主成分和回归分析方法在大气臭氧预报的应用——以北京夏季为例

利用北京市区两个典型观测站的大气臭氧(O3)及前体物浓度观测资料和气象要素观测数据,分析了影响大气O3浓度各要素的相关性,并采用主成分分析和逐步回归方法构造大气O3浓度统计预报方程.结果表明,大气O3与前体物一氧化氮(NO)、二氧化氮(NO2)和气象要素呈现较好相关性;发现由于变量间共线性问题,逐步回归方法不能给出可接受的回归方程,而采用主成分分析和逐步回归方法相结合,可避免共线性问题,由前体物浓度和气象要素给出较好的北京大气O3浓度统计预报方程,可决系数R2分别为0.78(IAPs2007)、0.88(IRSAs2007)和0.64(IAPs2005),能够有效地预报O3浓度的变化情况.

Principal component and multiple regression analysis predicting ozone concentrations:Case study in summer in Beijing

Data of atmospheric pollutants and meteorological variables in the summer of two stations of Beijing were employed to analysis the relationship between ozone precursors and meteorological parameters and to predict the concentration of ozone in the atmosphere using both multiple linear regression and Principal Component Analysis (PCA) methods. For both sites the pollutants (NO and NO2) and meteorological parameters (air temperature, relative humility) were highly correlated with ozone. We found that simple stepwise regression analysis fails to build accurate regression equations owing to the existence of multicollinearity among the independent variables. A variable selection method combination of PCA and stepwise regression analysis was used to obtain subsets of the predictor variables (ozone precursors and meteorological parameters) to produce a model free of the multicollinearity problem. The formulas were validated and have R2 values of the order of 0.78 (IAPs2007), 0.88 (IRSAs2007) and 0.64 (IAPs2005).