大气污染物浓度全方位动态监测是进行区域大气污染精细化防控的重要前提。为开展长三角地区小时分辨率PM_2.5浓度无缝制图，本研究通过耦合AOD缺失信息重建与多模数据融合技术，建立了一套能够有效集成卫星遥感、地面观测、数值模拟等多源异构数据资料的近地面PM_2.5浓度无缝制图方案，并据此生产了2015年—2020年长三角地区小时分辨率无缝PM_2.5浓度格点数据产品。结果表明：本研究生产的PM_2.5浓度无缝格点产品与国控站点观测数据的交叉验证相关系数达0.9，平均偏差不超过10 μg·m^-3。较于空间分布不均且相对稀疏的站点观测PM_2.5浓度资料，面域无缝PM_2.5浓度格点数据更能有效揭示长三角地区PM_2.5污染的时空变化特征；在2015年—2020年研究期内，其平均下降速率超过3 μg·m^-3·a^-1。本研究发展的PM_2.5浓度无缝制图方法和生产的相关数据产品有望为区域灰霾污染防控和PM_2.5暴露健康风险评估研究提供方法参考和基础数据支撑。

Monitoring concentrations of atmospheric particulate matters is essential to regional haze pollution prevention and control. Satellite-based Aerosol Optical Depth (AOD) data have been frequently used to map regional PM_2.5 concentrations. However, the resultant PM_2.5 concentration maps are always spatially incomplete due to significant data gaps in satellite-based AOD retrievals. This study aims to fill data gaps in AOD imageries to support spatially contiguous PM_2.5 concentration mapping on an hourly basis in the Yangtze River Delta.An integrated data fusion approach was developed to seamlessly gear up the missing AOD imputation and multimodal data fusion approaches. Specifically, all available Himawari-8 AOD observations during the daytime were fused to maximize hourly AOD coverage in each single snapshot. To further tackle data gaps in fused AOD maps, a virtual AOD monitoring network was constructed by estimating AOD at each state-controlled air quality monitoring station based on ground measured air pollutant concentration. This way enables us to extend the sparsely distributed aerosol monitoring network nationwide, which significantly improves the spatial coverage of AOD. Subsequently, the reconstructed satellite AOD and PM inferred AOD were fused with AOD simulations from MERRA-2 using the optimal interpolation method to generate spatially contiguous yet far more accurate AOD reanalysis. Spatially complete PM_2.5 concentration maps were finally generated on hourly basis over the study region using the random forest method.Ground validation results indicate that AOD values inferred from air quality measurements agree well with in situ AOD measurements, with R of 0.90 and RMSE of 0.13. The analyzed spatially complete AOD dataset has a correlation of 0.86 and RMSE of 0.16 compared with in situ AOD data, which is much higher than that of raw Himawari-8 AOD. The estimated PM_2.5 concentration data also have a promising accuracy, with R of 0.9 and mean absolute error of 9.87 μg m^-3 compared with in situ PM_2.5 measurements.Compared with sparsely distributed in situ PM_2.5 measurements, this spatially contiguous PM_2.5 concentration dataset has great advantages in assessing PM_2.5 variations in space and time in the Yangtze River Delta. Statistically significant decreasing trend over the whole study area also highlights the effectiveness of clean air actions in reducing PM_2.5 loadings across China. Overall, the proposed method can be practically used for future PM_2.5 mapping practices and the generated spatially contiguous PM_2.5 concentration dataset is a promising data source for the assessment of the human exposure risk to haze pollution.