基于非参数核回归的福州市PM2.5实证研究

Empirical Study on Fuzhou PM2.5 Based on Nonparametric Kernel Regression

采用非参数核回归方法,研究福州市PM2.5质量浓度与气象变量、其他污染物之间的关系.核回归拟合结果表明在控制其他变量取中位数时,PM2.5质量浓度与温度、风速呈现负相关,与PM10、CO质量浓度整体上呈现正相关,小范围内波动剧烈.PM2.5与NO2的关系表现出周期性,随着NO2质量浓度的不断增大,PM2.5质量浓度表现为先减小后增大再减小的趋势.PM2.5与SO2呈正相关.不同风向影响差异较大,其中当风向是东北风时PM2.5质量浓度较高.非参数核回归的拟合优度和均方误差均比逐步回归效果更优.

A nonparametric kernel regression method was used to study the relationship between PM2.5 concentration and meteorological variables and other pollutants in Fuzhou.Regression fitting results showed that when other variables were controlled to take the median,PM2.5 concentration was negatively correlated with temperature and wind speed,and positively correlated with PM10 and CO concentration as a whole,and fluctuated violently within a small range.The relationship between PM2.5 and NO2 showed periodicity.With the continuous increased of NO2 concentration,PM2.5 concentration tended to decrease first and then increase and then decrease.PM2.5 was positively correlated SO2.The influence of different wind directions varied greatly,among which PM2.5 concentration was higher when the wind direction was northeast wind.The goodness of fit and mean square error of nonparametric kernel regression were better than the stepwise regression.