Groundwater level prediction using machine learning algorithms in a drought-prone area
Quoc Bao Pham,Manish Kumar,Fabio Di Nunno,Ahmed Elbeltagi,Francesco Granata,Abu Reza Md. Towfiqul Islam,Swapan Talukdar,X. Cuong Nguyen,Ali Najah Ahmed,Duong Tran Anh
DOI: https://doi.org/10.1007/s00521-022-07009-7
2022-03-03
Neural Computing and Applications
Abstract:Groundwater resources (GWR) play a crucial role in agricultural crop production, daily life, and economic progress. Therefore, accurate prediction of groundwater (GW) level will aid in the sustainable management of GWR. A comparative study was conducted to evaluate the performance of seven different ML models, such as random tree (RT), random forest (RF), decision stump, M5P, support vector machine (SVM), locally weighted linear regression (LWLR), and reduce error pruning tree (REP Tree) for GW level (GWL) prediction. The long-term prediction was conducted using historical GWL, mean temperature, rainfall, and relative humidity datasets for the period 1981–2017 obtained from two wells in the northwestern region of Bangladesh. The whole dataset was divided into training (1981–2008) and testing (2008–2017) datasets. The output of the seven proposed models was evaluated using the root mean square error (RMSE), mean absolute error (MAE), relative absolute error (RAE), root relative squared error (RRSE), correlation coefficient (CC), and Taylor diagram. The results revealed that the Bagging-RT and Bagging-RF models outperformed other ML models. The Bagging-RT models can effectively improve prediction precision as compared to other models with RMSE of 0.60 m, MAE of 0.45 m, RAE of 27.47%, RRSE of 30.79%, and CC of 0.96 for Rajshahi and RMSE of 0.26 m, MAE of 0.18 m, RAE of 19.87%, RRSE of 24.17%, and 0.97 for Rangpur during training, and RMSE of 0.60 m, MAE of 0.40 m, RAE of 24.25%, RRSE of 29.99%, and CC of 0.96 for Rajshahi and RMSE of 0.38 m, MAE of 0.24 m, RAE of 23.55%, RRSE of 31.77%, and CC of 0.95 for Rangpur during testing stages, respectively. Our study offers an effective and practical approach to the forecast of GWL that could help to formulate policies for sustainable GWR management.
computer science, artificial intelligence