美国地质调查局(USGS) 2020年12月正式发布了基于Landsat热红外光谱数据生产的2级地表温度产品(Landsat Collection 2 Level-2 surface temperature, LC2L2ST), 但目前还鲜有该地表温度产品的相关研究报道。 由于美国地质调查局已宣布自2022年开始将不提供除该产品之外的其他地表温度数据, 因此有必要对该产品进行适时的评估。 在同类遥感卫星产品中, MODIS地表温度产品的质量最被广大用户认可, 应用也最为广泛, 因此首次将Landsat新型热红外地表温度产品与MODIS地表温度产品进行交互对比, 以评估新产品与MODIS地表温度产品的一致性。 分别选取了我国的不同地区(福州、 太湖、 银川、 敦煌)作为试验区, 以20对同日过空的LC2L2ST与MODIS地表温度影像为数据源进行交互对比。 影像涵盖植被、 水体、 建筑、 荒漠等地物以及不同的季节。 在试验区影像上选取一系列均质样区(ROI), 通过各样区的地表温度均值散点进行拟合回归分析, 研究二者地表温度间的差异及定量关系, 并提出彼此转换的模型。 结果表明, Landsat地表温度新产品与MODIS地表温度产品具有很高的相关性, 4个试验区的决定系数(R2)皆大于0.98, 集成4个试验区的总R2也接近0.98; 但LC2L2ST比MODIS的地表温度平均高0.90 ℃(RMSE=2.29 ℃)。 分析发现, 二者地表温度数据间的差异与其在空间分辨率、 观测角度、 地物类型和季节的不同有关。 从不同地物和季节来看, LC2L2ST在晚秋和冬季略低于MODIS的地表温度, 而在夏季极端高温的城镇、 荒漠地区则明显高于MODIS的地表温度, 且偏差可近7 ℃。 总的看来, Landsat新型地表温度产品与MODIS同类产品的相关性显著, 但在夏季的城市和沙漠地区的差异较大, 因此, LC2L2ST新产品在夏季高温季节的适用性仍有待进一步基于地面实测温度的验证。 鉴于两种地表温度数据产品仍存在着一定差距, 因此二者如要协同使用, 需要进行转换。 本研究基于4个试验区的560个ROI样区构建了二者地表温度间的转换方程, 并通过验证发现, 经转换后的两数据差异性得到大幅缩小。 因此, 必要的数据转换有利于二者数据的协同使用, 可为长时间序列的地表温度变化监测提供连续的遥感数据。

Landsat Collection 2 Level-2 Surface Temperature (LC2L2ST) was formally released in December 2020 by the U.S. Geological Survey (USGS). However, there are few reports on this new land surface temperature (LST) product. As this product will be the only LST data provided by the USGS starting in 2022, it is necessary to evaluate the product timely. Among various satellite LST products, the quality of the MODIS LST product is well recognized, and widely used. Therefore, this paper, for the first time, performed a cross-comparison between the new Landsat LST product and the MODIS LST product to examine the quality of the new product. Different regions in China (Fuzhou, Taihu, Yinchuan and Dunhuang) were selected as the test areas, and 20 pairs of LC2L2ST and MODIS LST synchronous images were used for the comparison. The images cover different land types, such as vegetation, water, town and deserts across different seasons. A total of 560 homogeneous regions of interest (ROI) were selected from the images of the test areas. The regression analysis was carried out to examine the fit of the ROIs and the quantitative relationship between the two LST products. The conversion model between them was also developed. The results showed that the new LC2L2ST product is highly correlated with the MODIS LST product. Each of the four test areas can achieve a coefficient of determination (R2) greater than 0.98. Integrating the 560 samples from the four test areas also obtain an R2 close to 0.98. Nevertheless, differences between the two products have also been founded. The LC2L2ST is 0.90℃ averagely higher than the MODIS LST (RMSE = 2.29 ℃). However, LC2L2ST can be slightly lower than MODIS LST in late fall and winter seasons but significantly higher than extremely hot summer seasons with a bias close to 7 ℃. The analysis revealed that the differences were related to spatial resolution, sensor viewing angles, land cover types and seasons. In general, the new LC2L2ST product strongly correlates with the MODIS LST, but significant differences were also observed in the summer months. Therefore, the new Landsat LST product must be further tested with in-situ measured LST data. Due to the differences in this paper, the two LST data products need to be converted when they must be collaboratively used. This study developed the conversion equation between the two LSTs based on the 560 ROIs. The verification found that the differences between the two data after conversion were greatly reduced. It is conducive to the cooperative use of the two LST data and providing continuous remote sensing data for long-term LST monitoring.