Laalaoui et al. (2026) Hybrid Ensemble Learning for TWSA Prediction in Water-Stressed Regions: A Case Study from Casablanca–Settat Region, Morocco

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: MDPI (MDPI AG)
• Year: 2026
• Authors: Youssef Laalaoui, Naïma El Assaoui, Oumaima Ouahine, Thanh Thi Nguyen, Ahmed Mohamed Ahmed Elsayed Mostafa Saqr
• DOI: 10.3390/hydrology13020053

Research Groups

• Not explicitly mentioned in the provided text (typically involves regional water management departments and environmental research units in Morocco).

Short Summary

This study develops a hybrid machine learning ensemble to estimate Terrestrial Water Storage Anomalies (TWSA) in Morocco’s Casablanca–Settat region by combining GRACE satellite data with environmental indicators. The framework achieves high predictive accuracy ($R^2 = 0.97$), providing a detailed spatial tool for monitoring groundwater depletion and informing regional water management.

Objective

• To develop and validate a hybrid machine learning framework for the high-resolution estimation of Terrestrial Water Storage Anomalies (TWSA) as a proxy for groundwater changes in a water-stressed region.

Study Configuration

• Spatial Scale: Regional (Casablanca–Settat region, Morocco).
• Temporal Scale: Spans the duration of the GRACE and GRACE Follow-On (GRACE-FO) satellite missions (2002–present).

Methodology and Data

• Models used: A stacked ensemble approach featuring six base learners: Huber regressor, Automatic Relevance Determination Regression (ARDR), Kernel Ridge, Long Short-Term Memory (LSTM), K-Nearest Neighbors (KNN), and Gradient Boosting. A Random Forest meta-learner was used to aggregate predictions.
• Data sources: Satellite-based observations from GRACE and GRACE-FO; environmental indicators including rainfall, evapotranspiration, and land use data.

Main Results

• Model Performance: The ensemble model achieved a coefficient of determination ($R^2$) of 0.97, a Root Mean Square Error (RMSE) of 0.13, and a Mean Absolute Error (MAE) of 0.108.
• Comparative Advantage: The hybrid framework significantly outperformed all single-model baselines in terms of accuracy and stability.
• Spatial Insights: The analysis identified clear correlations between groundwater depletion and specific land cover types/usage patterns, highlighting areas of high stress.

Contributions

• Methodological Innovation: Demonstrates the effectiveness of a multi-model stacking approach for downscaling or estimating coarse satellite gravimetry data at a regional scale.
• Practical Application: Provides a scalable tool for targeted aquifer recharge, irrigation management, and drought response planning in semi-arid environments.
• Policy Alignment: Directly supports Morocco’s national water sustainability initiatives by providing high-resolution data for groundwater-related storage changes.

Funding

• Not specified in the provided text.

Citation

@article{Laalaoui2026Hybrid,
  author = {Laalaoui, Youssef and Assaoui, Naïma El and Ouahine, Oumaima and Nguyen, Thanh Thi and Saqr, Ahmed Mohamed Ahmed Elsayed Mostafa},
  title = {Hybrid Ensemble Learning for TWSA Prediction in Water-Stressed Regions: A Case Study from Casablanca–Settat Region, Morocco},
  journal = {MDPI (MDPI AG)},
  year = {2026},
  doi = {10.3390/hydrology13020053},
  url = {https://doi.org/10.3390/hydrology13020053}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)