Study On Spatial And Temporal Distribution And Prediction Of Ozone Concentration In China

The development of the global economy has led to the rapid growth of the secondary industry,leading to increasingly serious global air pollution.Now that China has completed the building of a moderately prosperous society in an all-round way and achieved the first centenary goal,more attention should be paid to air pollution control and early warning.In our country is still facing the plight of a wide range of ozone（O₃）pollution,O₃ pollution situation is not optimistic,among them,the beijing-tianjin-hebei urban agglomeration is the political and cultural center,is also an important core,the northern economy O₃ overweight phenomenon has occurred frequently in recent years,directly affects the national O₃ pollution,O₃ pollution study of beijing-tianjin-hebei urban agglomeration and even the whole country extremely important practical significance.Nearby surface O₃ pollution can affect the entire ecological environment,and even threaten human health.However,the sources and causes of O₃pollution are complex and diverse,and the treatment is difficult.Therefore,it is extremely urgent to timely analyze the variation characteristics of the current O₃concentration pollution and establish a high-precision O₃concentration prediction method.In view of this,this paper uses pollutants(O₃-8h,PM_2.5,PM₁₀,SO₂,NO₂,CO)and meteorological elements（wind speed,precipitation,air pressure,sunshine duration,relative humidity）to analyze the hourly,daily,monthly and seasonal spatial and temporal distribution of O₃concentration in China’s large-scale and Beijing-Tianjin-Hebei urban agglomeration.The interaction between pollutants and meteorological elements and their effects on O₃concentration are discussed by using the geodetector method and correlation analysis.The analysis of the spatial and temporal distribution of O₃ pollution is a preliminary understanding of the occurrence mechanism of O₃pollution,which can only be treated after pollution rather than prevention.Therefore,based on the spatial and temporal distribution of O₃,the pollutants and meteorological elements after pretreatment are taken as the input data of the method.The support regression machine（SVR）and the extreme learning machine（ELM）method in machine learning are used to predict the O₃ concentration,and the multi-region method of the improved extreme learning machine is built to deeply analyze the influence of the interaction of different factors in each region on the O₃ concentration,and to predict the O₃ concentration in the next 24 hours.Then,MAE,MAPE,RMSE and R² were used to evaluate the accuracy of the prediction method.The main research results are as follows:（1）In July 2019,the hourly variation of O₃-8h concentration began to rise rapidly from 9 am and reached the maximum concentration at 19 am.In 2019,the diurnal variation of O₃-8h concentration in China and the Beijing-Tianjin-Hebei urban agglomeration showed an inverted"U"shape,with 7.98%of the cities in China failing to meet the standards,and 33.33%of the cities in Beijing-Tianjin-Hebei urban agglomeration failing to meet the standards.In the Beijing-Tianjin-Hebei urban agglomeration,there were substandard cities in May,June,July and September,and the substandard rate was as high as 78.57%in June.The O₃ pollution was the most serious in summer and the lowest in winter,with a non-compliance rate of 7.8%nationwide and 50%in the Beijing-Tianjin-Hebei urban agglomeration.O₃pollution in the Beijing-Tianjin-Hebei urban agglomeration is more serious than the national average level.O₃ is more likely to be generated in the daytime than at night,and the O₃concentration is higher in summer and lower in winter.（2）The annual mean mass concentration of O₃-8h in China is the highest in Jincheng and the lowest in Jixi.Only in some parts of southwest China,the mass concentration of O₃-8h in spring is higher than that in summer,autumn and winter.The annual mean mass concentration of O3-8h in the Beijing-Tianjin-Hebei urban agglomeration increased gradually from the north to the south,and the highest concentration was found in Anyang,which was consistent with Zhangjiakou,Beijing,Langfang,Cangzhou,Tangshan and Qinhuangdao in summer,and all exceeded the first-level limit.The Moran’s I index in China was 0.215,and the proportion of HH,LL,HL and LH was 42.7%,36.6%,16.3%and 5.8%,respectively.Hot spots are concentrated in North China,East China,Central China and Northwest China,while cold spots are mainly concentrated in Southwest China,South China,Central China and Northeast China.The spatial distribution of O₃-8h in Beijing-Tianjin-Hebei urban agglomeration is mainly that high value areas are adjacent to high value areas,low value areas are adjacent to low value areas,and a small part of high value areas are adjacent to low value areas.The mass concentration of O₃-8h in China presents a significant positive spatial correlation and aggregation.（3）The trend of O₃-8h average daily concentration was roughly"U"type,and SO₂,PM_2.5,PM₁₀,NO₂ and CO were inverted"U"type.The concentration of five pollutants was the highest in North China,and the lowest in Southwest China.The concentrations of PM₁₀ and PM_2.5 were high in Northwest China,while the concentrations of PM_2.5 were low in Southwest China,East China,Northeast China and South China.The concentrations of NO₂ and PM_2.5 were high in North China and low in Southwest China,while the concentrations of O₃-8h were high in North China and low in Southwest China.O₃-8h concentration is negatively correlated with pollutant factors,significantly negatively correlated with air pressure and relative humidity,and significantly positively correlated with wind speed,precipitation and sunshine duration.There was a positive correlation between sunshine and O₃-8h concentration in spring,and a negative correlation between autumn and winter.Pollutants and meteorological factors may affect the generation,transport and treatment of O₃concentration.（4）Compared with the worst CS-HK-ELM method in China,MAE,RMSE and MAPE errors were reduced by 73.7%,70.3%and 80.6%,respectively,by 64.5%,64.5%and 59.1%in Anyang,by 68.0%,66.0%and 68.3%in Beijing.Chengde reduced by 56.9%,63.0%and 60.6%,and Tianjin69.1%,65.1%and 65.1%;The MAE,RMSE and MAPE of traditional ELM were 7.395,1.408 and3.719 lower than those of PK-ELM in Anyang.The precision of RK-ELM method in China,Anyang,Chengde and Tianjin is higher than that of PK-ELM method.The best application area of all methods is the whole country,and the precision of SVR method in Beijing is the worst.The R² of CS-HK-ELM method in Anyang City was 0.901,which showed the highest goodness of fit,and the ELM method in Chengde City was the lowest.The prediction accuracy and goodness of fit of CS-HK-ELM method in each region are the highest.Therefore,the kernel function and optimization algorithm can not only improve the accuracy and stability of the ELM method,but also accurately capture the abrupt node of O3-8H mass concentration,and the optimization degree of the mixed kernel function is greater than that of the single kernel function,and the combination of the mixed kernel function and cuckoo algorithm has the best effect.