Influencing Factors of Long-term Variations on Gridded PM2.5 of Typical Regions in China Based on GAM Model

Abstract

This study investigates annually-averaged surface PM2.5 concentrations with the spatial resolution of 0.01°×0.01° to explore spatio-temporal variations and influencing factors of annual PM2.5 over typical regions in China during the period of 1998-2016, applying the generalized additive model (GAM). Regionally-averaged PM2.5 concentrations of five typical regions are ranked from high to low as follows:East China (40.5 μg·m-3) > North China (37.4 μg·m-3) > South China (27.8 μg·m-3) > Northeast China (23.7 μg·m-3) > Sichuan Basin (22.4 μg·m-3). The PM2.5 over Northeast China showed a linear increasing trend, while in other regions, PM2.5 tended to increase from 1998 to 2007 and decrease after 2007. PM2.5 concentrations over typical regions were all stably distributed which clearly exhibited areas with high PM2.5 values. For the single influencing factor GAM model of PM2.5 concentration, all influencing factors passed the significance test. The most influential factors with regard to the variations in the PM2.5 concentration differed among typical regions. In the multiple-influencing-factors-GAM model of PM2.5 concentration, all factors exhibited a non-linear relationship with PM2.5, and they accounted for 87.5%-92% (average 89.0%) of variations in the PM2.5 concentration, suggesting a good model fit. The most significant influencing factors on PM2.5 concentrations were YEAR and LON-LAT in all typical regions. Meteorological factors have different impacts on PM2.5 concentrations among the typical regions. The three most influential meteorological factors in the five typical regions ranked from high to low are as follows:tp > v10 > ssr for Northeast China; temp > tp > msl for North China; temp > tp > ssr for East and Central China; temp > RH > blh for South China; tp > temp > u10 for the Sichuan Basin. Our results demonstrated that the GAM model could quantitatively analyze influencing factors in long-term variations of the regional PM2.5 concentration, which is important for the assessment of PM2.5 pollution.