Applicability of MODIS lake surface temperature depends on season and lake size on the Tibetan Plateau

Lake surface temperature is a key parameter in understanding the variability of lake thermal condition and evaporation. MODIS-derived LST is widely used as a reference for lake-model validations and process studies in data-scarce regions. In this study, the accuracy and limitation of MODIS LST were examined on the Tibetan Plateau, where there are thousands of lakes. It is found that agreement between MODIS LST and in-situ subsurface (~1 m depth) temperature collected at six large lakes depends on the thermal phases. During lake turnover period (nighttime or from October to freeze-up date), the sink of surface water causes mixed with subsurface water. The MODIS LST was consistent with the in-situ data, indicating its high accuracy. During stratification period (from May to September), the lakes were thermally stratified due to intense solar heating and high salinity in some lakes; the daytime MODIS LST is systematically higher than the in-situ subsurface temperature, indicating it is credible. However, the MODIS LST has two limitations in this region. First, nighttime retrievals during monsoon season have considerable cold biases in monsoon-controlled region. This can be associated with shallow clouds or fog near the lake surface that occur frequently at night but are not well detected by MODIS. Second, the retrievals for narrow and small lakes have warm (cold) biases in the daytime (at night), perhaps due to proximity effect of mountains and land. The two situations are common across the Tibetan Plateau and thus severely restrict the applications of MODIS LST in lake studies.