Dynamic Vegetation Phenology Improves SWAT-C Simulation of Riverine Total Carbon and Reservoir Effects in the Middle Jinsha River Basin

Riverine carbon transport is jointly controlled by vegetation dynamics, hydrological processes, and reservoir operation, but these interactions remain difficult to quantify in data-limited mountain basins. In this study, we adopted a previously developed dynamic phenology-constrained SWAT-C model (DPC-SWAT-C) and applied it to the middle Jinsha River Basin. Annually varying start-of-season and end-of-season data were used as external inputs to constrain vegetation growth. The original SWAT-C and DPC-SWAT-C models were evaluated using monthly discharge and leaf area index observations, while total organic carbon (TOC), inorganic carbon (IC), and total carbon (TC) observations from two field campaigns in 2024 were used to evaluate spatial model performance. DPC-SWAT-C increased NSE from 0.22 to 0.64 for leaf area index, from 0.58 to 0.80 for TOC, from 0.68 to 0.82 for IC, and from 0.41 to 0.75 for TC, whereas discharge NSE remained nearly unchanged (0.762–0.764). The 2003–2024 simulations showed higher TC concentrations during the dry season and lower concentrations during the wet season. Under the reservoir-development scenario, downstream TC concentration peaks and temporal variability were reduced. Mean annual TC load at the reservoir outlet decreased from 2.729 × 106 to 2.238 × 106 t C yr−1, a reduction of 17.9%. These results show that observation-based phenological constraints improved vegetation and spatial carbon simulations in the basin and suggest that the planned Longpan Reservoir may reduce downstream carbon export.