Land use and land cover mapping in the Kenitra region (2017–2024) using sentinel-2 imagery, GIS and machine learning on google earth engine

Abstract

This study assessed land use and land cover (LULC) changes in the Kenitra region of Morocco from 2017 to 2024 using Sentinel-2 imagery and Google Earth Engine (GEE). Four machine learning algorithms—Random Forest, CART, SVM, and k-NN—were applied to classify six LULC classes: Built-up, Forest, Water, Bare Land, Wetlands, and Agriculture. The 2024 map was validated using field-collected data, with Random Forest achieving the highest accuracy (OA: 88.5%, Kappa: 0.861). Results revealed substantial urban expansion and bare land increase, alongside marked declines in forest, wetland, and agricultural areas. These trends signal land degradation, environmental stress, and shifting land use patterns under combined anthropogenic and climatic pressures. The findings underscore the importance of cloud-based remote sensing and machine learning for timely, high-resolution monitoring of environmental change, while also informing policy priorities. Recommended actions include controlled urban expansion, targeted restoration of degraded ecosystems, sustainable agricultural practices, and aquifer protection measures—essential for safeguarding land and water resources and guiding sustainable development in the Kenitra region.