Mangrove ecosystems play a crucial role in coastal biodiversity, carbon sequestration, and shoreline stabilization. Delineating mangrove and non-mangrove areas accurately remains a challenge using remote sensing data due to its presence in dynamic intertidal ocean and tidal wave variability. The present study brings out the wetland discrimination capability of ALOS PALSAR-2 (L-band), NovaSAR-1 (S-band), Sentinel-1 (C-band), and Sentinel-2 MSI using Support Vector Machine (SVM) classification.<br />The classification results using SAR, Sentinel-1 (C-band) achieved an overall accuracy of 91.1104% with a kappa coefficient of 0.8277, better than ALOS PALSAR-2 (L-band) and NovaSAR-1 (S-band). The better classification accuracy of 96.39% with kappa coefficient 0.92 is achieved with combination of L, C SAR and Sentinel-2 MVI data due to sensitivity of SAR to forest structure and moisture. The mangrove extent obtained through optical NDVI and MVI data is overestimated due to different approach of these vegetation index. The mangrove extent is overestimated using multi-band combination of SAR 38.6 sq km and SAR-Optical combination 39.62 sq km as compared to 31.06 sq km estimated by Global Mangrove Watch (GMW) in year 2020 and 30.00 km&sup2; in Forest Survey of India (FSI) 2021 report for Lothian, Sundarbans.<br />The study highlights that classification using multi-frequency-polarization SAR and cloud free optical imagery stacked with SAR data have advantage in frequently cloud covered study area. This research contributes to coastal resource management and conservation, for large-scale mangrove monitoring. Future studies should explore deep learning approaches and time-series analysis to improve the long-term sustainability of these critical ecosystems.