Study on the water level forecast in Chaohu Lake using the ensemble learning LightGBM model

Accurately predicting and forecasting the changes in lake water levels is of great significance for regional water security. In order to achieve rapid and accurate prediction of lake water level fluctuations driven by basinal inflow, this paper takes Chaohu Lake as an example and adopts the ensemble learning LightGBM algorithm to construct a daily average water level prediction model for Chaohu Lake driven by precipitation data. The model takes the daily average rainfall of the basin in one historical week and the historical daily average water level of the simulated hydrological station in one week as input, and is trained on data from 2019 to 2022 with 2023 for validation. The validation and sensitivity analysis results show that the constructed water level prediction model based on LightGBM has good robustness and accuracy, with root mean square errors of 0.03 m and 0.04 m for Zhongmiao and Chaohu sluice stations water level predictions, and Nash efficiency coefficients of 0.98 and 0.94 respectively. This model provides a reference for rapid prediction and analysis of lake water levels.