Retrieving PM2.5 Using Domestic GF-1 Satellite WFV Data

In recent years,a number of large-scale haze events nationwide have aroused the attention of the community to PM2.5.How to monitor PM2.5 has been highly valued by various governments and relevant departments.Taking Taiwan,Beijing and Fujian provinces as the research areas,this paper comprehensively uses the Gao Fen-1 satellite(GF-1)WFV remote sensing data and meteorological auxiliary data to research on the inversion method of regional PM2.5 mass concentration.The main research contents and results are as follows:(1)A cloud detection method for GF-1 satellite WFV data band characteristics is proposed.Cloud has great interference on satellite remote sensing retrieving aerosol.It needs to be removed from remote sensing images to ensure inversion accuracy.In view of the characteristics of the homemade GF-1 star 16 meter wide range multispectral remote sensing image,which only contains the visible near infrared band,a cloud detection method based on band operation and spatial texture features is proposed.This method can well detect cloud pixels over different phases and different underlying surfaces,which can meet the application needs of subsequent aerosol inversion.(2)An aerosol optical thickness(AOD)inversion method that can take into account the surface reflectance of high and low reflectance is provided.Integrating the advantages of dark pixel and deep blue algorithm,it provides a way to retrieve AOD over both low-reflectivity and high-reflectivity surface cover types.Based on this method,the AOD distribution at 550nm is retrieved from GF-1 star WFV data.Compared with MODIS aerosol products(MOD04),the spatial distributions of AOD from GF-1 are well consistent with those from MOD04(r>0.9);there is a significant correlation with the measured values of AERONET(r>0.85),and the 70%estimated results meet the accuracy requirement.Compared with single algorithm,this proposed method has distinctive advantage in terms of estimation accuracy and spatial coverage.(3)The remote sensing estimation model of PM2.5 mass concentration was constructed.The remote sensing estimation of PM2.5 based on direct relation model shows that there is a significant correlation between PM2.5 and AOD;the seasonal model has higher accuracy than the annual model;the optimal model between PM2.5 and AOD has regional and seasonal differences.The remote sensing estimation of PM2.5 based on stepwise regression model,the results show that AOD has the most obvious effect on the model results,followed by wind speed,temperature,planetary boundary layer height,solar radiation and relative humidity and other factors have little effect on the estimation results;inversion accuracy of model is generally not high.The remote sensing estimation of PM2.5 based on weighted regression model,the results show that relative to stepwise regression methods,modeling with a geographically weighted approach may yield higher accuracy;first,the stepwise regression method is used to select the variables,and then the geo weighted regression is used for modeling.This is advantageous for constructing an efficient and accurate PM2.5 inversion model;the inversion result is in good agreement with AOD and interpolation results,and the inversion result is more reliable.