The Study On Application Of Extreme Value Theory In Environmental Monitoring Data

Background and objective: Since the 21 st century,with the continuous development of the global economy and the continuous improvement of people’s health requirements,the problem of environmental pollution has become increasingly prominent,and the relationship between environmental pollution and population health has attracted widespread attention from the public and the government.A large amount of environmental health monitoring data in China have been accumulated in China.The rich data resources provide an important information basis for environmental health risk assessment and prediction,but this kind of data has not been fully mined and analyzed at present.Researchers or health workers are all concerning about the analysis of the excess of monitoring indicators at the selected reference level.From the perspective of data distribution,it is more practical to pay attention to its extreme values.The basic ideas and modeling strategies of statistical extreme value theory perfectly match the actual purpose of focusing on extreme values in environmental monitoring and risk assessment.Based on the air monitoring data of 13 districts or cities in Jiangsu Province from 2016 to 2018,this study explored the application of extreme value models from extreme value theory,BMM model and POT model,in air monitoring data in order to find appropriate and effective statistical method for Environmental monitoring and early warning.Main research contents: Based on the air pollution monitoring data in Jiangsu Province,systematically discuss: 1.The application strategies and goodness of fit of the two extreme value models in the analysis of extreme value data of environmental pollutants;2.Based on the analysis of the spatial distribution characteristics of the scale parameters and shape parameters of the two extreme value models,the ability of the parameters of the two models to describe the spatial changes of different air pollutant concentrations is compared;3.Based on the actual monitoring data in 2018,evaluate the prediction capabilities of the two extreme value models for different air pollutant data;Main results: 1.The P-P diagram and Q-Q diagram of the BMM and the POT show that the two extreme models can fit the distribution of extreme values of NO2,PM2.5,and O3-8h.The scatter points of the two graphs can be better dispersed in a straight line.However,in the P-P diagram of the PM2.5 BMM model,there are fewer discrete points in Xuzhou that deviate from the straight line.2.The scale parameters of the NO2 BMM model better describe the spatial change trend of the average daily NO2 concentration,especially the high and low value areas of the scale parameter distribution are basically consistent with the spatial locations of the high and low values of NO2 concentration.The spatial distribution of the scale parameters of the PM2.5 POT model in the high-value area in the northwest,the middle and low value areas in central Jiangsu and southern Jiangsu is basically consistent with the high-value area of daily average PM2.5 concentration and the low-medium value in southern Jiangsu.The spatial distribution positions of the high and low values of the scale parameters of the O3-8h BMM model and the POT model are basically consistent with the spatial distribution positions of the high and low values of the average daily concentration of O3-8h.3.The 5-month forecast value of the BMM model of NO2 has 7 cities where the forecast results are close to the actual monitoring results,the forecast value of the POT model in the same period has the forecast results of 10 cities that exceed or approximate the actual monitoring results,which can provide air pollution early warning information that is consistent with the actual monitoring.The 5-month forecast of the BMM model of 2.5PM has 1 city where the forecast results are close to the actual monitoring results,and the POT model’s forecast value of the same period has 7 cities where the forecast result exceeds or is close to the actual monitoring result,which can provide air pollution early warning information that is consistent with the actual monitoring.The 5-month forecast value of the BMM model of O3-8h has 10 cities where the forecast results are close to the actual monitoring results,the forecast value of the POT model in the same period has the forecast results of 13 cities that exceed or approximate the actual monitoring results,which can provide air pollution early warning information that is consistent with the actual monitoring.Compared with the BMM model,the POT model is more accurate in predicting the data of the three air pollutants.Main conclusions: Both the BMM model and POT model of extreme value theory can be applied to the distribution description and prediction analysis of extreme values of air monitoring data;The POT model based on GPD distribution is more sensitive to the extreme data distribution of air pollutants,and the prediction information is more accurate;The prediction results of the two extreme value models can provide air pollution early warning information consistent with the actual monitoring situation.