Flooding events have been frequently observed throughout the world and may have devastating effects on the environment. Mapping of flood extent is important for taking the necessary mitigation measures for future. The freely available Sentinel-1 radar and Sentinel-2 optical images allow analysis of flood extent with adequate spatial resolution. However, temporal resolution may be insufficient, and currently only events with the suitable satellite orbital passes can be analyzed with Sentinel sensors. In addition, clouds are often in scene during and shortly after a flood event, which hinders the use of optical imagery. Here, we investigated the complementary use of Sentinel-1 and Sentinel-2 data with a land cover classification approach based on random forest over a part of northern Türkiye, which frequently confronts floods and landslides. We expanded the feature set with principal components (PC) of gray-level co-occurrence matrix (GLCM) variables obtained from Sentinel-1 polarization and Sentinel-2 spectral bands, and also the normalized difference vegetation index (NDVI) and modified normalized difference water index (MNDWI) images produced from the optical data. The training and test data were manually extracted from pre- and post-event optical data. The findings demonstrated that using GLCM PCs significantly increased the overall accuracy (OA = 99% with GLCM and OA = 93% without GLCM) of the classification. Furthermore, the flooded vegetation differs in textural features when compared with the other inundated surfaces, and also permanent water. Therefore, by allowing the separation of flooded vegetation and the other flooded areas, the GLCM data considerably increased the map quality.