Groundwater Level Time Series Prediction Using Group Method of Data Handling and Deep Convolutional Neural Network With Recurrent Neural Networks

This study compared the performances of the GDMH and CNN-RNN algorithms for modeling the underground levels in 2 Erzurum wells. For model input combinations, lagged values exceeding 95% were selected from the PACF charts. According to the analysis results, the GDMH model performed more effectively in both wells. As a result of the analysis, both models produce promising results in GWL time series estimation. The study’s outputs are essential for effective water management for water resource managers, decision-makers, and planners.