Spatial Factor Analysis for Aerosol Optical Depth in Metropolises in China with Regard to Spatial Heterogeneity

Abstract

A substantial number of studies have analyzed how driving factors impact aerosols, but they have been little concerned with the spatial heterogeneity of aerosols and the factors that impact aerosols. The spatial distributions of the aerosol optical depth (AOD) retrieved by Moderate Resolution Imaging Spectrometer (MODIS) data at 550-nm and 3-km resolution for three highly developed metropolises, the Beijing-Tianjin-Hebei (BTH) region, the Yangtze River Delta (YRD), and the Pearl River Delta (PRD), in China during 2015 were analyzed. Different degrees of spatial heterogeneity of the AOD were found, which were indexed by Moran's I index giving values of 0.940, 0.715, and 0.680 in BTH, YRD, and PRD, respectively. For the spatial heterogeneity, geographically weighted regression (GWR) was employed to carry out a spatial factor analysis, where terrain, climate condition, urban development, and vegetation coverage were taken as the potential driving factors. The results of the GWR imply varying relationships between the AOD and the factors. The results were generally consistent with existing studies, but the results suggest the following: (1) Elevation increase would more likely lead to a strong negative impact on aerosols (with most of the coefficients ranging from −1.5~0 in the BTH, −1.5~0 in the YRD, and −1~0 in the PRD) in the places with greater elevations where the R-squared values are always larger than 0.5. However, the variation of elevations cannot explain the variation of aerosols in the places with relatively low elevations (with R-squared values approximately 0.1, ranging from 0 to 0.3, and approximately 0.1 in the BTH, YRD, and PRD), such as urban areas in the BTH and YRD. (2) The density of the built-up areas made a strong and positive impact on aerosols in the urban areas of the BTH (R-squared larger than 0.5), while the R-squared dropped to 0.1 in the places far away from the urban areas. (3) The vegetation coverage led to a stronger relief on the AOD in parts of the YRD and PRD (with coefficients less than −0.6 and ranging from −0.4~−0.6, respectively) where there is greater vegetation coverage, and led to a weaker relief on the AOD in the urban area of the PRD with a coefficient of approximately −0.2~−0.4.Atmosphere 2018, 9, 156 2 of 14 images [2,5] and their variations can now be fully studied [6]. Moreover, a factor analysis for aerosols has been widely carried out. The Ordinary Least Square (OLS) based regression is usually used to identify the relationships between aerosols and their impact factors [7][8][9].However, OLS assumes that the relationship between the dependent and independent variables is consistent and cannot handle the spatial data with heterogeneity, which makes it difficult to meet the assumptions and requirements of OLS [10]. Heterogeneity here includes two aspects: spatial autocorrelation and spatially non-stationarity. Spatial autocorrelation means that the value of a variable at a location depends on the value of the same variable at nearby locations [11,12]. Spatial non-stati...

Spatiotemporal Variability and Influencing Factors of Aerosol Optical Depth over the Pan Yangtze River Delta during the 2014–2017 Period

“…The close and negative association between DEM, SLP and AOD may be explained from three aspects: firstly, low-elevation and flat areas are more influenced by human activity such as industry and construction, and thus emitted more air pollutants [34,50]; secondly, high mountains in high-elevation areas can prevent the horizontal dispersion of air pollutants [19,96]; and lastly, f or the mid-latitude areas, precipitation usually increases with elevation [97], while precipitation is capable of bringing down aerosols [46]. Previous studies also showed the aerosol distribution is strongly affected by topography conditions [19,41,42,43].…”

“…The normalized difference vegetation index (NDVI) data from 2014–2017, representing the vegetation coverage of the study area, were acquired from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC) [84]. RESDC provides seasonal NDVI values and annual NDVI values at 1-km resolution [41]. From this website, we also obtained the 1-km resolution annual gross domestic product (GDP) and population density (POP) data of the study area in 2015 to reflect the anthropogenic emissions of pollutants from 2014 to 2017.…”

“…The seasonally and annually averaged AOD were derived by averaging the daily merged AOD. The monthly meteorological data (i.e., PREC, AWS, ATEM, ARH and PBLH) and NDVI data (both seasonal and annual data) [41] were converted to four-year seasonal and annual average data. After that step, the meteorological data (i.e., PREC, AWS, ATEM, ARH) at each site were interpolated to 3-km continuous raster data by the inverse distance weight interpolation method [38].…”

“…Variation in the spatial distribution of AOD results from a combination of multiple factors, such as topography, meteorology, vegetation, and socioeconomic factors [6,19,40,41,42]. Topography has been reported to be correlated negatively with the spatial pattern of AOD in several studies because of its strong relation with aerosol emissions and particle accumulation [19,41,42,43].…”

“…Variation in the spatial distribution of AOD results from a combination of multiple factors, such as topography, meteorology, vegetation, and socioeconomic factors [6,19,40,41,42]. Topography has been reported to be correlated negatively with the spatial pattern of AOD in several studies because of its strong relation with aerosol emissions and particle accumulation [19,41,42,43]. Meteorological variables, such as precipitation [19,44,45,46], wind speed [45,47,48], temperature [46,49,50,51], relative humidity [45,52] and planetary boundary layer height (PBLH) [53,54] play important roles in the diffusion, dilution, and accumulation of aerosol particles.…”