Desertification monitoring in arid oasis environments using Google Earth Engine machine learning and field based hydrogeological assessment

Oasis ecosystems, vital for water and food security in arid and semi-arid regions, are highly susceptible to degradation from climatic stress, fragile soils, and excessive groundwater withdrawal. This study assesses desertification dynamics in the Ternata Oasis (southeastern Morocco) by integrating remote sensing, machine learning (ML), hydrogeological fieldwork, and socioeconomic surveys. A multi-decadal monitoring framework (1984–2024) was developed using the full Landsat archive processed in Google Earth Engine, where a Gradient Tree Boosting (GTB) model was applied to map and track the spatial progression of land degradation over time. The GTB classifier achieved an overall accuracy of 87.2%, outperforming Random Forest (85.0%) and Classification and Regression Trees (CART) (82.0%), confirming its effectiveness for long-term desertification monitoring in arid environments. To contextualize the biophysical data, semi-structured interviews were conducted with long-term oasis farmers. Their insights were thematically coded and triangulated with observed desertification patterns and hydrochemical indicators. Farmers confirmed increasing salinity stress, prohibitively high well-deepening costs, youth outmigration, and a growing number of palm grove fires—largely attributed to the accumulation of dead or dying trees made more flammable by drought and salt toxicity. The results reveal a sharp decline in vegetation health. The healthy oasis area contracted significantly, while desertified land expanded. Groundwater levels dropped markedly, and water salinity exceeded critical thresholds for date palm survival. These findings underscore the combined impact of climate variability and anthropogenic overexploitation in accelerating desertification in oasis systems and highlight the urgent need for integrated water and land management strategies.