A Hybrid Machine Learning and Deep Learning Model for Time-Series Groundwater Level Prediction

Groundwater level forecasting is important for sustainable water resource management, especially in areas experiencing variable water availability. This research investigates the use of machine learning and deep learning methods to predict groundwater levels from historical data obtained from monitoring stations.The input data include groundwater measurements, temporal characteristics, and station information, which were preprocessed with interpolation and normalization techniques to ensure data quality.Different predictive models were deployed, such as Random Forest, Support Vector Machines (SVM), Long Short-Term Memory (LSTM) networks, and Bidirectional LSTM. In addition, a hybrid model involving Random Forest with LSTM was built for leveraging both structured and sequential learning features. Model accuracy was approximated through Mean Squared Error (MSE) and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> Score, with Random Forest being the most accurate in machine learning models (MSE: 0.0276, R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>: 0.9946). Bidirectional LSTM was outperforming the simple LSTM with an MSE of 0.0848 and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> Score of 0.9835. The hybrid model had better predictive ability with an MSE of 0.0410 and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> Score of 0.9920.The findings indicate the efficiency of hybrid methods in improving prediction accuracy, with the implication that integrating machine learning and deep learning methods can enhance groundwater forecasting. Future research will aim to incorporate remote sensing data, utilize federated learning for decentralized processing, and investigate transformer-based architectures for additional accuracy enhancements.