Key Laboratory for Urban Habitat Environmental Science and Technology, Shenzhen Graduate School, Peking University, Shenzhen 518055, PR China.
Department of Urban Planning and Design, The University of Hong Kong, Hong Kong, SAR, China; Shenzhen Institute of Research and Innovation, The University of Hong Kong, Shenzhen 518075, PR China.
Sci Total Environ. 2018 Mar 15;618:819-828. doi: 10.1016/j.scitotenv.2017.08.209. Epub 2017 Nov 11.
Satellite-derived aerosol optical depth (AOD) has been proven effective for estimating ground-level particles with an aerodynamic diameter <2.5μm (PM) concentrations. Using a time fixed effects regression model, we compared the capacity of two AOD sources, Moderate Resolution Imaging Spectroradiometer (MODIS) and Visible Infrared Imaging Radiometer Suite (VIIRS), to estimate ground-level PM concentrations over a heavily polluted region in China. Regarding high-quality AOD data, the results show that the VIIRS model performs better than the MODIS model with respect to all model accuracy evaluation indexes (e.g., the coefficient of determination, R, of the VIIRS and MODIS models are 0.76 and 0.71 during model fitting and 0.72 and 0.66 in cross validation, respectively), the potential for capturing high PM concentrations, and the precision of annual and seasonal PM estimates. However, the spatiotemporal coverage of the high-quality VIIRS AOD is inferior to that of the MODIS AOD. We attempted to include medium-quality VIIRS AOD data to eliminate this, while exploring its influence on the performance of the VIIRS model. The results show that it improves the spatiotemporal coverage of the VIIRS AOD dramatically especially in winter, although a decline in model accuracy occurred. Compared to the MODIS model, the VIIRS model with both high-quality and medium-quality AOD data performs comparably or even better with respect to some model accuracy evaluation indexes (e.g., the model overfitting degree of the VIIRS and MODIS models are 7.46% and 5.82%, respectively), the potential for capturing high PM concentrations, and the precision of annual and seasonal PM estimates. Nevertheless, the VIIRS models did not perform as well as the MODIS model in summer. This study reveals the advantages and disadvantages of the MODIS and VIIRS AOD in simulating ground-level PM concentrations, promoting research on satellite-based PM estimates.
卫星衍生气溶胶光学厚度（AOD）已被证明可有效估算空气动力学直径<2.5μm 的地面颗粒物（PM）浓度。我们使用时间固定效应回归模型，比较了两种 AOD 数据源——中分辨率成像光谱仪（MODIS）和可见红外成像辐射套件（VIIRS），在中国一个污染严重地区估算地面 PM 浓度的能力。关于高质量 AOD 数据，结果表明，VIIRS 模型在所有模型精度评估指标（例如，模型拟合时 VIIRS 和 MODIS 模型的决定系数，R 分别为 0.76 和 0.71，交叉验证时分别为 0.72 和 0.66）、捕捉高 PM 浓度的能力以及年际和季节性 PM 估算的精度方面，表现优于 MODIS 模型。然而，高质量 VIIRS AOD 的时空覆盖范围不如 MODIS AOD。我们试图纳入中等质量的 VIIRS AOD 数据来消除这一影响，同时探索其对 VIIRS 模型性能的影响。结果表明，尽管模型精度有所下降，但它极大地提高了 VIIRS AOD 的时空覆盖范围，尤其是在冬季。与 MODIS 模型相比，具有高质量和中等质量 AOD 数据的 VIIRS 模型在某些模型精度评估指标（例如，VIIRS 和 MODIS 模型的模型过拟合程度分别为 7.46%和 5.82%）、捕捉高 PM 浓度的能力以及年际和季节性 PM 估算的精度方面表现相当甚至更好。然而，VIIRS 模型在夏季的表现不如 MODIS 模型。本研究揭示了 MODIS 和 VIIRS AOD 在模拟地面 PM 浓度方面的优缺点，促进了基于卫星的 PM 估算研究。
