Medium and short term prediction of water temperature in deep water lake-reservoir based on data assimilation

The distribution and evolution of water temperature in deep-water lakes and reservoirs affect water movement, biochemical reactions, and the metabolic processes of aquatic organisms. Forecasting water temperature changes is necessary for lake and reservoir water quality management and ecological environment safety. In this study a lake and reservoir water temperature data assimilation system is constructed on the basis of the ensemble Kalman filter algorithm and the CE-QUAL-W2 model, which is able to comprehensively consider model parameters, boundary conditions, and the uncertainty of observation data. The system is applied to Daheiting Reservoir for 1-10 days of medium and short term water temperature forecasting using reservoir operation data and meteorological data as forecast conditions. The results show that when the number of sets is 100, the simulation error and the observation error are 10% and 1%, the assimilation system can consider higher calculation efficiency and simulation accuracy. Simultaneously correcting the model parameters and state variables can increase the water temperature simulation accuracy of the data assimilation system at different water depths by 41.2%-68.8% compared to the results without data assimilation. As the forecast period was extended from 1 to 10 days, the forecast error of each water depth increased from 0.22-0.35 °C to 0.77-1.09 °C. Regardless of whether the reservoir is in the stratified or mixed period, the data assimilation system can maintain high accuracy under the internal and external factors such as meteorological conditions and reservoir scheduling during the forecast period. The high-precision medium and short term water temperature forecast method can provide theoretical and technical support for lake and reservoir water supply and ecological security. © 2022 China Water Power Press. All rights reserved.