Mapping Vegetation-Covered Water Areas Using Sentinel-2 and RadarSat-2 Data: A Case Study of the Caohai Wetland in Guizhou Province

Vegetation-covered water bodies (VCW) are a vital component of wetlands, and their distribution information is crucial for studying the dynamic interactions between vegetation and water. However, due to vegetation obstruction, optical remote sensing has limitations in extracting such water bodies, as it typically identifies only open water areas effectively. In contrast, microwave remote sensing, with its vegetation-penetrating capability and specular reflection characteristics, provides a more comprehensive identification of wetland water bodies. Previous studies have shown that the additional water body areas (SW) identified by SAR but not by optical sensors are often accompanied by significant vegetation cover. However, a systematic assessment of SW’s potential in mapping VCW is still lacking. This study uses the Caohai Wetland in Guizhou, China, as an example, leveraging Sentinel-2A and RadarSat-2 imagery from adjacent periods and multiple water body extraction methods to extract SW and explore its performance in mapping VCW during the dry season. Results show that during the initial stage of vegetation senescence (7 January 2019), the use of SW achieved high accuracy in mapping VCW, with overall accuracy, kappa coefficient, and F1 score reaching 84.2%, 68.4%, and 85.3%, respectively. However, as vegetation senescence deepened (12 January 2020), these metrics dropped to 76.2%, 60.7%, and 87%, respectively, indicating a significant decline in accuracy. During the vegetation regrowth stage (7 April 2020), the overall accuracy, kappa coefficient, and F1 score were 71.1%, 57.2%, and 70.9%, respectively. As vegetation continued to grow (21 April 2019), these metrics improved to 79.4%, 67.2%, and 86.6%. In summary, SW extracted from high-resolution optical and SAR imagery can preliminarily map VCW during the dry season. Furthermore, its identification accuracy improves significantly with increasing vegetation density. This study provides a novel perspective for wetland water body monitoring and the study of vegetation-water interactions.