Abstract: To address the challenge of real-time monitoring of small water bodies in mountainous areas and to improve monitoring accuracy, we evaluated the water body extraction capabilities of GF-3 and Sentinel-1 radar satellite imagery.A local threshold method was applied to extract water body from radar images in multiple reservoirs in the Xiangxi River Basin.The extraction results were compared with those obtained from Sentinel-2 MSI imagery using the normalized difference water index (NDWI), followed by a quantitative analysis based on accuracy metrics.The results indicate that GF-3 has advantages in capturing water body changes, achieving a maximum recall rate of 0.952, but it is tend to be influenced by noise-induced errors in complex terrain.In contrast, Sentinel-1 demonstrates better accuracy and consistency, with a maximum precision rate of 0.913 and a maximum Intersection over Union (IoU) of 0.830.In the analysis of inundation frequency, GF-3 reveals richer spatiotemporal variations compared to Sentinel-1.The findings provide a scientific basis for the extraction of reservoir water body and water resource management under cloudy and rainy weather conditions.