Ground-level NO₂ Concentration Retrieval And Downscaling Research Based On Tropomi Data

The issue of air pollution has long been a concern worldwide,and China has also carried out a large amount of work to control air pollution.Pollutants such as SO₂,PM_2.5,and PM₁₀have achieved significant control effects,but the current situation of O₃pollution in China remains relatively serious.The high and concentrated emissions of NOx and VOCs are the primary reasons for O₃pollution in China in recent years.NO₂is one of the components of NOx,which has significant harm to human health and environmental safety.Therefore,effective monitoring and analysis of the spatial and temporal changes of near-surface NO₂concentration are crucial for China to improve air quality and living environment for residents.The NO₂concentration monitoring data of China’s national environmental monitoring stations from 2016 to 2020 has been selected to explore the spatiotemporal distribution patterns of near-surface NO₂concentration.This study developed monthly and daily near-surface NO2 concentration estimation models using random forest.Geographically weighted regression（GWR）and multiscale geographically weighted regression（MGWR）models were employed to downscale the inverted 0.05°×0.05°near-surface NO2 concentration data,resulting in a final resolution of 0.01°×0.01°.The main research results of this article are as follows.It is found that the near-surface NO₂concentration has increased and then decreased since 2016-2020,with the highest value appearing in 2017,and showing a downward trend overall.In terms of seasonal variation,the near-surface NO₂concentration exhibits the characteristics of winter>autumn>spring>summer,which is related to both human emissions and natural conditions.At the monthly scale,the near-surface NO₂concentration shows a U-shaped variation,with the monthly average minimum value appearing in July-August in the five-year period,influenced by summer high temperature,rainfall,and sufficient sunshine.In terms of spatial distribution,the near-surface NO₂concentration in northern China is higher than that in southern China,and that in the eastern region is higher than that in the western region.The high-value areas in the country are mainly distributed in the economically developed and densely populated North China Plain,the Yangtze River Delta region and the coal-rich and heavy industrial concentrated Fenwei Plain.Through analyzing the TROPOMI NO₂tropospheric column concentration data and ground monitoring station data,we found that their spatiotemporal distribution patterns are relatively consistent,with a correlation coefficient of 0.50.This indicates that the TROPOMI NO₂tropospheric column concentration data can be used to estimate the distribution of near-surface NO₂concentration.Based on the constructed near-surface NO₂concentration estimation dataset,this study used a random forest model to establish a monthly-scale and daily-scale near-surface NO₂concentration estimation model with a spatial resolution of 0.05°×0.05°,and conducted an estimation study on the near-surface NO₂concentration in 2020.After feature factor screening,the daily-scale estimation model established in this study was divided into four seasons:spring,summer,autumn,and winter,with good fitting ability for the winter and autumn models,and the R²of ten-fold cross-validation test sets were 0.75（RMSE=9.14ug/m³）and 0.73（RMSE=7.77ug/m³）,respectively.For the verification of the estimated daily-scale near-surface NO₂concentration results in the corresponding seasons of the model,the R²reached 0.67(RMSE=12.79ug/m³)and 0.70（RMSE=11.37ug/m³）,respectively.This study also selected the Beijing-Tianjin-Hebei,Yangtze River Delta,and Pearl River Delta regions as sub-regions for the study,and verified the daily average near-surface NO₂concentration,finding that in the autumn and winter seasons in the Beijing-Tianjin-Hebei region,the estimated daily NO₂concentration change trend was consistent with the measured NO₂concentration trend at ground monitoring sites,with an R²of 0.70.This study demonstrates that the model has some application value in the estimation of NO₂concentration in small areas.The study area for downscaling was selected to be Shanghai and surrounding areas.In this study,GWR and MGWR models were used to downscale the monthly mean concentration of near-surface NO₂based on high-resolution auxiliary data,resulting in 0.01°×0.01°monthly mean concentration of near-surface NO₂.By analyzing the fitting ability of the models,the best result of R²for GWR model was0.54（RMSE=9.51ug/m³）,while for MGWR model,the best R²was 0.55（RMSE=10.24ug/m³）,indicating that MGWR model had a more stable performance overall but still had room for improvement in RMSE.Through downscaling,the spatial resolution of near-surface NO₂concentration data was significantly improved.The GWR and MGWR models were able to capture the spatial distribution differences of NO₂concentration in small areas that are difficult to obtain with coarser resolution data,laying a foundation for further exploration of the spatial distribution of near-surface NO₂concentration and identification of point source pollution inside the city.In this study,random forest model,geographically weighted regression model and multi-scale geographically weighted regression model were used to estimate the near-surface NO₂concentration.The obtained near-surface NO₂concentration products with high spatial resolution and spatial-temporal continuity provide data support for the study of near-surface NO₂concentration distribution.