<p>Accurate prediction of groundwater level (GWL) and its associated drought is crucial for sustainable water resources management, particularly in arid and semi-arid regions. In this study, a hybrid modeling framework was developed by integrating advanced data preprocessing techniques with artificial intelligence and deep learning (DL) models to predict GWL and groundwater drought (GWD) in the Nahavand aquifer, western Iran. Despite the critical role of the Nahavand region—one of the main tributary basins of the Karkheh watershed and a vital source of agricultural and domestic water supply—no comprehensive investigation has yet been conducted to assess its water resources and drought dynamics. This research gap is particularly concerning given the accelerating rate of groundwater extraction from the aquifer. Two signal decomposition methods including wavelet transform (WT) and complete ensemble empirical mode decomposition (CEEMD) were employed to decompose the time series into sub-signals, which were then used as inputs to the long short-term memory (LSTM) and group method of data handling (GMDH) models. Hybrid models (W-LSTM, W-GMDH, CEEMD-LSTM, and CEEMD-GMDH) were constructed and evaluated using statistical performance indicators. The results revealed that the W-GMDH hybrid model outperformed the others, achieving a coefficient of determination (R²) of 0.954 and a root mean square error (RMSE) of 0.027&#xa0;m. The GWL forecasts generated by this model were used to compute the Groundwater Resource Index (GRI), indicating the occurrence of severe and prolonged droughts in the study area. Moreover, predictions for the first half of the 2024–2025 water year suggest continued GWD in the region. These findings highlight that combining signal decomposition techniques with AI-based models provides an efficient and reliable approach for groundwater prediction and drought assessment.</p>