基于Sentinel-1 SAR数据的青狮潭水库水体提取与蓄水量反演OA

Accurate long-term monitoring of water storage in mountainous reservoirs is often hampered by severe terrain-induced radar shadowing and frequent cloud-rain conditions,which seriously limit the effectiveness of conventional optical remote sensing.Taking Qingshitan Reservoir in Guilin,Guangxi as a case study,a SAR-based adaptive water extraction method is developed using long-term Sentinel-1 dual-polarization(VV/VH)data for the period 2015-2025.On the basis of the Sentinel-1 dual-polarization water index(SDWI),a composite water response is constructed by integrating SDWI,multi-polarization backscatter and texture features,while the DEM-based terrain information and multi-temporal water occurrence frequency are incorporated to build a robust adaptive thresholding scheme suitable for complex topography and varying hydrological conditions.High-accuracy water surface areas are derived from Sentinel-2 and other optical data and combined with in-situ reservoir storage measurements to construct an S-shaped area-storage relationship model,which is then used to reconstruct the water storage dynamics of Qingshitan Reservoir from 2015 to 2025.The results show that:(a)the adaptive method achieves overall accuracies of 0.907 and 0.953 during typical wet and dry seasons,respectively,demonstrating improved water identification capabilities under complex terrain and hydrological conditions;(b)the constructed area-storage model shows high fitting accuracy(R2=0.970 6),which effectively inverts the reservoir storage;and(c)the SAR-derived storage estimates agree well with observed reservoir storage,with an R2 of 0.953 and an RMSE of 219.8×103 m3.This study provides an economical and practical remote sensing-based long-term approach for monitoring storage dynamics in data-scarce mountainous reservoirs,with important implications for regional water resources management and ecological protection.