Study On Probability Estimation Of Haze In Beijing Based Cumulative Logistic Regression Model

The frequency of haze in China has attracted the attention of all sectors of society.In many areas,haze has been forecast as a warning for severe weather.With the rapid development of economy,energy consumption has risen sharply.Air pollution in Beijing has become more and more serious.Affected by coal-fired emissions and pollutants discharged from motor vehicle exhaust,the fine particulates have become the most influential factors affecting the air quality in Beijing.Logistic Regression Model is a special form of logarithmic linear model,which is a statistical analysis method for classification variables.Logistic regression analysis is flexible and simple.The cumulative Logistic regression model uses all existing data to evaluate the role of the independent variable in the logarithmic ratio of the reaction variables in or more than(or conversely)of a particular category.Haze pollution is closely related to meteorological factors,and when the emission of pollution sources is relatively stable,the dispersal and transmission capacity of the atmosphere mainly depends on meteorological elements.The diffusion of atmospheric pollutants is affected by the combination of wind,precipitation,temperature and humidity.First of all,this paper collected the January 1,2016 solstice every three hours on December 31,meteorological data and PM2.5 concentration data in Beijing,has carried on the classification on the degree of pollution haze weather according to standard of air quality announced by China's environment ministry;Secondly,the paper study the PM2.5 pollution in Beijing,daily distribution and seasonal distribution with descriptive statistical analysis,and use the correlation analysis methods to analyze the relationship between PM2.5 pollution in Beijing and meteorological factors including relative humidity,average wind speed,wind direction;Then the paper establish probability estimation model of haze in Beijing to explore the relationship between the probability of haze weather and seasonal factors,various meteorological factors such as pressure,temperature,humidity,precipitation,wind scale and so on,based on the binary classification logistic regression model.The relative humidity and the season and are found to be strongest factors.In winter,haze weather happens 2.73 times than that of summer and 1.6 times than that of spring.For each 10%increase in relative humidity,the probability of occurrence of haze weather has increased by 48%.Lastly,with the cumulative Logistic regression model,the paper study the probability of haze weather in Beijing.As the season and the wind direction differ from each other,the paper build separate cumulative Logistic regression models,and the meteorological factors are used as independent variables to find out the main meteorological factors which affect the haze weather under different season and different wind direction,at the same time the models are used to the estimate the probability of haze weather at all levels under different meteorological factors.