<![CDATA[Sand encroachment is among the most significant processes of land degradation in arid and semi-arid regions, producing complex spatial and temporal dynamics as dune formations expand at the expense of pastoral and agricultural land. This study used Sentinel-2 imagery and Random Forest classification on the Google Earth Engine (GEE) platform to map and analyse land-cover change in the wilayas of El-Bayadh and Naâma (south-western Algeria) between 2019 and 2025. Supervised classification of Sentinel-2A imagery produced land-cover maps for 2019 and 2025 (overall accuracy 88.9% and 91.5%; Kappa 0.87 and 0.90, respectively). Transition matrices quantified flows to and from the 'Dunes and Shifting Sands' class. Dune formations remained relatively stable, with 5.37 million ha unchanged, while the class expanded by 219,458 ha (+3.7%) overall. This expansion was predominantly fed by conversion from grazing land (627,335 ha), bare soil (57,395 ha) and agricultural land (32,959 ha), reflecting intensifying desertification linked to climate variability, overgrazing, deforestation and cropland abandonment. Conversely, 455,993 ha of former dune surfaces reverted to grazing land and 86,631 ha to bare soil, corresponding mainly to areas where natural vegetation recovered or where biological dune-fixation works were implemented. These results confirm the value of multi-temporal remote sensing combined with GEE for monitoring large-scale land degradation. Critically, the study identifies management opportunities: areas undergoing dune stabilization through vegetation recovery or anti-desertification engineering provide model zones for scaled intervention.]]>
