Moumane et al. (2026) Desertification monitoring in arid oasis environment using Google Earth Engine, machine learning, and field-based hydrogeological assessment

Identification

• Journal: Scientific Reports
• Year: 2026
• Date: 2026-02-21
• Authors: Adil Moumane, Youssef Azougarh, Abdelhaq Ait Enajar, Wafa Saleh Alkhuraiji, Ismail Bahdou, Jamal Al Karkouri, Faten Nahas, N. Ya. Rebouh, Youssef M. Youssef
• DOI: 10.1038/s41598-026-41216-9

Research Groups

• Department of Geography, Faculty of Humanities and Social Sciences, Ibn Tofail University, Kenitra, Morocco
• Materials and Environment Laboratory (LME), Faculty of Sciences, Ibn Zohr University, Agadir, Morocco
• Regional Center for Educational and Training Professions, Marrakech, Morocco
• Department of Geography and Environmental Sustainability, College of Humanities and Social Sciences, Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia
• Department of Geography, Abdelmalek Essaâdi University, Morocco
• Geography Department, College of Humanities and Social Sciences, King Saud University, Riyadh, Saudi Arabia
• Department of Environmental Management, Institute of Environmental Engineering, RUDN University, Moscow, Russia
• Geological and Geophysical Engineering Department, Faculty of Petroleum and Mining Engineering, Suez University, Suez, Egypt

Short Summary

This study assessed desertification dynamics in the Ternata Oasis (southeastern Morocco) over four decades (1984–2024) by integrating Google Earth Engine-based machine learning, remote sensing, hydrogeological fieldwork, and socioeconomic surveys. It revealed a significant decline in oasis vegetation, groundwater depletion, and salinization, driven by climate variability and anthropogenic overexploitation, with the Gradient Tree Boosting model achieving 87.2% accuracy for desertification mapping.

Objective

• To monitor long-term vegetation and desertification trends (1984–2024) using NDVI, MSAVI, EVI, and albedo indices from the full Landsat archive.
• To monitor the spatial and temporal progression of desertification from 1984 to 2024 by leveraging the full Landsat archive and applying Gradient Tree Boosting (GTB) within the Google Earth Engine environment.
• To validate satellite-based outputs through in-situ groundwater depth, salinity, and hydrochemical analysis.
• To integrate local socio-environmental knowledge to contextualize desertification processes.

Study Configuration

• Spatial Scale: Ternata Oasis, Middle Draa Valley, southeastern Morocco. Remote sensing data at 30 meters spatial resolution (Landsat). Spatial autocorrelation analysis used a 200 meters fishnet grid. Field measurements were point-based across 30 wells/farms.
• Temporal Scale: Multi-decadal monitoring from 1984 to 2024 for remote sensing and hydrological data. Field measurements were conducted in November 2022. Socioeconomic surveys were carried out from November 2022 to August 2025.

Methodology and Data

• Models used: Gradient Tree Boosting (GTB), Random Forest (RF), Classification and Regression Trees (CART) for land cover classification. Spectral indices: Normalized Difference Vegetation Index (NDVI), Modified Soil Adjusted Vegetation Index (MSAVI), Enhanced Vegetation Index (EVI), Broadband Albedo (α), Normalized Difference Water Index (NDWI). Spatial autocorrelation analysis: Global Moran’s I.
• Data sources:
  • Satellite: Landsat 5 TM, Landsat 7 ETM+, Landsat 8 OLI (surface reflectance imagery processed in Google Earth Engine). ASTER Global DEM (30 meters). High-resolution Google Earth imagery for validation.
  • Observation/Field: In-situ measurements from 30 wells (November 2022) including groundwater depth (meters below ground level), electrical conductivity (µS/cm), pH, salinity (parts per thousand), Total Dissolved Solids (g/L), chloride (mg/L), sulfate (mg/L), and nitrate (mg/L). Soil electrical conductivity (dS/m). Socioeconomic surveys via semi-structured interviews with 120 households/30 follow-up interviews (November 2022 – August 2025).
  • Administrative: Annual irrigation water releases (million cubic meters, Mm³) from El Mansour Eddahbi Dam (ORMVA Ouarzazate records, 1984–2022).

Main Results

• Vegetation Dynamics (1984–2024): NDVI, MSAVI, and EVI showed peaks in the late 1990s and around 2015, followed by a clear and sustained decline, indicating increasing vegetation stress. Mean surface albedo exhibited an inverse trend, increasing to 0.207 by 2024, reflecting surface brightening due to vegetation decline and exposed soil. A strong negative Pearson correlation (r = –0.82) was found between NDVI and albedo.
• Desertification Progression: The Ternata Oasis experienced three distinct desertification phases: Early (1984–1988, desert area expanded by ~1200 hectares), Sustained (2000–2006, desert cover surged by 3336 hectares, oasis area plummeted by 3195 hectares), and Recent Acceleration (2020–2024, desert area jumped by 2612 hectares, oasis extent fell by 2909 hectares, marking the fastest rate of change). The year 2021 was a "tipping point" where desert cover (2947 hectares) first exceeded oasis area (2721 hectares).
• Machine Learning Performance: The Gradient Tree Boosting (GTB) classifier achieved the highest overall accuracy of 87.2% (Kappa = 0.81) for desertification mapping in 2022, outperforming Random Forest (85.0%) and CART (82.0%).
• Hydrological Stress: Annual dam inflows to El Mansour Eddahbi Dam dramatically declined post-2015, reaching extreme lows (e.g., 77.8 Mm³ in 2022). The positive correlation between irrigation volumes and oasis area weakened significantly from r = 0.56 (1984–2004) to r = 0.28 (2005–2024), indicating a loss of ecological resilience. NDWI-derived maps showed a dramatic contraction of the reservoir surface area by 2024.
• Groundwater Depletion and Salinization: In 2022, mean well depth increased to 19.39 meters (from ~12.00 meters in 2010). Mean groundwater electrical conductivity (EC) rose to 4590 µS/cm (from 2376 µS/cm in 2010), exceeding the date palm tolerance threshold of 4 dS/m (4000 µS/cm). Mean Total Dissolved Solids (TDS) climbed to 4.31 g/L (from 1.751 g/L in 2010).
• Soil Salinity: Mean soil electrical conductivity was 8.01 dS/m, well above the 4 dS/m threshold for date palm yield decline, indicating widespread salinity stress.
• Socioeconomic Impacts: 40% of farmers reported consistent water table decline, and 60% ceased well deepening due to prohibitive costs (estimated 2500 USD for a 100 meters well). 80% reported severe soil salinization, and 90% noted salty drinking water. Youth out-migration and a dramatic surge in palm grove fires were also observed.
• Spatial Autocorrelation: Global Moran’s I for NDVI was 0.4593 (z-score = 31.17, p < 0.001), indicating a highly significant and strong positive spatial clustering of desertification, with areas farther from the Draa River exhibiting higher degradation.

Contributions

• Developed an integrated framework for oasis desertification assessment combining multi-temporal satellite imagery, machine learning classification (highlighting the superior performance of Gradient Tree Boosting), in-situ hydrochemical analyses, and local socioeconomic knowledge.
• Provided a multi-decadal (1984–2024) geospatial assessment of land degradation, hydrological stress, and socio-ecological vulnerability in the Ternata Oasis, utilizing the full Landsat archive within Google Earth Engine.
• Quantified the weakening link between dam-regulated irrigation releases and oasis vegetation health, demonstrating a loss of ecological resilience post-2005 due to intensified drought and upstream hydraulic interventions.
• Documented the severity of groundwater depletion and salinization through field measurements, linking it directly to vegetation decline and socioeconomic impacts on local communities.
• Offered a replicable interdisciplinary model for dryland desertification assessment that connects surface and subsurface processes with the lived experiences of affected communities, providing crucial evidence for sustainable water and land management strategies.

Funding

• Princess Nourah bint Abdulrahman University Researchers Supporting Project number (PNURSP2025R680), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia.

Citation

@article{Moumane2026Desertification,
  author = {Moumane, Adil and Azougarh, Youssef and Enajar, Abdelhaq Ait and Alkhuraiji, Wafa Saleh and Bahdou, Ismail and Karkouri, Jamal Al and Nahas, Faten and Rebouh, N. Ya. and Youssef, Youssef M.},
  title = {Desertification monitoring in arid oasis environment using Google Earth Engine, machine learning, and field-based hydrogeological assessment},
  journal = {Scientific Reports},
  year = {2026},
  doi = {10.1038/s41598-026-41216-9},
  url = {https://doi.org/10.1038/s41598-026-41216-9}
}