成都气溶胶散射吸湿增长因子多变量影响分析

Impact Analysis of Multivariable of Aerosol Scattering Hygroscopic Growth Factor in Chengdu

基于成都市2017年10—12月逐时“干”状态气溶胶散射系数和吸收系数观测数据,结合该时段同时次大气能见度(V)、相对湿度(RH)和二氧化氮(NO_(2))浓度监测资料,采用光学综合法计算气溶胶散射吸湿增长因子(f),探究f因子对RH和气溶胶组分协同变化的复杂响应关系。首先,利用相关性分析和方差膨胀因子相结合的方法诊断RH、黑碳气溶胶(BC)质量浓度(ρ_(BC))、ρ_(BC)/PM_(1)质量浓度(ρ_(PM1))、ρ_(BC)/PM_(2.5)质量浓度(ρ_(PM2.5))、ρ_(BC)/PM_(10)质量浓度(ρ_(PM1)0)、ρ_(PM1)/ρ_(PM2.5)、ρ_(PM1)/ρ_(PM10)和ρ_(PM2.5)/ρ_(PM10)之间的多重共线性问题,提出f因子解释变量集(RH、ρ_(BC)、ρ_(BC)/ρ_(PM2.5)、ρ_(PM1)/ρ_(PM2.5)和ρ_(PM2.5)/ρ_(PM10))。其次,针对f因子单变量广义可加模型(GAM)的分析结果表明,RH虽然是f因子的决定性影响因素,但f因子与其他解释变量,尤其是ρ_(BC)也呈显著非线性关系(均通过α=0.01的显著性检验)。最后,以RH、ρ_(BC)、ρ_(BC)/ρ_(PM2.5)、ρ_(PM1)/ρ_(PM2.5)和ρ_(PM2.5)/ρ_(PM10)为自变量构建了f因子多变量影响的GAM模型,其对应的调整后的判定系数和压轴回归决定系数分别为0.787和0.797,气溶胶散射吸湿增长因子的模拟效果得到显著提升。

Based on the hourly observation data of“dry”aerosol scattering coefficient and absorption coefficient,as well as the simultaneous monitoring data including visibility(V),relative humidity(RH),and nitrogen dioxide(NO_(2))during October to December 2017 in Chengdu,aerosol scattering hygroscopic growth factor(f)was calculated by optical synthesis method,and then the complex response-relationship of f to synergistic action between RH and aerosol components was further investigated.Firstly,there were strong multicollinearity among different indicators including RH,ρ_(BC),ρ_(BC)/ρ_(PM1),ρ_(BC)/ρ_(PM2.5),ρBC/ρ_(PM1)0,ρ_(PM1)/ρ_(PM2.5),ρ_(PM1)/ρ_(PM10),andρ_(PM2.5)/ρ_(PM1)0 by using the method of correlation analysis combined with variance inflation factor(ρ_(BC),ρ_(PM1),ρ_(PM2.5)andρ_(PM10)represent mass concentrations of BC,PM_(1),PM_(2.5),and PM_(10),respectively).Subsequently,the explanatory variable set of f which included RH,ρ_(BC),ρ_(BC)/PM_(2.5),ρ_(PM1)/ρ_(PM2.5),andρ_(PM2.5)/ρ_(PM1)0 were obtained.Secondly,the diagnostic results of f single-variable by adopting generalized additive model(GAM)validated that although RH was the decisive influencing factor of f,f also had significant nonlinear relationship with other explanatory variables(all passed the significance test ofα=0.01),especiallyρ_(BC).Finally,the model of the multivariable influence factors of f were constructed based on GAM and the corresponding independent variables including RH,ρ_(BC),ρ_(BC)/ρ_(PM2.5),ρ_(PM1)/ρ_(PM2.5)andρ_(PM2.5)/ρ_(PM10),respectively.The corresponding adjustment determination coefficient(R 2)and the determination coefficient of reduced major axis regression were 0.787 and 0.797,respectively,which significantly improved the simulation effect of aerosol scattering hygroscopic growth factor.