耦合主被动遥感数据与空气质量模式的PM2.5优化估算研究

Most of the research on PM2.5 concentration in remote sensing estimation has not considered the spatial-temporal differentiation between region AOD and PM2.5 ratio, and the simulation precision of air quality model is limited by the accuracy of the source inventory. From the point of using multi-source remote sensing information to improve the accuracy of model simulation, This project aims at: (1) the method of temporal and spatial expansion of AERONET AOD observation based on deep learning, to get more AOD observation data; (2) the model AOD profile correction method based on CALIPSO extinction coefficient information, to obtain more accurate AOD profile space distribution; (3) introducing optimal estimation method, to construct the near-surface PM2.5 concentration estimation model of optimal based on multi-source active and passive remote sensing information combined with CAMx model; (4) taking Beijing-Tianjin-Hebei as a target area,   to verify the accuracy of the PM2.5 estimation model constructed in this study and perform uncertainty analysis, to assess the validity of the optimal estimation method. Through the optimal estimation method system for PM2.5 concentration constructed in this study, higher spatial-temporal resolution and less uncertainties can be obtained for PM2.5 concentration, and supports for understanding the regional PM2.5 pollution situations and improving dynamic monitoring capability of PM2.5 can be provided.

目前遥感估算PM2.5浓度的研究大多未考虑区域AOD与PM2.5比率的时空分异性，而空气质量模式的模拟精度则受限于源清单的准确性。本项目拟从利用多源主被动遥感信息提高模式模拟精度的角度入手，研究：（1）基于深度学习的AERONET气溶胶观测时空扩展方法，获取更多点位的气溶胶观测数据；（2）基于CALIPSO消光系数信息的模式AOD廓线校正方法，提供更为准确的AOD廓线空间分布；（3）引入最优估计方法，构建耦合多源主被动遥感信息和CAMx模式的近地面PM2.5浓度优化估算模型；（4）以京津冀为靶区，对本研究构建的PM2.5估算模型进行精度验证及不确定性分析，评估本研究提出的优化估算方法的有效性。通过本研究构建的PM2.5浓度优化估算方法体系，可以获取较高时空分辨率 、不确定性较小的PM2.5浓度，为更精细的了解区域PM2.5污染状况、提高PM2.5动态监动态监测能力提供支持。

本项目的科学问题是建立蕴含物理机制的PM2.5浓度估算模型，获取高时空分辨率、不确定性小的PM2.5时空分布格局。为此，项目首先应用深度学习算法构建大气污染物浓度观测数据、气象观测数据与 AERONET 气溶胶观测数据间的关系模型，对 AERONET 气溶胶数据进行时空扩展，获取更多点位的气溶胶数据作为气溶胶遥感数据产品及模式模拟 AOD 浓度订正的基础资料。其次，研究利用 CALIPSO 气溶胶观测信息对 CAMx 模式模拟的气溶胶光学厚度廓线的校正方法，进一步降低模式模拟结果的不确定性，提高根据气溶胶柱浓度估算近地面 PM2.5浓度的精度。在廓线校正的基础上，采用最优估计方法对 CAMx 模式模拟 AOD 浓度进行优化，进而构建基于多源遥感信息的近地面PM2.5浓度优化估算模型，实现对 AOD 以及近地面PM2.5浓度的优化估算。为验证优化算法的可靠性，项目以京津冀地区为靶区，开展PM2.5浓度遥感反演和精度验证试验，结果显示评估决定系数提高到0.50以上，均方根误差控制在30微克每立方米以内。本研究的成果可为提高区域PM2.5动态监测能力及公共卫生政策制订提供数据支持。