Establishment Of A Near-Surface Ozone Prediction Model In Shijiazhuang

Under the background of global warming,ozone（O₃）pollution events occur frequently,which has become the key problem to affect the air quality in China.Real-time monitoring of O₃ data by automatic ozone monitoring stations has been set up in various cities in China.However,due to the small number and uneven distribution of monitoring sites,it is not possible to provide more accurate regional O₃ concentrations.Therefore,the prediction model is an effective tool to capture regional O₃ concentration.As a typical city of O₃ pollution in Shijiazhuang,it is urgent to develop a prediction model for regional O₃monitoring.This paper proposes a background calculation method to study the seasonal and hourly pollution characteristics of O₃ using the production and monitoring values of O₃,taking Shijiazhuang as an example.The relationship between O₃ and precursor factors such as NO₂,VOCs,PM_2.5,CO,and meteorological conditions such as UVB,temperature,relative humidity,wind speed and wind direction was analyzed.Based on the three models of Stepwise Linear Regression（SLR）,Random Forest（RF）and Gradient Boosting Decision Tress（GBDT）,the factors were optimized and the O₃ prediction model was built and evaluated.The research conclusions are as follows:1)Calculating of primary pollutants in Shijiazhuang.It was found that the O₃ pollution was mainly concentrated in June to July,showing a significant single peak change.The peak of O₃ monitoring value,background value and production value occurred in June,June to July and September respectively,and the lowest value occurred in December,October to December and December respectively.The contribution of the O₃ production value to the monitoring value was as follows:autumn（October and November）>winter（December,January,February）>summer-I（August and September）>spring（March,April,May）>summer-II（June and July）,and the main contribution time was as follows:autumn>summer-I>spring>winter>summer-II,which the proportion of the hour scale is more than 70%.2)According to the O₃/NOx and NO₂/NO ratios,Shijiazhuang was legally identified as a VOCs-controlled area.And the main contribution of NOx to O₃ was analyzed.The partial correlation coefficient method was used to determine the correlation between O₃and PM_2.5,and the Pearson correlation coefficient method was used to determine the correlation between O₃ and CO.It was found that PM_2.5 was positively correlated with O₃in summer and autumn,while CO was negatively correlated with O₃ in all seasons except summer.3)UVB was found to have a moderate positive correlation with O₃ in different seasons,through correlation analysis.The variation trend of O₃ at different stages of temperature,relative humidity and atmospheric pressure was identified by using the piecewise method.The effects of wind speed and direction on the horizontal transport of O₃ in Shijiazhuang were analyzed by using the polar coordinates of wind speed and wind direction.The main potential sources of O₃ pollution in Shijiazhuang are cities within 500 km of the city.4)The importance and significance of the introduced factors are evaluated through the classification tree of RF and SLR,and the optimization of the introduced factors is realized.RF and GBDT were used to establish seasonal prediction models for O₃production value and monitoring value,respectively.The R² was higher than the threshold value 0.6,and RMSE<20μg·m^-3.The model can achieve good prediction at both low and high concentrations.At the same time,the model is improved by using SLR to improve the prediction accuracy,and the generalization of the model is ensured by the historical simultaneous test and regional test.