Spatial And Temporal Characteristics Of The Two-way Feedback Between Meteorology And Aerosol Pollution

The two-way feedback between unfavorable meteorological conditions and accumulated aerosol pollution is a controlling factor for the PM_2.5 explosive growth in cumulative stages in Beijing.However,the variation characteristics of the two-way feedback in different periods and regions are still uncertain.Based on the multi-source observation data of PM_2.5,vertical meteorological elements and radiation,and machine learning techniques including temporal clustering analysis and Light GBM,this study focuses on analyzing the spatial and temporal characteristics of the two-way feedback effect and constructing PM_2.5 historical datasets to pave the way for the historical characterization of the feedback.From the temporal variation of the feedback before and after typical individual cases in Beijing,the two-way feedback in winter 2017/18under the control of emission reduction and favorable meteorological conditions is substantially weaker than that in previous winters,and the biases of near-surface temperature,specific humidity and relative humidity caused by the feedback on winter pollution days only accounts for 38%,65%and 36%of the meteorological variation caused by aerosols in the previous winter,respectively;During the top 10%of the most severe pollution days,the PM_2.5 increase due to the meteorological feedback in winter2017/18 was only 49%of that in the previous winter.In contrast to Winter 2017/18,large-scale circulation changes led to an increase in aerosol pollution in Beijing in March 2018,while the two-way feedback effect further worsened pollution.The above results suggest that effective pollution control and more favorable meteorological conditions lead to a further decrease in PM_2.5 reduction,but increased emissions and deteriorating meteorological conditions lead to an additional increase in PM_2.5.Based on the analysis of typical cases,this paper evaluates the two-way feedback effects under different meteorological conditions in Beijing area in winter since 2013.The results of the cluster analysis show that the process of upper warming and lower cooling in the boundary layer leads to a continuous increase in unfavorable degrees of meteorological conditions,while the two-way feedback effect becomes increasingly striking in the cooling effect on lower layers.Comparing the changes in meteorological conditions before and after 2017,it was found that moderately and heavily unfavorable conditions in winters of 2017?2019 decreased from 18.8%to 13.3%in proportion,particularly heavily unfavorable conditions with the most significant two-way feedback effect decreased by 63%,where more favorable meteorological conditions contribute to 40%of the PM_2.5 reduction in winters of 2017?2019.In terms of prediction results of an autoregressive integrated moving average(ARIMA)model based on long-term balloon observations since 1991,the proportion of heavily unfavorable conditions with the most significant two-way feedback effect will increase from 3.4%in Winter 2019to 4.8%in Winter 2020.Therefore,we suggest that the target for PM_2.5 reductions in the following winter need to take deteriorating meteorological conditions into account and should not be set too high.Following the analysis of the temporal characteristics of the two-way feedback effect in Beijing,this paper continues to gain insight into the feedback effect in different haze regions in China,and finds that,except for the two-way feedback effect in the Sichuan Basin,which is subject to middle and high level clouds,all key regions in China experience different degrees of feedback effects,with the degree of feedback in the North China Plain,the Guanzhong Plain,and the Northeast Plain being higher than that in the Two Lakes Basin,the Yangtze River Delta,and the Pearl River Delta.In most regions,the feedback effect of accumulated pollution leading to deteriorating meteorological conditions dominates most of the PM_2.5 explosive growth.The study of historical characteristics of two-way feedbacks requires PM_2.5 historical datasets as a basis,so this paper also constructs a machine learning model by extracting spatial characteristic variables,which achieves high precision prediction of PM_2.5 at hourly(R²=0.75),day(R²=0.84),and month(R²=0.88)by year(R²=0.87)scales.The model can reconstruct the historical PM_2.5 dataset since 1960,and pave the way for two-way feedback historical feature analysis.