【摘要】 为探究地面气溶胶质量浓度遥感反演模型在年际及区域尺度上的适用性,确定一种可以在区域尺度上业务运行的反演模型,基于2015-2016年河南省28个城市MODIS气溶胶光学厚度、PM_2.5质量浓度、地面气象观测能见度和相对湿度,分别采用半经验模型和非线性多元回归模型建立了分季节反演模型,并基于2017年观测数据对两种模型的精度进行了对比评价。结果表明:半经验模型中标高订正的AOD与经过湿度订正的PM_2.5质量浓度建立反演模型效果最好,四个季节反演模型决定系数R²分别为0.3938、0.3507、0.3488及0.4212。非线性多元回归模型最优组合为AOD、气溶胶波长指数、气溶胶标高与相对湿度,四个季节反演模型决定系数R²分别为0.4295、0.3598、0.4099和0.5616。两种反演模型验证结果均通过0.001的显著性检验,非线性多元回归模型的验证决定系数大于半经验模型的,春季和冬季非线性多元回归模型的验证斜率更接近1,而夏季和秋季半经验模型验证斜率更接近1。非线性多元回归模型建模及验证相关性更高,且可有效避免高相对湿度条件下异常值的出现,业务适用性更好。更多还原

【Abstract】 In order to explore the applicability of surface aerosol mass concentration remote sensing inversion models at the interannual and regional scales, and determine an inversion model that can operate on the regional scale, a semi-empirical model and a nonlinear multifactor regression model are employed to establish a seasonal retrieval model based on MODIS Aerosol Optical Depth（AOD）, PM_2.5 mass concentration, observed visibility and relative humidity at the meteorological stations in 28 cities of Henan province during 2015-2016. In addition, with the 2017 observation data, the accuracy of the two models is compared and evaluated. The results show that the combination of aerosol height modified AOD and relative modified PM_2.5 concentration shows the best performance in semi-empirical method with the determination coefficient R² being 0.3938, 0.3507, 0.3488 and 0.4212 respectively in four seasonal models（99.9% confidence）.The combination of AOD, Angstrom Exponent, aerosol height and relative humidity shows the best performance in logarithm multifactor nonlinear regression method with the determination coefficient R² being 0.4295, 0.3598, 0.4099 and 0.5616 respectively in four seasonal models（99.9% confidence）.The verification results of the two methods are at the 99.9% confidence level, while logarithm multifactor nonlinear method shows higher determination coefficient R². Logarithm multifactor nonlinear regression method shows a better slope（close to 1） in spring and winter verification, while in summer and autumn the slope verified with the semi-empirical method is better, close to 1. So, the modeling of non-linear multifactor regression model and verification are more correlated, and can effectively avoid the occurrence of abnormal values under high relative humidity conditions, so it has a better applicability in operation.更多还原