LUR Models For Simulating The Spatial Distribution Of PM_2.5 Concentration In Zhejiang Province

inhaled PM_2.5 can cause serious impact on human health,The knowledge of spatial distribution of atmospheric particulates may be still the academic focus and problems to be solved for a long time.Although traditional method of site monitoring can accurately detect the local PM_2.5 mass concentration,it is unable to forecast the quality of PM_2.5 concentrations in a wide range of spatial distribution.Using of land use regression model to simulate the distribution of PM_2.5 air pollution concentration is increasingly mature.Land use regression?LUR?model was employed to simulate the spatial distribution of PM_2.5 in zhejiang province based on environmental air quality monitoring data.20 predictors of influence PM_2.5 mass concentration including meteorological factor,land use factor,topography factor and other factors was selected.the results show that R2 avarage value?0.69?and R2 Adjusted avarage value?0.53?for GWR models were superior than R2 avarage value?0.53?and R2 Adjusted avarage value?0.41?for OLS models based on the 31 sites in zhejiang province,but there were no significant differences between both AIC index.while the R2 avarage value and R2 Adjusted avarage value for GWR models were superior to using OLS models fitting equation of R2 avarage value and R2 Adjusted avarage value based on the 10 sites in area of hangzhou,and AIC value of GWR models was obviously higher than that AIC value of OLS models.The result indicate that using LUR model to simulate the large-scale area of near surface PM_2.5 concentrations is effective.Forecasting method based on GWR is better than OLS.The method provided in this paper has a certain reference value for the further research of PM_2.5 estimation model.Secondly,Land use regression?LUR?models were used to estimate the spatial distribution of paticulate matter(PM_2.5)in Hangzhou City.In total,more than 60 variables for the land use regression models were generated to characterize the road network,land use,meteorology and other factors.The geographical weighted regression algorithm?GWR?was used to build PM_2.5 LUR models for the different seasons.The adjusted R2 values for the PM_2.5 LUR models for Spring,Summer and Autumn were 0.76,0.70,0.73 and 0.76 respectively,The LUR models explained more than 80% of the spatial variability for PM_2.5.The spatial pattern of PM_2.5 changed with the seasons,The high concentrations of PM_2.5 were more dispersed in the central areas of hangzhou,and a clear ares of low concentrations was evident in the southwest of the study regions.The approach of modeling the spatial distribution of PM_2.5 using LUR models has potential usefulness for exposure assessment in health studies.Finallly,Using remote sensing data to retrieve AOD data from MODIS based on 152 sites in Zhejiang Province,The meteorological factors of the air monitoring stations were obtained by the inverse distance interpolation,and the stepwise multiple linear regression model and the BP artificial neural network were established.The results shows: Using of BP artificial neural network to establish the fitting equation of the R2 value and R2 Adjusted value are better than the use of stepwise multiple linear regression method,the R2 value was 0.56,the RMSE value was 16.15,MAE value was 12.88 by using BP artificial neural network.the R2 value was 0.44,the RMSE value was 20.48,MAE value was 14.6 by using stepwise multiple linear regression.The application of artificial neural network model can better realize the the expression of nonlinear implicit spatial distribution of near surface PM_2.5.