and parts most of occurred the YRB. in This northern was related parts of to thefrequent YRB. sandstorm This was related events to in frequentspring [35 sandstorm,36]. events in spring [35, 36].

Table 3 Annual and seasonal mean values of each PM2.5 component over the YRB during 1980–2017, in units of µg m−3.

Remote Sens. 2019, 11, 460 12 of 25

Table 3. Annual and seasonal mean values of each PM2.5 component over the YRB during 1980–2017, in units of µg m−3.

PM2.5 SO4 BC OC Dust2.5 SS2.5 Annual 23.43 6.80 1.88 4.74 5.33 0.24 MAM 26.00 6.75 1.68 4.85 8.63 0.26 JJA 18.47 5.47 1.44 4.16 3.45 0.23 SON 22.63 7.13 1.98 4.45 4.37 0.24 DJF 26.63 7.83 2.42 5.50 4.88 0.23 Remote Sens. 2019, 11, x FOR PEER REVIEW 13 of 27 Remote Sens. 2019, 11, x FOR PEER REVIEW 13 of 27

Figure 7. Seasonal variations of each MERRA-2 PM2.5 component: (left) the PM2.5, (middle) SO4 and Figure 7. Seasonal variations of each MERRA-2 PM2.5 component: (left) the PM2.5,(middle) SO4 and (Figureright) BC7. Seasonal over the variationsYRB during of 1980–2017. each MERRA-2 PM2.5 component: (left) the PM2.5, (middle) SO4 and (right) BC(right over) theBC over YRB the during YRB during 1980–2017. 1980–2017.

Figure 8. Seasonal variations of each MERRA-2 PM2.5 component: (left) OC, (middle) Dust2.5 and 2.5 2.5 Figure 8. Seasonal variations of each MERRA-2(right PM) SS2.5 .component: (left) OC, (middle) Dust and Figure 8. Seasonal variations of each MERRA-2 PM2.5 component: (left) OC, (middle) Dust2.5 and (right) SS2.5. (right) SS2.5.

Remote Sens. 2019, 11, 460 13 of 25

Remote Sens. 2019, 11, x FOR PEER REVIEW 14 of 27 3.2.3. Long-Term Trends of MERRA-2 PM2.5 Components

3.2.3Figure Long-term9 exhibited trends the of interannual MERRA-2 PM variability2.5 components of each MERRA-2 PM 2.5 component over the YRB µ −3 −1 during 1980–2017.Figure 9 exhibited A significant the interannual increasing variability trend of of 11.4E-4 each MERRA-2g m PMyr 2.5 componentwas observed over in the PM YRB2.5 for theduring whole 1980–2017. period. Notably, A significant the growth increasing trend trend of PM of2.5 11.4E-4was effectively µg m-3 yr-1 curbed was observed after 2007. in PM To2.5 reduce for the the uncertaintywhole period. in the Notably, MERRA-2 the PM growth2.5 trend, trend we of also PM compared2.5 was effectively it with thecurbed interannual after 2007. variability To reduce of the GWR PMuncertainty2.5 from 1998 in tothe 2017 MERRA-2 (Figure PMA3).2.5 Thetrend, results we also showed compared that both it with PM 2.5theproducts interannual had variability similar growth of trends,GWR and PM there2.5 from was 1998 a to significant 2017 (Figure turning A3). The point results from showed increasing that toboth decreasing PM2.5 products around had 2007–2009. similar Similarly,growth He trends, et al. and [10] there revealed was that a significant an over-increasing turning point trend from in MERRA-2 increasing AOD to decreasing over the around YRB was curbed2007–2009. after 2008. Similarly, The reason He et wasal. [10] probably revealed due that to thean effectiveover-increasing implementation trend in MERRA-2 of energy AOD conservation over the YRB was curbed after 2008. The reason was probably due to the effective implementation of and emission reduction policies in China [10]. Comparable trends were also found in SO4, BC and OC, confirmingenergy conservation that reduction and in emission anthropogenic reduction emissions policies was in theChina dominant [10]. Comparable factor for thetrends improvement were also in found in SO4, BC and OC, confirming that reduction in anthropogenic emissions was the dominant−3 −1 air quality. In addition, Dust2.5 and SS2.5 witnessed slightly increasing trends of 7.42E-5 µg m yr factor for the improvement in air quality. In addition, Dust2.5 and SS2.5 witnessed slightly increasing and 1.63E-5 µg m−3 yr−1 after 1980, respectively. Nevetheless, whether meterological factors impacted trends of 7.42E-5 µg m−3 yr−1 and 1.63E-5 µg m−3 yr−1 after 1980, respectively. Nevetheless, whether air quality requires further analysis. meterological factors impacted air quality requires further analysis.

Figure 9. Annual mean trends in MERRA-2 PM and its ﬁve major components over the YRB from Figure 9. Annual mean trends in MERRA-2 PM2.52.5 and its five major components over the YRB from −3 19801980 to 2017,to 2017, in unitsin units of ofµg µg m m-3..

TheThe annual annual mean mean trends trends of of each each MERRA-2MERRA-2 PM PM2.52.5 componentcomponent at atgrid grid level level were were estimated estimated by a by a linearlinear regression, regression, andand theirtheir correspondingcorresponding significant significant levels levels were were further further calculated calculated based based on onthe the −3 −1 MKMK method method (Figure (Figure 10 ). 10). Overall, Overall, a a significant significant increasing increasing trend trend in in PM PM2.52.5 (0.7 (0.7µgµg m m−3 yryr−1) was) was observedobserved over over the the YRD, YRD, CC CC and and SB. SB.However, However, GWR GWR exhibited exhibited a significant a significant decreasing decreasing trend trend in PM in 2.5 in thePM2.5 mountains in the mountains of the of northern the northern YRB YRB (Figure (Figure A4 ),A4), which which was was not not observed observed in in MERRA-2 MERRA-2 PM PM2.5.2.5 . RecentRecent studies studies demonstrated demonstrated that that thethe decreasing PM PM2.52.5 trendtrend in inthose those regions regions might might be a beresponse a response to to thethe implementationimplementation of of forest forest ecology ecology projects projects over over the the middle middle and and upper upper reaches reaches of ofthe the YRB YRB since since the 1990s [44]. In terms of each PM2.5 component, SO4, BC and OC witnessed significant increases, the 1990s [44]. In terms of each PM2.5 component, SO4, BC and OC witnessed significant increases,

Remote Sens. 2019, 11, 460 14 of 25

Remote Sens. 2019, 11, x FOR PEER REVIEW 15 of 27 especially over the YRD, CC and SB. By contrast, signiﬁcant growth trends in Dust2.5 and SS2.5 occurred especially over the YRD, CC and SB. By contrast, significant growth trends in Dust2.5 and SS2.5 over the SYR and the eastern coastal regions, respectively. occurred over the SYR and the eastern coastal regions, respectively.

Figure 10. Spatial patterns of each MERRA-2 PM component over the YRB from 1980 to 2017. Figure 10. Spatial patterns of each MERRA-2 PM2.5 component2.5 over the YRB from 1980 to 2017. The Thepixels pixels marked marked with with black black dots dots represent represent significant significant trends trends (p < 0.05). (p < 0.05). 4. Discussion 4. Discussion To quantify the impact of meteorological variables on PM2.5 changes, the slope coefficients βi To quantify the impact of meteorological variables on PM2.5 changes, the slope coefficients βi between daily mean PM2.5 concentrations and six meteorological factors from 1980 to 2017 were between daily mean PM2.5 concentrations and six meteorological factors from 1980 to 2017 were calculatedcalculated using using the the multiple multiple linear linear regression regression (MLR)(MLR) model (Equation (Equation 9). (9)). Assuming Assuming that that all other all other meteorologicalmeteorological variables variables remained remained fixed, fixed, these these slope coefficientsslope coefficients actually actually represented represented the sensitivities the ofsensitivities PM2.5 to unit of PM changes2.5 to unit in changes each meteorological in each meteorological factor. Infactor. the In MLR the MLR process, process, there there should should be no multicollinearitybe no multicollinearity between between the independent the independent variables variables (i.e., (i.e., meteorological meteorological factors), factors), otherwise otherwise they wouldthey bewould eliminated be eliminated [26,27 ]. [26, In 27]. the correlationIn the correlation matrix matrix (Table (Table A1), evenA1), eventhough though there there were were several correlations of more than 0.5, most of them failed to pass the 95% significance test, indicating that

Remote Sens. 2019, 11, 460 15 of 25

Remote Sens. 2019, 11, x FOR PEER REVIEW 16 of 27 each meteorological variable was not signiﬁcantly related to any other. Therefore, we concluded that collinearityseveral correlations was minimal of and more the than MLR 0.5, analysis most of could them be failed performed. to pass Figure the 95%11 showed significance the adjusted test, coefﬁcientsindicating of that determination each meteorological (adjusted-R variable2) at was grid not level. significantly Higher adjusted-Rrelated to any2 values other. indicated Therefore, that we concluded that collinearity was minimal and the MLR analysis could be performed. Figure 11 the MLR model performed better. The highest adjusted-R2 values (~0.67) were mostly concentrated showed the adjusted coefficients of determination (adjusted-R2) at grid level. Higher adjusted-R2 over the SYR, indicating that the local meteorological conditions could explain up to 67% of the values indicated that the MLR model performed better. The highest adjusted-R2 values (~0.67) were daily PM changes. By comparison, the values were lowest over the YRD and CC, where less mostly 2.5concentrated over the SYR, indicating that the local meteorological conditions could explain than 20% of the daily PM changes were due to meteorological variations. Over those regions, up to 67% of the daily PM2.52.5 changes. By comparison, the values were lowest over the YRD and CC, anthropogenic emissions from frequent industrialization and urbanization may be the major reason where less than 20% of the daily PM2.5 changes were due to meteorological variations. Over those 2 forregions, high PM anthropogenic2.5 concentrations. emissions In Figure from A5 frequent, adjusted-R industrializationvalues showed and seasonal urbanization variations, may with be the high 2 valuesmajor (adjusted-R reason for > high 0.5) in PM spring2.5 concentrations. and summer In over Figure the SYR, A5, suggesting adjusted-R2 that values the local showed meteorological seasonal conditionsvariations, could with accounthigh values for (adjusted-R more than250% > 0.5) of in PM spring2.5 changes. and summer This over may the be SYR, closely suggesting related that tothe 2 2 summerthe local monsoon meteorological climate conditions [26,27]. However, could account the relatively for more low than adjusted-R 50% of PM2.5values changes.(adjusted-R This may

22 FigureFigure 11. 11.Coefﬁcients Coefficients of of determination determination (R )) of of the the MLR MLR model model between between PM PM2.5 2.5 andand the the six six meteorologicalmeteorological variables. variables.

4.1.4.1 Sensitivity Sensitivity of of PM PM2.52.5Components Components to Surface Surface Wind Wind Speed Speed

FigureFigure 12 12depicts depicts correlations correlations of of each each PM PM2.52.5component component with withsurface surface windwind speed,speed, asas measuredmeasured by theby MLR the regression MLR regression coefficient coefficient (β1) in Equation(β1) in Equation (9). Red 9. grids Red represent grids represent positive positive correlations, correlations, whereas greenwhereas grids aregreen negative grids correlations. are negative In correlations. Figure 12a, In the Figure PM2.5 is 12(a), significantly the PM2.5 anti-correlated is significantly with surfaceanti-correlated wind speed with throughout surface wind the YRB. speed High throughout wind speeds the YRB. facilitated High circulation wind speeds and facilitated dilution of 2.5 aerosolcirculation particles, and leading dilution to of a aerosol consistent particles, decrease leading in PM to a2.5 consistent. Negative decrease correlations in PM between. Negative PM 2.5 concentrationscorrelations between and surface PM2.5 wind concentrations speeds were and also surface observed wind in speeds the United were also States observed of America in the [26 ], United States of America [26], Europe [27–29, 31] and China [32–34]. Yang et al. [33] discovered a Europe [27–29,31] and China [32–34]. Yang et al. [33] discovered a negative wind-PM2.5 correlation throughoutnegative wind-PM China except2.5 correlation Hainan Island.throughout Compared China except with theHainan USA Island. and Europe, Compared China with had the the USA worst and Europe, China had the worst atmospheric dispersion conditions with annual stagnation atmospheric dispersion conditions with annual stagnation frequency of 29%, leading to serious air frequency of 29%, leading to serious air pollutions especially in winter [31]. However, a significant pollutions especially in winter [31]. However, a significant positive correlation occurred in the SYR, positive correlation occurred in the SYR, where a plenty of dust particles were concentrated (see where a plenty of dust particles were concentrated (see Section 3.2). It was probably due to the Section 3.2). It was probably due to the strong wind speed dependence of dust particles, further strong wind speed dependence of dust particles, further confirmed by the Dust -wind regression in confirmed by the Dust2.5-wind regression in Figure 12 (e). Similarly, in Saudi Arabia2.5 (an arid region), Figure 12e. Similarly, in Saudi Arabia (an arid region), both PM and PM were positively related to both PM2.5 and PM10 were positively related to wind speed [30].2.5 Additionally,10 in Figure A6 and windTable speed A2, [most30]. Additionally,of the YRB experienced in Figure A6an andaverage Table decrease A2, most of 0.05 of the m YRBs-1 per experienced year in wind an speed average −1 decreasefrom 1980 of 0.05 to m2017. s perBased year on in the wind MLR speed model, from the 1980 average to 2017. PM Based2.5-wind on regression the MLR model, coefficient the average was −3 −1 PMapproximately2.5-wind regression −2.42 µg coefficient m-3 m-1 s, was leading approximately to an average−2.42 increaseµg m of 0.12m µgs, m leading-3 per year to anin averagePM2.5 −3 increaseover the of YRB. 0.12 µ However,g m per all year other in meteorological PM2.5 over the variables YRB. However, remained all fixed other (Table meteorological A2). variables remained fixed (Table A2).

Remote Sens. 2019, 11, 460 16 of 25 Remote Sens. 2019, 11, x FOR PEER REVIEW 17 of 27

Figure 12. Multiple linear regression coefficients (β1) of each PM2.5 component with surface wind Figure 12. Multiple linear regression−3 −1 coefficients (β1) of each PM2.5 component with surface wind speed (Wind), in units of µg m m s. (a–f) refer to the regression coefficients of PM2.5, SO4, BC, OC, speed (Wind), in units of µg m−3 m−1 s. Figure 12 (a), (b), (c), (d), (e) and (f) refer to the regression Dust2.5 and SS2.5 with surface wind speed, respectively. coefficients of PM2.5, SO4, BC, OC, Dust2.5 and SS2.5 with surface wind speed, respectively. 4.2. Sensitivity of PM2.5 Components to Surface Temperature 4.2 Sensitivity of PM2.5 Components to Surface Temperature In Figure 13, surface temperature (T) was significantly positively correlated with each PM2.5 componentIn Figure across 13, surface the YRB. temperature Similar results (T) werewas significantlyalso reported positively on global [27correlated] and regional with [each26] scales. PM2.5 componentWestervelt across et al. the [27 ]YRB. discovered Similar aresults robust were positive also relationshipreported on betweenglobal [27] surface and regional temperature [26] scales. and WesterveltPM2.5 throughout et al. [27] the discovered continents a and robust oceans positive except relationship East and South between Asia. surfaceThis positive temperature correlation and PMmay2.5 throughout be related tothe temperature-dependent continents and oceans reaction except East rates and and South changes Asia. in oxidantThis positive abundance, correlation that mayis, be the related oxidation to temperature-dependent of aerosol precursors (such reaction as sulfurrates and dioxide) changes could in oxidant be enhanced abundance, with higherthat is, thetemperature oxidation [27 of]. aerosol However, precursors in China, Yang (such et al. as [33 sulfur] revealed dioxide) that surface could temperature be enhanced was with negatively higher temperaturerelated to PM [27].2.5 in However, autumn but in positivity China, Yang correlated et al. with [33] PM revealed2.5 in winter. that Dawson surface et temperature al. [45] found was an negativelyaveraged related negative to effect PM2.5 of in temperature autumn but on PMpositivity2.5 east ofcorrelated the USA, with primarily PM2.5 due in towinter. the volatilization Dawson et ofal. [45]nitrate found at an high averaged temperature. negative Since effect nitrate of temperature aerosols wereon PM not2.5 east assimilated of the USA, in the primarily MERRA-2 due PM to 2.5the volatilizationreanalysis, there of nitrate was little at high negative temperature. relationship Since between nitrate PM aerosols2.5 and were temperature not assimilated over the in YRB. the In addition, Figure A6 and Table A2 showed an average increase of 0.03 K per year in temperature from MERRA-2 PM2.5 reanalysis, there was little negative relationship between PM2.5 and temperature over1980 the to YRB. 2017. In Based addition, on the Figure MLR modelA6 and between Table A2 temperature showed an and average the PM increase2.5 (Figure of 0.0313a), K the per average year in −3 −1 −3 sensitivity of 0.11 µg m K could lead to PM2.5 growth of up to 0.003 µg m per year. Overall, temperature from 1980 to 2017. Based on the MLR model between temperature and the PM2.5 −3 −1 the temperature effect on PM2.5 (0.11 µg m K ) was weaker than the influence of wind speed (Figure 13(a)), the average sensitivity of 0.11 µg m−3 K−1 could lead to PM2.5 growth of up to 0.003 µg (−2.42 µg m−3 m−1 s) over the YRB (Table A2). m-3 per year. Overall, the temperature effect on PM2.5 (0.11 µg m-3 K-1) was weaker than the influence of wind speed (−2.42 µg m−3 m−1 s) over the YRB (Table A2).

RemoteRemote Sens. Sens.2019 2019, 11,, 11 460, x FOR PEER REVIEW 18 of17 27 of 25 Remote Sens. 2019, 11, x FOR PEER REVIEW 18 of 27

FigureFigure 13. 13.Multiple Multiple linear linear regression regression coefﬁcients coefficients (β22)) of of each each PM PM2.52.5 componentcomponent with with surface surface Figure 13. Multiple linear regression coefficients (β2) of each PM2.5 component with surface temperaturetemperature (T), (T), in in units units of ofµ µgg mm−-33 KK-1.− Figure1.(a– f13) refer(a), (b), to (c), the (d), regression (e) and (f) coefﬁcients refer to the of regression PM , SO , temperature (T), in units of µg m-3 K-1. Figure 13 (a), (b), (c), (d), (e) and (f) refer to the regression2.5 4 BC,coefficients OC, Dust of andPM2.5 SS, SO4with, BC, OC, surface Dust temperature,2.5 and SS2.5 with respectively. surface temperature, respectively. coefficients2.5 of PM2.5, SO2.54, BC, OC, Dust2.5 and SS2.5 with surface temperature, respectively.

4.3.4.3 Sensitivity Sensitivity of of PM PM2.52.5Components Components to Precipitation 4.3 Sensitivity of PM2.5 Components to Precipitation

FigureFigure 14 14shows shows the the slope slope coefﬁcients coefficients between between eacheach PM2.5 componentcomponent and and precipitation. precipitation. Each Each Figure 14 shows the slope coefficients between each PM2.5 component and precipitation. Each PMPM2.5 2.5component component was wassigniﬁcantly significantly anti-correlated anti-correlated with with precipitation precipitation in in most most areas areas of of the the YRB, YRB, PM2.5 component was significantly anti-correlated with precipitation in most areas of the YRB, primarilyprimarily due due to to wet wet scavengingscavenging [26–35]. [26–35]. However, However, a positive a positive relationship relationship was observed was observed in the SYR, in the SYR,where where dust dust particle particle was was the the main main air air pollutant pollutant (see (see Section Section 3.2) 3.2 and) and precipitation precipitation did did not not occur occur frequentlyfrequently [46 [46–48].–48]. Even Even if if there there was was a a smallsmall amountamount of of precipitation precipitation over over the the SYR, SYR, it may it may have have only only increasedincreased the the relative relative humidity, humidity, resulting resulting inin thethe growthgrowth of hygroscopic hygroscopic particles particles [27]. [27 ].Precipitation Precipitation changeschanges were were typically typically more more positive positive than than negative negative from from 1980 1980 toto 2017,2017, alongalong withwith anan annual trend trend of –1 of 0.02 mm−1 day–1 (Figure A6 and Table A2). Overall, based on the MLR model of PM2.5 and 0.02of mm 0.02 day mm (Figure day (Figure A6 and A6 Table and A2 Table). Overall, A2). Overall, based on based the MLR on the model MLR of PM model2.5 and of PMprecipitation2.5 and precipitation (Figure 14(a)), the average sensitivity of −0.34 µg m−3 mm−1 day could lead to a (Figureprecipitation 14a), the (Figure average 14(a)), sensitivity the average of −0.34 sensitivityµg m−3 mm of −0.34−1 day µg could m−3 leadmm−1to day a decrease could lead of 0.007 to a µg −3 −decrease3 of 0.007 µg m−3 per year for PM2.5, which was comparable to the effect of temperature on m decreaseper year of for0.007 PM µg2.5 ,m which per year was comparablefor PM2.5, which to the was effect comparable of temperature to the effect on PM of 2.5temperature(Table A2 ).on PM2.5 (Table A2). PM2.5 (Table A2).

Figure 14. Multiple linear regression coefﬁcients (β3) of each PM2.5 component with surface precipitation (Precip), in units of µg m−3 mm−1 day. (a–f) refer to the regression coefﬁcients of PM , SO , BC, OC, Dust and SS with precipitation, respectively. 2.5 4 2.5 2.5 Remote Sens. 2019, 11, x FOR PEER REVIEW 19 of 27

Figure 14. Multiple linear regression coefficients (β3) of each PM2.5 component with surface Remote Sens.precipitation2019, 11, 460 (Precip), in units of µg m−3 mm−1 day. Figure 14 (a), (b), (c), (d), (e) and (f) refer to the 18 of 25

regression coefficients of PM2.5, SO4, BC, OC, Dust2.5 and SS2.5 with precipitation, respectively.

4.4. Sensitivity of PM2.5 Components to Relative Humidity 4.4 Sensitivity of PM2.5 Components to Relative Humidity Figure 15 shows the slope coefficients for each PM component and relative humidity (RH). Figure 15 shows the slope coefficients for each PM2.5 component and relative humidity (RH). As canAs can be seenbe seen in Figure in Figure 15a, 15(a), PM2.5 PMwas2.5positively was positively correlated correlated with RHwith in RH most in areasmost ofareas the of YRB, the withYRB, the exceptionwith the of exception the YRD. of Throughout the YRD. Throughout the YRB, significant the YRB, significant positive correlations positive correlations with RH with were RH also were found in SOalso4, BCfound and in OC, SO4, likely BC and due OC, to thelikely dependence due to the of dependence sulfate and of carbonaceous sulfate and carbonaceous PM2.5 formation PM2.5 on RHformation [26,27]. In on contrast, RH [26, Dust 27].2.5 Inhad contrast, an anti-correlation Dust2.5 had with an anti-correlation RH, which probably with RH, explained which the probably negative associationexplained of the PM negative2.5 with RHassociation in theYRD. of PM Yang2.5 with et al.RH [ 33in] the found YRD. apositive Yang et correlational. [33] found between a positive PM 2.5 andcorrelation RH in north between China PM and2.5 a and negative RH in correlation north China in and south a negative China.A correlation negative PM in south2.5-RH China. regression A wasnegative also estimated PM2.5-RH over regression the Pearl was River also Delta estimated Region [over34]. They the Pearl considered River Delta that when Region RH [34]. was They greater considered that when RH was greater than 80%, rainfall often occurred, leading to a decrease in than 80%, rainfall often occurred, leading to a decrease in PM2.5 in the atmosphere [34]. In this study, YRDPM had2.5 in a the temperate atmosphere monsoon [34]. In climate, this study, which YRD made hadprecipitation a temperate monsoon events occur climate, in most which days made with precipitation events occur in most days with high RH (> 80%), especially in summer. Thus, there high RH (>80%), especially in summer. Thus, there was a negative PM2.5-RH regression over the was a negative PM2.5-RH regression over the YRD. However, with the relatively low − slope3 − 1 YRD. However, with the relatively low slope coefficient for total PM2.5 and RH (0.05 µg m % ), coefficient for total PM2.5 and RH (0.05 µg m-3 %-1), the sensitivity of PM2.5 to RH was small (Table the sensitivity of PM to RH was small (Table A2). A2). 2.5

Figure 15. Multiple linear regression coefﬁcients (β ) of each PM component with relative humidity Figure 15. Multiple linear regression coefficients4 (β4) of each2.5 PM2.5 component with relative −3 −1 (RH),humidity in units (RH), of µ ing munits% of µg. Them−3 % blank−1. The grids blank indicate grids indicate invalid invalid values. values. (a–f) Figure refer to 15 the (a), regression (b), (c), coefﬁcients(d), (e) and of PM(f) refer2.5, SO to 4the, BC, regression OC, Dust coefficients2.5 and SS 2.5of PMwith2.5, relativeSO4, BC, humidity, OC, Dust2.5 respectively. and SS2.5 with relative humidity, respectively. 4.5. Sensitivity of PM2.5 Components to Total Cloud Cover

4.5 Sensitivity of PM2.5 Components to Total Cloud Cover In Figure 16, anti-correlations between each PM2.5 component and the total cloud cover were observedIn mostlyFigure 16, over anti-correlations the middle and between lower each reaches PM2.5 ofcomponent the YRB. and In thosethe total regions, cloud cloudscover were could partiallyobserved prevent mostly aerosol over precursors the middle (such and lower as sulfur reaches dioxide) of the from YRB. oxidizing, In those and regions, thus, clouds reduced could PM 2.5 concentrationspartially prevent in the aerosol atmosphere precursors [27]. However, (such as sulfur positive dioxide) regressions from between oxidizing, each and PM thus,2.5 component reduced andPM the2.5 total concentrations cloud cover in occurred the atmosphere in the SYR, [27]. possibly However, due to positive the enhancements regressions in between in-cloud each production PM2.5 of sulfatecomponent aerosols. and Overall,the total thecloud average cover sensitivity occurred in of PMthe 2.5SYR,to thepossibly total clouddue to cover the enhancements was approximately in − − −0.07in-cloudµg m production3 % 1 (Figure of sulfate 16a and aerosols. Table A2Overall,). the average sensitivity of PM2.5 to the total cloud cover was approximately −0.07 µg m−3 %−1 (Figure 16(a) and Table A2).

Remote Sens. 2019, 11, 460 19 of 25 Remote Sens. 2019, 11, x FOR PEER REVIEW 20 of 27 Remote Sens. 2019, 11, x FOR PEER REVIEW 20 of 27

Figure 16. Multiple linear regression coefficients (β ) of each PM component with total cloud cover Figure 16. Multiple linear regression coefficients (β5) of each PM2.52.5 component with total cloud cover − − Figure 16. Multiple linear3 regression1 coefficients (β5) of each PM2.5 component with total cloud cover (cloud),(cloud), in in units units of ofµg µg m m−3% %−1..( Figurea–f) refer 16 (a), to (b), the (c), regression (d), (e) and coefficients (f) refer to of the PM regression2.5, SO4, BC, coefficients OC, Dust 2.5 (cloud), in units of µg m−3 %−1. Figure 16 (a), (b), (c), (d), (e) and (f) refer to the regression coefficients andof SSPM2.52.5with, SO4,total BC, OC, cloud Dust cover,2.5 and respectively. SS2.5 with total cloud cover, respectively. of PM2.5, SO4, BC, OC, Dust2.5 and SS2.5 with total cloud cover, respectively. 4.6. Sensitivity of PM2.5 Components to Boundary Layer Height 4.6 Sensitivity of PM2.5 Components to Boundary Layer Height 4.6 Sensitivity of PM2.5 Components to Boundary Layer Height In Figure 17, the relationship of each PM2.5 component with boundary layer height (BLH) was In Figure 17, the relationship of each PM2.5 component with boundary layer height (BLH) was In Figure 17, the relationship of each PM2.5 component with boundary layer height (BLH) was robustlyrobustly negative negative throughout throughout the the YRB. YRB. The Thereason reason may be may that be a thatwell-developed a well-developed BLH was BLH conducive was robustly negative throughout the YRB. The reason may be that a well-developed BLH was toconducive the diffusion to the and diffusion dilution and of PMdilution2.5 particles of PM2.5 [particles31]. Recently, [31]. Recently, Wang et Wang al. [31 et] reportedal. [31] reported that BLH conducive to the diffusion and dilution of PM2.5 particles [31]. Recently, Wang et al. [31] reported wasthat lower BLHthan was thelower normal than the during normal heavy during pollution heavy pollution in Beijing, in China.Beijing, DuringChina. During 1980–2017, 1980–2017, the YRB that BLH was lower than the normal during heavy pollution in Beijing, China. During 1980–2017, experiencedthe YRB experienced a significant a significant decrease decrease of −0.98 of m −0.98 peryear m per in year BLH, in especiallyBLH, especially over theover YRD, the YRD, CC andCC SB the YRB experienced a significant decrease of −0.98 m per year in BLH, especially over the YRD, CC (Figureand SB A6 (Figure). Based A6). on theBased MLR on modelthe MLR between model between BLH and BLH PM2.5 and(Figure PM2.5 (Figure17a and 17(a) Table and A2 ),Table the averageA2), and SB (Figure A6). Based−3 on−1 the MLR−3 model−1 between BLH and PM2.5 (Figure 17(a)−3 and−3 Table A2), sensitivitythe average of 0.006sensitivityµg m of 0.006m µgcould m leadm could to PM lead2.5 growth to PM2.5 of growth up to of 0.006 up toµg 0.006 m µgper m year. per year. the average sensitivity of 0.006 µg m−3 m−1 could lead to PM2.5 growth of up to 0.006 µg m−3 per year.

Figure 17. Multiple linear regression coefficients (β6) of each PM2.5 component with boundary layer Figure 17. Multiple linear regression coefﬁcients (β6) of each PM2.5 component with boundary layer Figure 17. Multiple linear regression-3 -1 coefficients (β6) of each PM2.5 component with boundary layer heightheight (BLH), (BLH), in in units units of ofµg µg m− m3 m m−1.(. Figurea–f) refer 17 (a), to the (b), regression (c), (d), (e) coefﬁcients and (f) refer of PM to the , regression SO , BC, OC, height (BLH), in units of µg m-3 m-1. Figure 17 (a), (b), (c), (d), (e) and (f) refer to the2.5 regression4 Dustcoefficientsand SS of PMwith2.5, SO boundary4, BC, OC, layer Dust height,2.5 and SS respectively.2.5 with boundary layer height, respectively. coefficients2.5 of2.5 PM2.5, SO4, BC, OC, Dust2.5 and SS2.5 with boundary layer height, respectively. 5. Conclusions

Based on the MLR model, the sensitivities of PM2.5 and its major components to meteorological variables were quantiﬁed using daily mean MERRA-2 PM2.5 and meteorology reanalysis data over the Remote Sens. 2019, 11, 460 20 of 25

YRB from 1980 to 2017. First of all, in order to verify the accuracy of the MERRA-2 PM2.5 reanalysis, it was compared with ground-based measurements derived from 476 air quality monitoring sites in the YRB. The results suggested that MERRA-2 PM2.5 was underestimated relative to their corresponding ground observations throughout the YRB. In addition to the lack of nitrate aerosols in GOCART, the bias in MERRA-2 AOD assimilation could also partly explain the underestimation of MERRA-2 PM2.5. −3 The YRB witnessed severe air pollution, with its annual mean PM2.5 concentration of 23.43 µg m . In particular, approximately 67.49% of the pixels in the YRB exceeded the annual mean threshold of −3 10 µg m recommended by the WHO. The high PM2.5 concentrations were observed over the YRD, CC and SB in winter, while the low values occurred in the SYR in summer. For each PM2.5 component, SO4, BC and OC showed similar spatiotemporal distributions with PM2.5 concentration. The temperate monsoon climate could account for the seasonal and spatial variations of PM2.5 concentrations in the −3 YRB. However, Dust2.5 had the highest concentrations of 8.63 µg m over the SYR and the northern areas of the YRB in spring, which came from local arid sources. Most of the SS2.5 were concentrated in the eastern coastal areas and there was no significant seasonal variation. Over-increasing trends (1980 onward) in all PM2.5 components were confirmed in the YRB; however, they were curbed after 2007. Based on the MLR model, we found the highest adjusted-R2 (0.67) in the SYR, suggesting that the six meteorological variables could explain up to 67% of the daily PM2.5 changes in that region. Furthermore, each PM2.5 component was significantly anti-correlated with surface wind speed and precipitation throughout the YRB. High wind speeds and precipitations facilitated the circulation and dilution of PM2.5 from the atmosphere. Nevertheless, a positive wind-PM2.5 regression was observed in the SYR, probably due to the dependence of Dust2.5 on wind. Furthermore, SYR witnessed a positive relationship between precipitation and PM2.5, which was likely associated with an increase in relative humidity (RH). Additionally, boundary layer height (BLH) was robustly anti-correlated with each −3 −1 PM2.5 component. However, the impact of BLH on PM2.5 (−0.006 µg m m ) was far less than those of wind speed (−2.42 µg m−3 m−1 s) and precipitation (−0.34 µg m−3 mm−1 day). In contrast, surface temperature was positively correlated with each PM2.5 component across the YRB, and the −3 −1 sensitivity of PM2.5 to temperature was approximately 0.11 µg m K . This was possibly due to the enhancement in oxidation of aerosol precursors at high temperatures. The RH-PM2.5 regression showed regional dependencies, with negative correlation in the YRD and positive correlation in the other areas. A negative relationship between total cloud cover and PM2.5 occurred in the middle −3 −1 and lower reaches of the YRB. However, the sensitivities of PM2.5 to RH (0.05 µg m % ) and total cloud cover (0.07 µg m−3 %−1) were relatively small. Overall, this study highlighted the impact of meteorological conditions on PM2.5, even though it was relatively minor in comparison with the anthropogenic emissions effect [27]. When assessing the effects of emission reductions and further formulating mitigation policies, the unfavorable atmospheric diffusion conditions should not be ignored in China.

Author Contributions: L.H. designed the research; H.Z., Z.Z., X.C. and L.H. performed the experiments and analyzed the data; L.H. He wrote the manuscript; A.L. and P.H. revised the manuscript. Funding: This work was financially supported by National Major Scientific Instruments and Equipment Development Project (2012YQ10022507), National Natural Science Foundation of China (No.41601044, No.41571400, No. 41704012), the Special Fund for Basic Scientific Research of Central Colleges, China University of Geosciences, Wuhan (No. CUG150631, CUGL170401, CUGCJ1704). Conflicts of Interest: The authors declare no conflict of interest. RemoteRemote Sens. Sens.2019 2019, 11,, 11 460, x FOR PEER REVIEW 22 of21 27 of 25

Remote Sens. 2019, 11, x FOR PEER REVIEW 22 of 27 Appendix A Remote Sens. 2019, 11, x FOR PEER REVIEW 22 of 27

Figure A1. Scatter plot between GWR and ground-based annual mean PM 2.5 concentrations derived from 476 air quality monitoring stations of the YRB for the period of Jan. 2015 – Dec. 2016. The black lineFigure is a A1. 1:1 Scatterline and plot the betweenred line isGWR a linear and fit ground-based line. annual mean PM 2.5 concentrations derived from 476 air quality monitoring stations of the YRB for the period of Jan. 2015 – Dec. 2016. The black FigureFigure A1. A1.Scatter Scatter plot plot between between GWR GWR and and ground-based ground-based annualannual mean PM PM2.52.5 concentrationsconcentrations derived derived fromlinefrom 476 is 476 a air 1:1 air quality line quality and monitoring themonitoring red line stations isstations a linear ofof fit thethe line. YRBYRB for for the the period period of of Jan. Jan. 2015 2015–Dec. – Dec. 2016. 2016. The The black black lineline is a is 1:1 a 1:1 line line and and the the red red line line is is a a linear linear ﬁt fit line. line.

Figure A2. The spatial (top) and frequency (bottom) distributions of multiyear-averaged surface

FigurePM2.5 A2. concentrationsThe spatial (top)derived and from frequency MERRA-2 (bottom) (a-b), distributionsGWR-1 (c-d) over of multiyear-averaged the YRB for the period surface of 1998 PM 2.5 Figure A2. The spatial (top) and frequency (bottom) distributions of multiyear-averaged surface concentrations– 2016. derived from MERRA-2 (a,b), GWR-1 (c,d) over the YRB for the period of 1998–2016. PMFigure2.5 concentrations A2. The spatial derived (top) from and MERRA-2 frequency (a-b), (bottom) GWR-1 distributions (c-d) over ofthe multiyear-averaged YRB for the period of surface 1998 – 2016. PM2.5 concentrations derived from MERRA-2 (a-b), GWR-1 (c-d) over the YRB for the period of 1998 – 2016.

Figure A3. Temporal variations of annual mean PM concentration derived from MERRA-2 (black) Figure A3. Temporal variations of annual mean PM2.52.5 concentration derived from MERRA-2 (black) andand GWR GWR (red) (red) over over the the YRB YRB from from 1998 1998 to to 2016. 2016. Figure A3. Temporal variations of annual mean PM2.5 concentration derived from MERRA-2 (black) and GWR (red) over the YRB from 1998 to 2016. Figure A3. Temporal variations of annual mean PM2.5 concentration derived from MERRA-2 (black) and GWR (red) over the YRB from 1998 to 2016.

Remote Sens. 2019, 11, x FOR PEER REVIEW 23 of 27 Remote Sens. 2019, 11, 460 22 of 25 Remote Sens. 2019, 11, x FOR PEER REVIEW 23 of 27

Figure A4. Spatial distribution of PM2.5 annual trend derived form MERRA-2 (a) and GWR (b) over the YRB from 1998 to 2016. Grids marked by grey points represented PM2.5 trends that did not FigureFigure A4. A4.Spatial Spatial distribution distribution of of PM PM2.5annual annual trend trend derived derived formform MERRA-2 (a) (a )and and GWR GWR (b) (b over) over exceed a 95% significant level. 2.5 the YRBthe YRB from from 1998 1998 to 2016. to 2016. Grids Grids marked marked by grey by points grey points represented represented PM2.5 PMtrends2.5 trends that did that not did exceed not exceed a 95% significant level. a 95%Table signiﬁcant A1. Correlation level. matrix between meteorological variables temperature (T), precipitation (Precip), total cloud cover (Cloud), wind speed (Wind), relative humidity (RH) and Boundary layer TableTable A1. Correlation A1. Correlation matrix matrix between between meteorological meteorological variables variables temperature temperature (T), precipitation (T), precipitation (Precip), height (BLH). * represents p-value < 0.05. total(Precip), cloudcover total cloud (Cloud), cover wind (Cloud), speed wind (Wind), speed relative (Wind), humidityrelative humidity (RH) and (RH) Boundary and Boundary layer layer height (BLH).height * represents (BLH). * representsp-valueT < p-value 0.05. Wind < 0.05. Precip RH Cloud BLH

T 1 T T Wind Wind Precip Precip RH RH Cloud Cloud BLH BLH Wind 0.514 1 T T 1 1

PrecipWind Wind -0.202 0.514 0.514 -0.089 1 1 1 Precip −0.202 −0.089 1 RHPrecip 0.074 -0.202 -0.334 -0.089 0.359 1 1 RH 0.074 −0.334 0.359 1 Cloud 0.274 0.26 0.519* 0.28 1 RH Cloud 0.074 0.274 -0.334 0.26 0.3590.519 * 0.28 1 1 BLHCloud -0.559BLH 0.274 −0.559 -0.543 0.26− 0.543 -0.222 0.519−0.222 * 0.351 0.351 0.28 −0.416 -0.416 1 1 1 BLH -0.559 -0.543 -0.222 0.351 -0.416 1

2 FigureFigure A5. A5. CoefﬁcientsCoefficients of determination (R (R2) ) of of the the MLR MLR model model between between PM PM2.5 and2.5 and the the six six meteorologicalmeteorological variables variables inin springspring ((a),a), summer (b), (b), autumn autumn (c) (c and) and winter winter (d). (d ). Figure A5. Coefficients of determination (R2) of the MLR model between PM2.5 and the six meteorological variables in spring (a), summer (b), autumn (c) and winter (d).

Remote Sens. 2019, 11, 460 23 of 25 Remote Sens. 2019, 11, x FOR PEER REVIEW 24 of 27

Figure A6. Annual trends in meteorological factors over the YRB from 1980 to 2017. Figure A6. Annual trends in meteorological factors over the YRB from 1980 to 2017.

Table A2. Summary of regression coefficients between surface PM2.5 and meteorological variables, Table A2. Summary of regression coefficients between surface PM2.5 and meteorological variables, annual trends in meteorological variables and theoretical annual trends in PM2.5 based on MLR model. annual trends in meteorological variables and theoretical annual trends in PM2.5 based on MLR β1 ... β6 refer to regression coefficients between PM2.5 and wind speed (Wind), temperature (T), model. β1 … β6 refer to regression coefficients between PM2.5 and wind speed (Wind), temperature precipitation (Precip), relative humidity (RH), total cloud cover (Cloud) and Boundary layer height (T), precipitation (Precip), relative humidity (RH), total cloud cover (Cloud) and Boundary layer (BLH), respectively. height (BLH), respectively. Regression Coefficients Annual trends in Meteorological Variables Theoretical Annual Trends in PM Theoretical 2.5 −3 −1 Annual−1 trends in −3 β1 (−2.42 µg m m s) Wind (0.05 m s ) 0.12 µg m −3 −1 annual −3 β2 (0.11 µg m KRegression) coefficients T (0.03 K)Meteorological 0.003 µg m −3 −1 −1 trends in −3 β3 (−0.34 µg m mm day) Precip (0.02 mm dayvariables) 0.007 µg m −3 −1 2.5 −3 β4 (0.05 µg m % ) RH (−0.16%) PM−0.008 µg m −3 −1 −3 β5 (−0.07 µg m %β1 (-2.42) µg m-3 m-1 s) Cloud (−0.05%)Wind (0.05 m s-1) 0.12 0.0004µg m-3µ g m β (0.006 µg m−3 m−1) BLH (−0.98 m) 0.006 µg m−3 6 β2 (0.11 µg m-3 K-1) T (0.03 K) 0.003 µg m-3

β3 (-0.34 µg m-3 mm-1 day) Precip (0.02 mm day-1) 0.007 µg m-3 References β4 (0.05 µg m-3 %-1) RH (-0.16 %) -0.008 µg m-3 β5 (-0.07 µg m-3 %-1) Cloud (-0.05 %) 0.0004 µg m-3 1. IPCC. Climate change 2013: the physical science basis. In Contribution of Working Group I to the Fifth β6 (0.006 µg m-3 m-1) BLH (-0.98 m) 0.006 µg m-3 Assessment Report of the Intergovernmental Panel on Climate Change; Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex, V., Midgley, P.M., Eds.; Cambridge University References Press: Cambridge, UK; New York, NY, USA, 2013; p. 1535. 2. 1. Guo,IPCC. J.; Xia, Climate F.; Zhang, change Y.; 2013: Liu, H.; the Li, physical J.; Lou, science M.; Zhai, basis. P. Impact In Contribution of diurnal of Workingvariability Group and I meteorological to the Fifth factorsAssessment on the Report PM2.5 -AOD of the Intergovernmental relationship: Implications Panel on Climate for PM Change2.5 remote Stocker sensing.; T.F., Qin,Environ. D.,Plattner, Pollut. 2017 G.-K.,, 221 , 94–104.Tignor, [CrossRef M., Allen,][PubMed S.K., Boschung,] J., Nauels, A., Xia, Y., Bex, V.,Midgley, P.M., Eds.; Cambridge 3. He,University J.; Gong, Press: S.; Yu, Cambridge, Y.; Yu, L.; United Wu, L.; Kingdom Mao, H.; and Li, New R. AirYork, pollution NY, USA, characteristics 2013, pp. 1535. and their relation to 2. meteorologicalGuo, J.; Xia, F.; conditions Zhang, Y.; duringLiu, H.; 2014–2015Li, J.; Lou, M.; in majorZhai, P. Chinese Impact cities.of diurnalEnviron. variability Pollut. and2017 meteorological, 223, 484–496. [CrossReffactors ][onPubMed the PM ]2.5-AOD relationship: Implications for PM 2.5 remote sensing. Environ. Pollut. 2017, 221, 4. Zhang,94–104. M.; Wang, L.; Gong, W.; Ma, Y.; Liu, B. Aerosol optical properties and direct radiative effects over

3. CentralHe, J.; China. Gong, RemoteS.; Yu, Sens.Y.; Yu,2017 L.; ,Wu,9, 997. L.; [Mao,CrossRef H.; Li,] R. Air pollution characteristics and their relation to meteorological conditions during 2014–2015 in major Chinese cities. Environ. Pollut. 2017, 223, 484–496. 5. Zhang, M.; Ma, Y.; Wang, L.; Gong, W.; Hu, B.; Shi, Y. Spatial-temporal characteristics of aerosol loading over 4. Zhang, M.; Wang, L.; Gong, W.; Ma, Y.; Liu, B. Aerosol optical properties and direct radiative effects over the Yangtze River Basin during 2001–2015. Int. J. Climatol. 2018, 38, 2138–2152. [CrossRef] Central China. Remote Sens. 2017, 9, 997. https://doi.org/10.3390/rs9100997. 6. Qin, W.; Liu, Y.; Wang, L.; Lin, A.; Xia, X.; Che, H.; Zhang, M. Characteristic and driving factors of aerosol 5. Zhang, M.; Ma, Y.; Wang, L.; Gong, W.; Hu, B.; Shi, Y. Spatial-temporal characteristics of aerosol loading optical depth over mainland China during 1980–2017. Remote Sens. 2018, 10, 1064. [CrossRef] over the Yangtze River Basin during 2001–2015. Int. J. Climatol. 2018, 38, 2138–2152. 7. 6. Lelieveld,Qin, W.; J.;Liu, Evans, Y.; Wang, J.S.; L.; Fnais, Lin, M.;A.; Xia, Giannadaki, X.; Che, H.; D.; Zhang, Pozzer, M. A.Characteristic The contribution and driving of outdoor factors of air aerosol pollution sourcesoptical to premature depth over mortality mainland on a global China scale. duringNature 1980–2017.2015, 525, 367–371.Remote [CrossRef Sens. 2018][PubMed, 10, ] 1064. 8. He,https://doi.org/10.3390/rs10071064. L.; Wang, L.; Lin, A.; Zhang, M.; Bilal, M.; Tao, M. Aerosol Optical Properties and Associated Direct Radiative Forcing over the Yangtze River Basin during 2001–2015. Remote Sens. 2017, 9, 746. [CrossRef] Remote Sens. 2019, 11, 460 24 of 25

9. He, L.; Wang, L.; Lin, A.; Zhang, M.; Bilal, M.; Wei, J. Performance of the NPP–VIIRS and aqua–MODIS Aerosol Optical Depth Products over the Yangtze River Basin. Remote Sens. 2018, 10, 117. [CrossRef] 10. He, L.; Wang, L.; Lin, A.; Zhang, M.; Xia, X.; Tao, M.; Zhou, H. What drives changes in aerosol properties over the Yangtze River Basin in past four decades? Atmos. Environ. 2018, 190, 269–283. [CrossRef] 11. Wang, L.; Gong, W.; Xia, X.; Zhu, J.; Li, J.; Zhu, Z. Long–term observations of aerosol optical properties at Wuhan, an urban site in Central China. Atmos. Environ. 2015, 101, 94–102. [CrossRef] 12. Che, H.; Xia, X.; Goloub, P.; Holben, B.; Zhao, H.; Wang, Y.; Zhang, X.; Wang, H.; Blarel, L.; Damiri, B. Ground-based aerosol climatology of China: aerosol optical depths from the China Aerosol Remote Sensing Network (CARSNET) 2002-2013. Atmos. Chem. Phys. 2015, 15, 7619–7652. [CrossRef] 13. Wei, J.; Sun, L. Comparison and evaluation of different modis aerosol optical depth products over the beijing-tianjin-hebei region in china. IEEE J-STARS 2016, PP(99), 1–10. [CrossRef] 14. Wang, L.; Lu, Y.; Zou, L.; Feng, L.; Wei, J.; Qin, W.; Niu, Z. Prediction of diffuse solar radiation based on multiple variables in China. Renew. Sust. Energ. Rev. 2019, 103, 151–216. [CrossRef] 15. Feng, L.; Lin, A.; Wang, L.; Qin, W.; Gong, W. Evaluation of sunshine-based models for predicting diffuse solar radiation in China. Renew. Sust. Energ. Rev. 2018, 94, 168–182. [CrossRef] 16. Feng, L.; Qin, W.; Wang, L.; Lin, A.; Zhang, M. Comparison of Artiﬁcial Intelligence and Physical Models for Forecasting Photosynthetically-Active Radiation. Remote Sens. 2018, 10, 1855. [CrossRef] 17. Wang, L.; Chen, Y.; Niu, Y.; Salazar, G.A.; Gong, W. Analysis of atmospheric turbidity in clear skies at Wuhan, Central China. J. Earth Sci. 2017, 28, 729–738. [CrossRef] 18. Wang, L.; Kisi, O.; Zounemat-Kermani, M.; Zhu, Z.; Gong, W.; Niu, Z.; Liu, Z. Prediction of solar radiation in China using different adaptive neuro-fuzzy methods and M5 model tree. Int. J. Climatol. 2017, 37, 1141–1155. [CrossRef] 19. Wang, L.; Kisi, O.; Zounemat-Kermani, M.; Salazar, G.A.; Zhu, Z.; Gong, W. Solar radiation prediction using different techniques: model evaluation and comparison. Renew. Sust. Energ. Rev. 2016, 61, 384–397. [CrossRef] 20. Zhang, M.; Ma, Y.; Gong, W.; Wang, L.; Xia, X.; Che, H.; Hu, B.; Liu, B. Aerosol radiative effect in UV, VIS, NIR, and SW spectra under haze and high-humidity urban conditions. Atmos. Environ. 2017, 166, 9–21. [CrossRef] 21. Guo, J.; He, J.; Liu, H.; Miao, Y.; Liu, H.; Zhai, P. Impact of various emission control schemes on air quality using WRF-Chem during APEC China 2014. Atmos. Environ. 2016, 140, 311–319. [CrossRef] 22. Ren, Y.; Li, H.; Meng, F.; Wang, G.; Zhang, H.; Yang, T.; Li, W.; Ji, Y.; Bi, F.; Wang, X. Impact of emission controls on air quality in Beijing during the 2015 China Victory Day Parade: Implication from organic aerosols. Atmos. Environ. 2019, 198, 207–214. [CrossRef] 23. Larkin, A.; Van Donkelaar, A.; Geddes, J.A.; Martin, R.V.; Hystad, P. Relationships between changes in urban characteristics and air quality in East Asia from 2000 to 2010. Environ. Sci. Technol. 2016, 50, 9142–9149. [CrossRef][PubMed] 24. Li, T.; Wang, Y.; Zhao, D. Environmental Kuznets curve in China: new evidence from dynamic panel analysis. Energy Policy 2016, 91, 138–147. [CrossRef] 25. Liu, Y.; Zhou, Y.; Wu, W. Assessing the impact of population, income and technology on energy consumption and industrial pollutant emissions in China. Appl. Energy 2015, 155, 904–917. [CrossRef]

26. Tai, A.P.; Mickley, L.J.; Jacob, D.J. Correlations between ﬁne particulate matter (PM2.5) and meteorological variables in the United States: Implications for the sensitivity of PM2.5 to climate change. Atmos. Environ. 2010, 44, 3976–3984. [CrossRef] 27. Westervelt, D.M.; Horowitz, L.W.; Naik, V.; Tai, A.P.K.; Fiore, A.M.; Mauzerall, D.L. Quantifying

PM2.5-meteorology sensitivities in a global climate model. Atmos. Environ. 2016, 142, 43–56. [CrossRef] 28. Harrison, R.M.; Laxen, D.; Moorcroft, S.; Laxen, K. Processes affecting concentrations of ﬁne particulate

matter (PM2.5) in the UK atmosphere. Atmos. Environ. 2012, 46, 115–124. [CrossRef] 29. Kassomenos, P.A.; Vardoulakis, S.; Chaloulakou, A.; Paschalidou, A.K.; Grivas, G.; Borge, R.; Lumbreras, J.

Study of PM10 and PM2.5 levels in three European cities: Analysis of intra and inter urban variations. Atmos. Environ. 2014, 87, 153–163. [CrossRef]

30. Munir, S.; Habeebullah, T.M.; Mohammed, A.M.; Morsy, E.A.; Rehan, M.; Ali, K. Analysing PM2.5 and its association with PM10 and meteorology in the arid climate of Makkah, Saudi Arabia. Aerosol Air Qual. Res. 2017, 17, 453–464. [CrossRef] Remote Sens. 2019, 11, 460 25 of 25

31. Wang, X.; Dickinson, R.E.; Su, L.; Zhou, C.; Wang, K. PM2.5 pollution in China and how it has been exacerbated by terrain and meteorological conditions. Bull. Am. Meteoro. Soc. 2018, 99, 105–119. [CrossRef] 32. Chen, Y.; Schleicher, N.; Fricker, M.; Cen, K.; Liu, X.-L.; Kaminski, U.; Yu, Y.; Wu, X.-F.; Norra, S. Long-term

variation of black carbon and PM2.5 in Beijing, China with respect to meteorological conditions and governmental measures. Environ. Pollut. 2016, 212, 269–278. [CrossRef][PubMed]

33. Yang, Q.; Yuan, Q.; Li, T.; Shen, H.; Zhang, L. The relationships between PM2.5 and meteorological factors in China: Seasonal and regional variations. Int. J. Environ. Res. Public Health 2017, 14, 1510.

34. Lin, Y.; Zou, J.; Yang, W.; Li, C.Q. A review of recent advances in research on PM2.5 in China. Int. J. Environ. Res. Public Health 2018, 15, 438. [CrossRef][PubMed] 35. Li, L.; Qian, J.; Ou, C.Q.; Zhou, Y.X.; Guo, C.; Guo, Y. Spatial and temporal analysis of Air Pollution Index and its timescale-dependent relationship with meteorological factors in Guangzhou, China, 2001–2011. Environ. Pollut. 2014, 190, 75–81. [CrossRef][PubMed]

36. Zhang, T.; Zhu, Z.; Gong, W.; Zhu, Z.; Sun, K.; Wang, L.; Xu, K. Estimation of ultrahigh resolution PM2.5 concentrations in urban areas using 160 m Gaofen-1 AOD retrievals. Remote Sens. Environ. 2018, 216, 91–104. [CrossRef] 37. Buchard, V.; da Silva, A.M.; Randles, C.A.; Colarco, P.; Ferrare, R.; Hair, J.; Winker, D. Evaluation of the

surface PM2.5 in Version 1 of the NASA MERRA Aerosol Reanalysis over the United States. Atmos. Environ. 2016, 125, 100–111. [CrossRef] 38. Randles, C.A.; Da Silva, A.M.; Buchard, V.; Colarco, P.R.; Darmenov, A.; Govindaraju, R.; Shinozuka, Y. The MERRA-2 aerosol reanalysis, 1980 onward. Part I: System description and data assimilation evaluation. J. Climate 2017, 30, 6823–6850. [CrossRef][PubMed] 39. Buchard, V.; Randles, C.A.; da Silva, A.M.; Darmenov, A.; Colarco, P.R.; Govindaraju, R.; Yu, H. The MERRA-2 aerosol reanalysis, 1980 onward. Part II: Evaluation and case studies. J. Climate 2017, 30, 6851–6872. [CrossRef]

40. Song, Z.; Fu, D.; Zhang, X.; Wu, Y.; Xia, X.; He, J.; Che, H. Diurnal and seasonal variability of PM2.5 and AOD in North China plain: Comparison of MERRA-2 products and ground measurements. Atmos. Environ. 2018, 191, 70–78. [CrossRef] 41. Van Donkelaar, A.; Martin, R.V.; Spurr, R.J.D.; Drury, E.; Remer, L.A.; Levy, R.C.; Wang, J. Optimal estimation for global ground-level ﬁne particulate matter concentrations. J. Geophys. Res. Atmos. 2013, 118, 5621–5636. [CrossRef] 42. Van Donkelaar, A.; Martin, R.V.; Brauer, M.; Kahn, R.; Levy, R.; Verduzco, C.; Villeneuve, P.J. Global estimates of ambient ﬁne particulate matter concentrations from satellite-based aerosol optical depth: development and application. Environ. Health Perspect. 2010, 118, 847–855. [CrossRef][PubMed] 43. Huang, J.; Kondragunta, S.; Laszlo, I.; Liu, H.; Remer, L.A.; Zhang, H.; Petrenko, M. Validation and expected error estimation of Suomi-NPP VIIRS aerosol optical thickness and Ångström exponent with AERONET. J. Geophys. Res. Atmos. 2016, 121, 7139–7160. [CrossRef] 44. Peng, S.-S.; Piao, S.; Zeng, Z.; Ciais, P.; Zhou, L.; Li, L.Z.; Zeng, H. Afforestation in China cools local land surface temperature. Proc. Natl. Acad. Sci. USA 2014, 121, 2915–2919. [CrossRef][PubMed]

45. Dawson, J.P.; Adams, P.J.; Pandis, S.N. Sensitivity of PM2.5 to climate in the Eastern US: a modeling case study. Atmos. Chem. Phys. 2007, 7, 4295–4309. [CrossRef] 46. Zhou, H.; Luo, Z.; Zhou, Z.; Li, Q.; Zhong, B.; Lu, B.; Hsu, H. Impact of different kinematic empirical parameters processing strategies on temporal gravity ﬁeld model determination. J. Geophys. Res. Solid Earth 2018.[CrossRef] 47. Zhou, H.; Luo, Z.; Tangdamrongsub, N.; Zhou, Z.; He, L.; Xu, C.; Li, Q.; Wu, Y. Identifying Flood Events over the Poyang Lake Basin Using Multiple Satellite Remote Sensing Observations, Hydrological Models and In Situ Data. Remote Sens. 2018, 10, 713. [CrossRef] 48. Zhou, H.; Luo, Z.; Tangdamrongsub, N.; Wang, L.; He, L.; Xu, C.; Li, Q. Characterizing drought and ﬂood events over the Yangtze River Basin using the HUST-Grace2016 solution and ancillary data. Remote Sens. 2017, 9, 1100. [CrossRef]

© 2019 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 (http://creativecommons.org/licenses/by/4.0/).