Cross evaluation of optical versus sentinel-1 and 2 fusion data for tropical ecosystem mapping

Advances in satellite technology have greatly enhanced ecosystem mapping, a key aspect of environmental studies. This study applies a hybrid model to assess the efficiency of combining Sentinel-1 and Sentinel-2 data versus using Sentinel-2 alone across three priority ecosystems in Bangladesh: wetlands (Hakaluki Haor), riverine areas (Padma-Jamuna confluence), and mangroves (Sundarbans). Results indicate that integrating Sentinel-1 improves mapping accuracy across diverse ecosystems. Unlike previous studies that demonstrated the general advantages of Sentinel-1 and 2 fusions, this study provides a cross-ecosystem evaluation across mangrove, riverine, and wetland environments using a unified RF–SVM hybrid model, offering comparative insights on how fusion benefits vary by ecosystem type. The study utilizes C-band dual-polarization Synthetic Aperture Radar (SAR) data from Sentinel-1 alongside four spectral bands-blue, green, red, and near-infrared from Sentinel-2 to analyze imagery collected between December 2023 and February 2024. The classification was performed using a hybrid SVM-enhanced Random Forest model, in which SVM-derived probability features were incorporated into the RF structure to improve separability of spectrally similar classes. The results demonstrate that the fusion of Sentinel-1 and Sentinel-2 data substantially improves classification performance, such as overall accuracies of 94.17% for mangroves, 87.30% for riverine, and 85.96% for wetland ecosystems. In contrast, the use of singular Sentinel-2 imagery yields lower accuracies of 91.56%, 85.21%, and 82.51% for the respective ecosystems. These accuracy scores were evaluated using balanced training–testing splits. Beyond numerical accuracy gains, the fusion approach enhances class separability for structurally similar vegetation types, provides robust performance across diverse landscape conditions (CV < 2.1–4.8%), and enables ecologically meaningful discrimination that optical data alone cannot achieve. The integration of radar data is found to provide critical information, especially in environments with dense vegetation or cloud cover, where optical data alone is insufficient. This study highlights the limitations of Sentinel-2 imagery in capturing complex ecosystem details and underscores the need for Sentinel-1 integration. Data fusion enhances accuracy, deepens ecological insights, and supports effective conservation strategies.