MULTIPLE REGRESSION METHOD FOR ESTIMATING CONCENTRATION OF PM2.5 USING REMOTE SENSING AND METEOROLOGICAL DATA

摘要

Retrieval of Aerosol Optical Thickness (AOT) of Wuhan city by MODIS L1B data, and the humidity correction and vertical correction of AOT were carried out. Based on processing of meteorological data including relative humidity (RH), surface temperature (ST), wind speed (WS), pressure (PRE), and introducing meteorological data to the Relational Model of AOT-PM2.5, were established multiple linear and nonlinear regression models for estimating concentration of PM2.5 in Wuhan city. The models were compared and analysed to choose optimum modelling method and important influence factors. The results showed that the annual correlation coefficient of multiple linear and nonlinear regression models were 0.513 and 0.607, respectively. Multiple nonlinear regression model has the advantage over multiple linear model. The values of influence degree of meteorological data in two models were 21.7 and 13.2%, which shows that the meteorological factor has important influence on AOT-PM2.5 relationship; the prediction effect of two models in spring is similar to in spring, but the multiple nonlinear regression models have obvious advantage in summer and autumn.