Abstract: Lake ice phenology affects the lake environment and lake-air interaction through the ice-covered period. In order to investigate the trends and influencing factors of lake ice phenology in freshwater lakes on the Tibetan Plateau, the FLake model was employed after improvement. The model validation used MODIS surface temperature data based reconstruction of the ice-on date time series in 1979—2021 for Lakes Ngoring and Gyaring. Sensitivity experiments were conducted using the improved model. The results showed that the present FLake-SELF model improved the accuracy of the ice-on date calculation by 35%, and the reconstructed sequences showed that the ice-on dates of Lakes Ngoring and Gyaring have been delayed by 4.5 d/(10 a) and 3.8 d/(10 a), respectively. Increasing air temperature, decreasing wind speed, and increasing vapor pressure are the most important meteorological factors leading to the delay of ice-on date, and interannual variability in these factors were the most important reasons for the interannual fluctuations of ice-on dates in the two lakes. In the same climate conditions, the different lake depth was the reason for the differences in ice-on date and their delay rate between the two lakes. The deeper the lake was, the greater was the interannual fluctuation in ice-on date, the greater was the rate of change of ice-on date, and the more sensitive was the ice-on date to climate change.