Search

Study regions: The study area encompasses two distinct sub-basins within the High Atlas Mountains: Oukaimeden in the Rheraya and Tichki in the Mgoun Valley. Study focus: The research integrates remote sensing data, particularly the Normalized-Difference Snow Index (NDSI) from the MODIS Sensor, with machine learning (ML) and deep learning (DL) models to predict daily snow depth (DSD) at a local scale. The models evaluated include two ML approaches: Support Vector Regression (SVR) and eXtreme Gradient Boosting (XGBoost) and four DL models: 1-Dimensional Convolutional Neural Network (1D-CNN), Long Short-Term Memory Networks (LSTM), Gated Recurrent Unit (GRU), and Bi-directional Long Short-Term Memory Network (Bi-LSTM). The dataset was processed and normalized for optimal performance, and hyperparameters were fine-tuned using a randomized search method. New hydrological insights for the region: The Results highlight the efficacy of AI-based approaches for snow depth prediction, with SVR achieving the best performance (Root Mean Square Error of 2–5 cm and an average coefficient of determination of 0.97). This study reveals that incorporating lag times of snow depth data significantly enhances predictive accuracy. These findings underscore the potential of integrating remote sensing with AI techniques to improve hydrological modeling and water resource planning in data-scarce regions like the Atlas Mountains.

Study region: The upstream part of the Essaouira basin, a data-scare region in Morocco, Northwestern Africa. Study focus: The scarcity of hydro-climate data is a significant challenge found in several regions worldwide, where qualitative and quantitative water resource information remains limited. Estimating and predicting groundwater levels (GWL) in such areas is a significant challenge in producing knowledge for effective water resource management. To address this issue, the present study aimed to use the Soil and Water Assessment Tool (SWAT) model in conjunction with downscaled total water storage (TWS) data (9 km) obtained from Gravity Recovery And Climate Experiment (GRACE) and machine learning techniques, specifically random forest (RF) and support vector machine (SVM), to estimate and predict the variation in GWL. New hydrological insights for the region: This study constitutes a first of its kind in the study area; the SWAT model was set up for 10 years, with a warm-up period from 2000 to 2001, calibration from 2002 to 2007, and validation from 2008 to 2010. The statistical indices (Coefficient of Determination (R²) ≥ 0.73, R² ≥ 0.78, Nash–Sutcliffe model efficiency coefficient (NSE) ≥ 0.67, NSE ≥ 0.80 respectively for calibration and validation) highlight a significant correlation, implying the model's capability to faithfully reproduce the streamflow. The downscaled TWS demonstrates an impressive ability to identify and monitor fluctuations in GWL. Using machine learning algorithms (RF and SVR), the prediction of GWL yielded satisfactory results, NSE = 0.78 and root mean square error (RMSE) = 0.33, NSE = 0.51 and RMSE = 0.49 for the RF and SVR, respectively. Despite some limitations, our approach provided promising results in GWL prediction, with the possibility of expanding to other data-scarce regions.