Estimation of regional PM2.5 concentration based on EBK-GWR
In view of the limited number and uneven distribution of existing air monitoring stations in our country,it is difficult to obtain continuous regional PM2.5 data.On the basis of pressure(PRS),temperature(TEM),rela-tive humidity(RHU)obtained from meteorological stations in Guangdong and Guangxi,and precipitable water va-por(PWV)data obtained from subordinate and adjacent radiosonde stations,the TEM,PRS,RHU and PWV da-ta of different seasons in 2019 are acquired by inverse distance weighted interpolation.By comprehensively analy-zing their correlation with PM2.5,the geographically weighted regression(GWR)and empirical Bayesian Kriging geographically weighted regression(EBK-GWR)are constructed to estimate the spatial distribution of PM2.5 con-centrations.The results show that EBK-GWR model's PM2.5 concentration estimation effect is significantly better than GWR model in different seasons,where the root mean square error(RMSE)and mean absolute error(MAE)improves most significantly in spring and autumn.Compared with GWR model,the estimation accuracy of RMSE and MAE improves 16.67％and 13.88％in spring,14.13％and 13.04％in autumn,respectively.The improve-ment effect in summer and winter is lower than in spring and autumn,but the improvement remains at about 10％compared with GWR model,and the RMSE and MAE estimated in different seasons are both less than 4 μg/m3.
万方数据
