Rivers and adjacent riparian zones receive and process vast quantities of carbon, and play vital roles in global carbon cycles, especially proportional to the small fraction of the landscape they occupy. To effectively model carbon dynamics we need to understand environmental and biotic drivers of decomposition rates, and those mechanistic data are especially lacking in the tropics and Global South. Here, we couple data from a global-scale decomposition experiment (CELLDEX: Cellulose Decomposition Experiment), the HydroSHEDS databases, the TRY plant trait database, and meta-analyses on litter decomposition experiments to create and validate global maps of decomposition in rivers and riparian zones. The decomposition experiment used a standardized substrate (cellulose as cotton strips) in 550 reference sites located in each of Earth’s major biomes spanning 140 degrees of latitude. We related cellulose decay rates to a suite of subwatershed environmental variables using boosted regression trees, which explained 87% of variation in measured cotton decay rates (78% in riparian zones). The modeled cotton decay rates estimate carbon-processing “potential” given that natural litter has more complex chemistry that can modify decay rates. We tested whether our estimates of carbon processing potential could predict site-specific rates of natural leaf litter decomposition by comparing modeled data to a published dataset that contained 978 unique observations from 422 stream reaches and represented 31 genera of leaves. In combination with genus-level leaf traits, our estimates of carbon-processing potential explained 82% of the variation in empirical decomposition rates of natural leaves. These geospatial models of carbon processing potential will allow researchers to generate a priori estimates of decomposition rates in streams that have a mechanistic basis that accounts for both litter traits and environmental drivers of decomposition.