Evaluation of three land surface temperature products from Landsat series using in-situ measurements

Three operational long-term land surface temperature (LST) products from Landsat series are available to the community until now, i.e., U.S. Geological Survey (USGS) LST, Instituto Português do Mar e da Atmosfera (IPMA) LST and China University of Geosciences (CUG) LST. A comprehensive assessment of these LST products is essential for their subsequent applications in energy, water and carbon cycle modeling. In this study, an evaluation of these three Landsat LST products was performed using in-situ LST measurements from five networks (SURFRAD, ARM, HiWATER, BRSN and NDBC) for the period of 2009-2019. Results reveal that the overall accuracies of CUG LST with bias (RMSE) of 0.54 K (2.19 K) and IPMA LST with bias (RMSE) of 0.59 K (2.34 K) are marginally superior to USGS LST with bias (RMSE) of 0.96 K (2.51 K). The RMSE of USGS LST is about 0.3 K less than IPMA/CUG LST at water surface sites and is about 0.4 K higher than IPMA/CUG LST at cropland and shrubland sites. As for tundra, grassland and forest sites, the RMSE of three Landsat LST products are similar, the RMSE difference among three Landsat LST products is