Continuous spatiotemporal variations of riverine dissolved organic carbon in the Yangtze River Basin revealed by a river-reach-scale machine learning-based model

Study region The Yangtze River Basin (YRB), China Study focus Riverine dissolved organic carbon (DOC) is critical to the global carbon cycle, while its continuous spatiotemporal variations remain poorly understood in large rivers. This study developed an estimation model at the river-reach scale using interpretable machine learning to simulate long-term (2000–2020) monthly riverine DOC concentration (C DOC ) in the YRB, identifying its continuous spatiotemporal variations, export characteristics, major predictors, and their effects. These findings highlight the increasing trends and transport-limited characteristics of riverine DOC in the YRB, which are important for understanding the riverine carbon cycle processes. New hydrological insights for the region Seven major predictors representing geography, vegetation, climate, and river attributes had nonlinear effects on spatiotemporal variations in riverine C DOC . Elevation showed dominant negative effects with local low-elevation positive impacts. Generally, riverine C DOC responded positively to vegetation productivity, precipitation, and air temperature, but negatively to river attributes. Annual riverine C DOC averaged 4.69 ± 0.44 mg/L, higher in midstream small tributaries of the YRB. Approximately 23.6% of river reaches exhibited significant C DOC increases, primarily in the up-midstream regions. Riverine DOC export to the ocean was 2.02 ± 0.35 Tg C/yr, with a transport-limited regime. Moreover, mainstream C DOC showed an initial decrease from the source followed by a fluctuating increase and a final decrease toward the estuary, whereas most of its major tributaries generally exhibited fluctuating downward trends.