Modeling the fluctuations of groundwater level by employing ensemble deep learning techniques

This study proposes two techniques: Deep Learning (DL) and Ensemble (EDL) to predict groundwater level (GWL) for five wells in Malaysia. Two scenarios were proposed, scenario-1 (S1): GWL from 4 was used as inputs the fifth well scenario-2 (S2): time series with lag up 20 days all wells. The results S1 prove that ensemble EDL generally performs superior DL estimation of each station using data remaining four except Paya Indah Wetland which method provide better estimates compared EDL. Regarding S2, also exhibits performance predicting daily stations model. Implementing decreased RMSE, NAE RRMSE by 11.6%, 27.3% 22.3% increased R, Spearman rho Kendall tau 0.4%, 1.1% 3.5%, respectively. Moreover, S2 shows a high precision within less lag, ranging between 2 DL. Therefore, model has potential managing sustainability