Abstract: The riparian areas of lakes and reservoirs with diverse land use types have a large fluctuation of carbon sequestration potential affected by land management strategies. An accurate estimation of carbon storage and sequestration potential is a critical step for carbon neutrality practices. The carbon-storage estimation has been widely implemented in forest and farm land ecosystem, but has been rarely investigated for the riparian areas of lakes and reservoirs. To quantify the carbon storage of the riparian areas of lakes and reservoirs, this study selected the riparian areas of Shahe Reservoir (surface water area of 12 km2 and mean water depth of 7 m) in a mountain area of eastern China as the study area. A raster-based model was developed to estimate carbon storage based on the InVEST model, widely used in case studies across the world. The estimated results were comparable with those from previous publications, implying the reasonability of the model. The developed model was then used to investigate the response of carbon storage and sequestration potential to land management strategies, such as optimization of cultivation and fertilization, and mixed planting. Our investigation results revealed that the riparian areas of Shahe Reservoir had a total carbon storage of 3.25×105 t (i.e., 9 961 t/km2) in 2022. Forests, farmlands and tea lands contributed to 33.24%, 24.34% and 8.62% of the total carbon storage. Cultivation optimization had the largest potential in carbon sequestration (22 357 t). Our study demonstrated the significant advantage of the developed raster-based model in estimating carbon storage, and can be potentially used in other lakes and reservoirs via parameter measurements. However, the developed model also included considerable model uncertainties, which required significant efforts to minimize. For example, the parameter of carbon density was a critical and sensitive parameter in the developed model, but demonstrated a large variation among different land use types. Therefore, further work is needed to improve the accuracy of model input data.