Research On The Estimation And Spatiotemporal Distribution Characteristics Of Near-surface Ozone Based On Multi-source Dat

With the rapid development of industrialization and urbanization in our country,pollution gas emission increases substantially,the problem of air pollution is becoming more and more serious.It seriously endangers human health.With the launch of various remote sensing satellites,they have provided massive data for the monitoring of atmospheric composition.Therefore,Sichuan Province was selected as the research area in this paper,and multi-source data including nitrogen dioxide（NO₂）data provided by Sentine-5p satellite,ERA5 meteorological reanalysis data,DEM data,land use data（LUCC）data and vegetation index（NDVI）data were selected as the estimated variables.Meanwhile,the multi-source data and ozone concentration data provided by environmental monitoring stations were matched,trained and fitted for verification.Therefore,the ozone concentration in the whole Sichuan Province was estimated,and the variation characteristics of ozone concentration were analyzed on the monthly,seasonal,annual and spatial scales.The results show that:（i）In this study,the daily average data of Sichuan Province from March 2019 to February 2022 were used.The matched total data set is input into the GTWR model,GWR-LMM model,CNN model,RNN model,GTWR+CNN model,GTWR+RNN model,GWR-LMM+RNN model,GTWR+IRF model and GWR-LMM+IRF model for training and fitting comparison.In order to determine the highest accuracy of the estimation model.（ii）Compared with other models,GWR-LMM+IRF model has the best fitting effect.R²,RMSE and MAE fitted by GWR-LMM+IRF model are 0.98,5.33μg·m^-3and 3.74μg·m^-3,respectively.R²,RMSE and MAE of cross validation were 0.92,11.1μg·m^-3and 8.16μg·m^-3,respectively.The results show that this model can obtain the spatial and temporal distribution characteristics of ozone in Sichuan Province with high accuracy,and provide a reasonable reference for air pollution assessment and control.（iii）From the perspective of time distribution,the variation trend of monthly mean ozone concentration in Sichuan Province from January to December is approximately bimodal（"M"type）.It showed an obvious upward trend from January to May,reached the first peak in May,decreased slightly from May to June,began to rise from June to August,reached the second peak in August,showed a gradient decline from August to September,and reached the lowest point in December.The most polluted months are April to August.There are obvious seasonal differences in ozone concentration in Sichuan Province,and the relationship of ozone concentration in different seasons was as follows:Summer>Autumn>Spring>Winter.（iv）From the perspective of spatial distribution,GWR-LMM+IRF model can accurately estimate the spatial distribution characteristics of ozone pollution in Sichuan Province.On the whole,the ozone concentration of Sichuan Province is high in the east and low in the west,and the local pollution degree is high.The high pollution area is mainly distributed in the rapidly developing and densely populated eastern areas,while the low ozone area is mainly distributed in the underdeveloped and sparsely populated western areas.（v）It is important to identify the key factors affecting the concentration of ground-level ozone in Sichuan Province for fine control of air quality.Among the factors selected in this study,temperature,light radiation intensity and precursors are the key factors affecting ozone concentration in Sichuan Province.