This research aims to improve the prediction of groundwater levels by integrating numerical and machine learning models.
Established a multi-model framework combining MODFLOW, ANN, and LSTM.
Conducted calibration and validation using monitoring well data in Eastern Beijing.
Compared prediction accuracy through multiple scenarios for LSTM and numerical models.
Numerical model reproduced groundwater level variations with relative errors within 1.5 m for over 91.67% monitoring wells.
ANN model outperformed LSTM with NSEs greater than 0.92 and RMSEs smaller than 1.51 m during training and validation.
RMSE values for LSTM and numerical models ranged from 0.054 to 0.187 and 0.012 to 0.121, respectively, indicating varying levels of prediction accuracy.