MODIS地表温度产品在青藏高原连续多年冻土区的适用性分析

利用自动气象站观测的长波辐射计算得到的地表温度对MODIS地表温度(LST)产品在青藏高原中部连续多年冻土区的精度进行验证, 并利用具有较高空间分辨率的Landsat 5 TM和Landsat 7 ETM+反演的地表温度与MODIS LST产品进行了对比分析. 结果表明: 白天MODIS LST产品的平均绝对误差(MAE)和均方根误差(RMSE)分别约为3.42~4.41 ℃和4.41~5.29 ℃, 夜晚MODIS产品MAE和RMSE分别为2.15~2.90 ℃和3.05~3.78 ℃, 精度高于白天; MODIS LST与TM、ETM+反演的地表温度一致性较好, 相关系数分别达到0.85和0.95. 说明MODIS LST产品在连续多年冻土区的适用性较高, 是研究多年冻土地表热状况的一个非常好的数据源. 而且, 不同空间尺度的遥感数据之间一致性较好, 可考虑将多源遥感数据应用于多年冻土热状况监测研究.

Assessing the applicability of MODIS land surface temperature products in continuous permafrost regions in the central Tibetan Plateau

Land surface temperature (LST) is a crucial parameter of surface energy budget and controls the thermal regime of the active layer in permafrost regions. However, there are limited observed datasets available for the Tibetan Plateau, with greater bias in the LST products from remote sensing data. In this study, the LST calculated with the upward and downward long-wave radiation data from the three automatic weather stations were compared with the MODIS LST products. Meanwhile, the LST retrieved from Landsat 5 TM and Landsat 7 ETM+ was also compared with MODIS LST to explain the consistency between remote sensing data with different resolutions. In daytime, the mean absolute error (MAE) and the root-mean-square error (RMSE) of MODIS LST products were about 3.42~4.41 ℃ and 4.41~5.29 ℃, respectively. In nighttime, the MAE and the RMSE of MODIS LST products were about 2.15~2.9 ℃ and 3.05~3.78 ℃, respectively, indicating that the nighttime MODIS LST have a better accuracy. Both of the LSTs retrieved from TM and ETM+ have a good consistency with MODIS LST, with the correlation coefficients of greater than 0.85 and 0.95 for TM and ETM+, respectively. MODIS LST products have a high applicability in continuous permafrost regions in the Tibetan Plateau, which is a potential dataset for monitoring permafrost thermal regime. The consistency between Landsat TM/ETM+ LST and MODIS LST is also high. The results can be used to monitor and model permafrost with remote sensing data of multi-sources