High-frequency monitoring of reservoir inundated areas and water storage changes can facilitate the evaluation of reservoir functions, the calculation of evaporative and seepage losses, as well as the calibration of hydrological models. Because optical remote sensing images are affected by clouds, the integration of optical data with SAR backscattering coefficients (SAR backscatters) was regarded as a better approach towards high-frequency inundated area monitoring. However, because the discrepancy between inundated pixels' SAR backscatters and the surrounding land surfaces' backscatters may differ significantly among different regions and seasons, traditional methods can hardly obtain consistent high-quality retrievals of reservoir inundated areas over large areas. In this study, by developing both reservoir- and monthly-specific Support Vector Machine (SVM) models, we improved the integration of SAR backscatters with optical remote sensing-based surface water dynamics. Among 785 large reservoirs in China, our method obtained monthly inundated areas at 721 reservoirs during 2017‒2021. In addition, we utilized multisource satellite altimetry records as well as Digital Elevation Models (DEMs) observed at different periods to fit reservoir bathymetry, and then transformed the inundated areas into monthly water storage changes at 662 reservoirs, accounting for 93% of the total storage capacity of all large reservoirs. In terms of the validation against in-situ measurements at 80 reservoirs across China, the quality of monthly inundated area monitoring is superior to existing datasets by more than 20%, while the R2 and RMSE of our reservoir water storage change estimates reached 0.79±0.18 and 21%±14%, respectively.