Patterns of streamflow and connectivity in freshwater ecosystems have been altered by human activities globally. There has been considerable progress in recent decades in understanding how flow affects ecological patterns and processes, but largely from small-scale monitoring of discrete events. Scientists are now increasingly being tasked with performing flow-ecology assessments at large spatial and temporal extents to inform environmental flow policy and management decisions. Consequently, upscaling observed flow-ecology relationships in space and time is a current and pressing research challenge for freshwater ecology. One promising approach for extrapolating ecological patterns is the use of mechanistic models, which provide a causal representation for processes in terms of interactions between physical and biological components of the system. We describe the application of a mechanistic model to estimate gross primary production (GPP) for rivers in the Murray Darling Basin. Production was modelled as a function of light (modified by the effect of flow on depth and light attenuation) and water temperature, and results validated with data collected during a large-scale monitoring project. The model explained approximately 60% of the variation in daily GPP and was improved by the inclusion of an autoregressive term, which represents autotrophic biomass. We envisage that this mechanistic model, combined with estimates of river depth and turbidity, will be valuable for estimating the fixation of carbon for river food webs at unsurveyed locations across the landscape.