Rivers transport globally significant amounts of carbon (C) from terrestrial ecosystems to ocean margins. Understanding and quantifying the sources and respective contributions to riverine C has emerged as an important biogeochemical problem that can be approached through natural-abundance isotope mass balance. Traditionally, the sources of riverine C have been identified either qualitatively or quantitatively through application of static mixing models. However, both source signatures and contributions can vary significantly with time. Here we apply two time-based mixing models to a study of six rivers draining the northeast U.S. In the first model, a time-averaged mixing model (TAMM), we vary only the source isotopic (δ13C and Δ14C) signatures. In the second model, a time-varying mixing model (TVMM), we allow both isotopic signatures and contributions to vary with time. Based on results from the TVMM, drivers of variation in riverine particulate organic C (POC), dissolved organic C (DOC), and dissolved inorganic C (DIC) include stream discharge, stream discharge and water temperature, and water temperature and vegetation phenology, respectively. Major sources include C3plant material, algal material and slow-turnover soil OC (“slow SOC”), which together account for 50%–100% (95% CI) of riverine POC; C3plant material and slow SOC, which together typically account for 60%–100% (95% CI) of DOC; and atmospheric exchange which alone typically accounts for 40%–60% (95% CI) of DIC. Seasonal change in relative contributions from algal material, slow SOC, and photosynthesis (in response to the identified drivers) dominates the observed variation in POC, DOC and DIC, respectively. The TVMM is a novel tool to identify component contributions under more realistic non-static conditions, and with potential application to a broad range of biogeochemical studies.