Wetlands are considered among the most productive ecosystems on Earth, as they shelter a diversity of species and maintain ecological balance. However, their ongoing degradation threatens biodiversity and ecosystems, underscoring the need for regular and long-term monitoring. The Sidi Moussa Oualidia wetland complex, a Ramsar site in Morocco, is a critical habitat for migratory birds and an invaluable ecological resource known for its complex landscape patterns. In this study, we present a framework in Google Earth Engine (GEE) that fuses optical, radar, texture, and terrain data with both pixel- and object-based classification to map and classify wetlands at 10 m resolution. We first generate a cloud-free Sentinel-2 composite using Scene Classification masking and pansharpen the 20 m SWIR band (B11) to 10 m, enabling precise computation of NDWI, MNDWI, and GLCM texture indices. A Sentinel-1 VV/VH ratio and SRTM-derived slope are added to the stack. A pixel-level Random Forest (RF) classifier is trained on stratified samples to produce an initial map. We then segment the RGB composite into superpixels via Simple Non-Iterative Clustering (SNIC) and assign each superpixel its majority RF class, smoothing speckle and salt-and-pepper noise while preserving ecologically meaningful object boundaries. Validation against ground truth points yields an overall accuracy of 94 % and a Kappa of 0.91—an 8 % improvement over pixel-only results. Our first-of-its-kind approach in Morocco, designed to capture the complex spatial patterns of heterogeneous wetland environments, provides a promising solution for operational wetland monitoring and supports informed spatial decision-making in water resource management and ecological conservation.