Knowledge about the timing and magnitude of CO2 fluxes in soils and groundwater is a critical component of the global carbon budget. However, significant uncertainty is associated with predictions of carbon fluxes in soils due to water-rock interactions, redox reactions and other biogeochemical processes that affect carbon turnover and distribution in the subsurface. The objective of this study is to infer the relative contribution of different pathways (such as atmospheric exchange, precipitation/dissolution of carbonate minerals, and biotic heterotrophic and chemolithoautotrophic reactions) on carbon fluxes at a flood plain site in Rifle, Colorado. A 2-D biogeochemical reactive transport model has been developed for the unsaturated-saturated zone of the Rifle site. Results from model simulations suggest the need to include microbial contributions from chemolithoautotrophic processes (e.g., sulfur and iron oxidation) and temperature gradients to match locally-observed high CO2 concentrations in the unsaturated zone. Ignoring these processes leads to a significant underprediction of carbon fluxes at the site.