Analysis Of O₃ Pollution Characteristics And Sensitivity To Meteorological Elements In Bengbu City

Meteorological conditions play important roles in the formation of ozone(O₃).In this study,the correlation analysis method was used to select the factors that have strong influence on the maximum 8-hour average O₃(O₃-8h)concentrations in Bengbu city from 43meteorological elements and 5 pollutant influencing factors,and then the prediction model was constructed by multiple linear regression method and BP neural network method.It was found that the performance of BP neural network model is superior to multiple linear regression prediction model.Based on BP neural network prediction model,the sensitivity of O₃ concentration to major meteorological factors in Bengbu city was quantitatively calculated by scenario analysis.This study also analyzed the multi-time scale characteristics of O₃pollution in Bengbu city and built 4 seasonal prediction models for O₃ concentration.The main research results are as follows:(1)The mean O₃-8h concentration in Bengbu showed an increasing trend before 2018 and then decreased.The monthly variation of O₃-8h concentration presents a'bimodal'distribution feature,with the peak values concentrated in April-June and September.In the diurnal variation,O₃-8h concentration reached the lowest at 7 and the highest at about 15.The weekly variation of O₃-8h concentration showed a trend of decreasing first,rising then decreasing again.The O₃-8h concentration after 12:00 noon presents the'weekend effect'that it is higher on weekends than on weekdays.(2)O₃-8h concentration is strongly positively correlated with temperature,positively correlated with absolute water vapor content and negatively correlated with relative humidity,positively correlated with radiation,negatively correlated with wind speed,negatively correlated with NO₂ and CO concentration,and negatively correlated with air pressure.The determination coefficient R² of the multiple linear regression prediction model for O₃-8h concentration is 0.739,and the average prediction accuracy in 2020 is 82.2%.The R²of BP neural network model of O₃-8h concentration is 0.791,and the average prediction accuracy in2020 is 84.0%.The prediction ability of the BP neural network model is in the order of spring,summer,autumn and winter,and the overall effect of the model is better than that of the multiple linear regression model.(3)The sensitivity of O₃-8h concentration to temperature,relative humidity,wind speed and air pressure was analyzed using scenario analysis method based on BP neural network prediction model.Sensitivity results were 3.2?g·m^-3·?^-1,-0.2?g·m^-3·%^-1,-3.4?g·m^-3·m^-1·s,-0.2?g·m^-3·h Pa^-1.(4)The BP neural network prediction model of O₃-8h concentration in four seasons in Bengbu city was established.Compared with the annual prediction model,the average forecast accuracy of the seasonal prediction model increased by 0.9%,2.1%,4.1%and 7.4%in spring,summer,autumn and winter respectively.