Abstract: To accurately predict the aerosol optical depth (AOD) of typical urban agglomerations in eastern China, based on MODIS data from 2010 to 2019, the spatial and temporal differences of AOD between Beijing-TianjinHebei region, Yangtze River Delta and Pearl River Delta and within them were analyzed. The AOD prediction model based on the combination of wavelet transform and BP neural network was built to predict AOD in typical urban agglomerations. The results show that: 1) the peak value of aerosol concentration in all urban agglomerations occurs in summer, and the average AOD of Beijing-Tianjin-Hebei region is the highest, followed by Yangtze River Delta and Pearl River Delta. 2) the analysis of AOD influencing factors shows that GDP index, population density and temperature are positively correlated with AOD, while normalized difference vegetation index (NDVI), precipitation and wind speed are negatively correlated with AOD. 3) the mean absolute error (MAE) of AOD prediction results in each region is lower than 0.12, error is less than BP neural network and R2 is greater than 0.75, indicating that the model can effectively improve AOD prediction ability compared with BP neural network.

Key words: typical urban agglomeration, aerosol optical depth, analysis on spatial and temporal evolution; aerosol optical depth prediction