Comparative assessment of hybrid machine learning models for accurate groundwater level prediction in a data-limited aquifer
Description
This dataset contains hydro-climatic and groundwater level (GWL) observations for the Astaneh-Kuchesfahan aquifer, Iran, covering a 256-month period (2002–2020). The data includes monthly records of GWL, precipitation, temperature, evapotranspiration, and water depth from two observation wells (C1 and C2). This dataset was compiled to evaluate the performance of various hybrid machine learning models—including ANN-SFOA, LSTM-TROA, ANFIS-EGOA, RF-Wavelet, and GAN-RF—in predicting groundwater levels. It is particularly valuable for research focused on time-series analysis, hydrogeological modeling, and the application of machine learning in water resource management. By providing multi-scale patterns, this dataset enables the study of predictive model generalization and resistance to overfitting. It supports the research findings regarding the superiority of the RF-Wavelet model in denoising and enhancing predictive accuracy. Users are encouraged to cite the associated publication when utilizing this data for further scientific investigations.