空气质量指数及其含量的极值模型分析

Extreme value model analysis of air quality index and content

选取北京市2016年1月~2018年3月的日空气质量指数数据,建立广义Pareto分布模型分析空气质量指数的超阈值序列,利用轮廓似然估计确定精确的参数置信区间,得到重现水平和中度、重度、严重污染天气发生的概率.将极值理论与相关结构Copula相结合,建立二元超阈值模型对空气质量指数含量PM2.5与PM10之间的极值相关性进行了研究,结果显示其中一个超限值的条件下另一个也超过限值的概率很大,两者具有较大的尾部相关性,它们的叠加作用加重了空气污染的程度.

Based on the daily air quality index data of Beijing from January 2016 to March 2018,a generalized Pareto distribution model was established to analyze the threshold excesses sequence of air quality index.Used the profile likelihood estimate to determine the accurate parameter confidence interval.The return level and the probability of moderate-pollution,severe-pollution were obtained.Combining the extreme value theorywith the Copula,built a bivariate threshold excess model to show the extreme dependent between PM 2.5 and the PM 10.The result showed the probability of one exceeding the limit was very high in the case of another exceeding the limit value.The two have a large tail dependence,and the superposition increased the degree of air pollution.