Evaluating lake surface water temperature as an indicator of ice phenology in Selin Co, Tibetan Plateau
Abstract. Satellite-derived lake surface water temperature (LSWT) is increasingly used to determine lake ice phenology, commonly assuming that freezing and melting occur when LSWT crosses 0 °C. Here, we tested this assumption in Selin Co, a large saline lake on the central Tibetan Plateau, using MODIS LSWT (2001–2025), ERA5-Land meteorological data, optical ice phenology observations spanning 20 ice seasons, and in situ water-temperature and conductivity profiles. Within the perennial lake area, nighttime LSWT increased by 0.55 °C per decade, while daytime LSWT showed no significant trend, resulting in a 0.39 °C per decade decline in diurnal range. Freezing began 3–5 weeks after air temperature fell below 0 °C, when whole-lake mean LSWT remained near 4 °C, and complete freeze-up followed ~46 days later. Because the lake’s practical salinity of 8 lowers the temperature of maximum density to 2.2 °C, the water column continued convective cooling below 4 °C before freeze-up. Thus, complete freeze-up was governed primarily by lake-water cooling rather than the onset of atmospheric cooling, whereas ice melt began weeks before air temperature became positive and was associated with winter warming and solar radiation. A 0 °C LSWT threshold produced 2–5-week biases for three of four ice events and reproduced only complete freeze-up well after threshold correction. These results highlight the roles of lake depth, salinity, and heat storage in satellite-based lake ice phenology retrievals for large saline lakes.