New MAIAC AOD Product Based High Resolution PM2.5 Spatial-Temporal Distribution Change at Urban Scale –––– Case Study of Wuhan

In China, PM2.5 pollution has been an urgent problem affecting human health for urban areas. However, due to limited observation data, long period change of PM2.5 distribution at urban scale has not received much attention. In this study, we use a machine learning method, Gradient Boost Decision Tree (GBDT) with powerful predictive capacity, to derive and to analyze the winter PM2.5 spatial-temporal distribution change of Wuhan between two periods (December, 2005 to February, 2008 and December, 2015 to February, 2018). The new 1km Multi-Angle Implementation of Atmospheric Correction (MAIAC) AOD product is used as the primary predictor which can ensure PM2.5 distribution with high resolution at urban scale. It is found that PM2.5 pollution in winter of Wuhan is serious with averaged concentration value being more than 75 μg/m ³ in both periods. In addition, PM2.5 concentration shows a decreased trend in the central areas of Wuhan but an increased trend in some suburb areas because of urban expansion and construction.