Modeling temporal and spatial variation in atmospheric aerosols by geographically and temporally weighted regression with principal component analysis

摘要

Changes in the distribution of aerosol optical depth (AOD) are closely related to climate change, air quality, environmental pollution and human health. However, null values often appear in regions of AOD data retrieved by satellite. To address this problem, we propose geographically and temporally weighted regression with principal component analysis (PCA-GTWR), which aims to make full use of the advantages of geographically and temporally weighted regression (GTWR) and principal component analysis (PCA). Taking the prediction of the AOD in Beijing as an example, the PCA-GTWR model predicted that the monthly average AOD data would have an MAE, RMSE, R~2, R_j~2 and regression coefficient of 0.0705, 0.0954, 0.8705, 0.8703, and 0.7913, respectively, in April 2015; 0.0587, 0.0757, 0.8628, 0.8627, and 0.7939, respectively, in May 2015; and 0.1059, 0.1376, 0.8185, 0.8184 and 0.7633, respectively, in June 2015. This result shows that the PCA-GTWR model can be effectively applied to AOD data prediction. The research content of this paper is of great significance to research on climate change, air quality and environmental pollution.