On the predictability of lake surface temperature using air temperature in a changing climate: A case study for Lake Tahoe (U.S.A.)

Can we predict long-term trends of lake surface temperature based on air temperature alone? We explore this question by analyzing the performance of a hybrid model (air2water) as a predictive tool for defining scenarios of lake surface temperature in the framework of climate change studies. Employing Lake Tahoe (U.S.A.) as a case study, we apply the model using different air temperature datasets (in situ measurements, gridded observations, and downscaled General Circulation Models). Through a data-driven calibration of the model parameters based on surface water temperature records, we show that air2water provides good performance (root mean square error ∼ 0.5°C, on a monthly scale) regardless of the input dataset. The model is able to accurately capture the historical long-term trend and interannual fluctuations over decades (from 1969 to present), using only 7 yr of monthly measurements of surface water temperature for calibration. Additionally, when used to predict future surface water temperature of the lake, air2water produces the same projections irrespective of the air temperature dataset used to drive the model. This is certainly desirable, but not immediately expected when using a relatively simple model. Overall, the results suggest the high potential and robustness of air2water as a predictive tool for climate change assessment. Lake surface temperature warming of up to 1.1°C (RCP 4.5) and 2.9°C (RCP 8.5) was simulated at the end of the 21st century during summer months in Lake Tahoe. Such a scenario, if realized, would lead to serious consequences on lake water chemistry, primary productivity, plankton community structure, and nutrient cycling.