<p>Using the WRF-Lake model, this study tried to improve the prediction of lake surface water temperature (LSWT) and how it affects the atmospheric conditions over the hypersaline Urmia Lake (UL). Several changes were made to the lake model to fix its problems with showing how the temperature changes in UL. Because the salinity was abnormally high in this lake, changes had to be made to the formulas for water density, freezing point, and saturation vapor pressure. A dynamic data assimilation method was also used to use in-situ observations to keep the lake surface temperature up to date. For the cold season of 2016–2017, model simulations were run and the model’s performance was assessed using field observations, reanalysis data, and satellite retrievals. The results indicated significant discrepancies between the results from model with default configuration and in-situ observations. The model underestimated LSWT and had cold biases, particularly during nighttime. After trying out different combinations, the SLake model, which included all of the changes, worked the best. It enhanced the accuracy of LSWT prediction and eliminated thermal stratification patterns. The model was still not able to reproduce the lake’s complicated thermal behavior; hence, a dynamic LSWT assimilation method was employed that used in-situ data to keep the lake surface temperature up to date. This new method cut down the cold bias in LSWT by a large amount and improved the simulation of near-surface air temperatures significantly. The dynamic LSWT updates also improved simulations of lake evaporation, with the SLake_LSWT configuration outperformed other models and GLEAM predictions. The study shows importance of inclusion of physical properties of hypersaline lakes in to numerical models to have realistic results.</p>