Publication: Advanced machine learning algorithm to predict the implication of climate change on groundwater level for protecting aquifer from depletion
Date
2024
Authors
Ahmed Osman A.I.
Latif S.D.
Wee Boo K.B.
Ahmed A.N.
Huang Y.F.
El-Shafie A.
Journal Title
Journal ISSN
Volume Title
Publisher
Elsevier B.V.
Abstract
Due to the impact of climate change, the groundwater level (GWL) has been declining recently in Malaysia, which is essential to protect the groundwater aquifer against depletion. Therefore, the current study aimed to propose an accurate GWL prediction model using advanced machine learning (ML) algorithms in five populated towns, namely Jenderam, Bangi, Beranang, Kajang, and Paya Indah Wetland which are in Selangor, Malaysia. The models developed, used 11 months of previously recorded daily meteorological data of rainfall, temperature, evaporation, and GWL data from one selected well for each town, to predict 1-day, 3-day, and 5-day horizons GWL. For all five locations, four ML algorithms have been trained, tested, and then evaluated: long short-term memory (LSTM), extreme gradient boost (XGBoost), Artificial Neural Network (ANN), and Support Vector Regression (SVR). Further, the best model among the four proposed models is used to predict daily GWL from January 2030 to December 2039 using projected rainfall and temperature data extracted from the Intercomparison Project Phase 5 (CMIP5) climate model. Applying the same 3 different input combinations for the models, the results showed that all the locations, including the GWL time-series data, improved the prediction accuracy significantly in all four models. Using testing data for 1-day ahead GWL prediction at Paya Indah Wetland as the best-performing location, XGBoost achieved the highest prediction performance with root mean squared error (RMSE) of 0.026 followed by LSTM, ANN, and SVR with RMSE of 0.027, 0.050, and 0.085 respectively. Ultimately, the results obtained in this study serve as a great benchmark for future GWL prediction using LSTM and XGBoost algorithm and give an insight into the influence of climate change on future GWL. Further, the findings can help local water resource managers draft resealable accuracy water resource plans in the state of Selangor for the next decade. ? 2024 Elsevier B.V.
Description
Keywords
Malaysia , Aquifers , Climate change , Climate models , Forecasting , Groundwater resources , Hydrogeology , Learning algorithms , Learning systems , Location , Long short-term memory , Mean square error , Rain , Support vector machines , 'current , Ground water level , Groundwater aquifer , Machine learning algorithms , Machine-learning , Malaysia , Prediction modelling , Root mean squared errors , Support vector regressions , Waters resources , aquifer , climate change , groundwater resource , machine learning , prediction , regression analysis , Wetlands