Many organisms living in temporary streams rely on remnant surface water to survive during extended dry periods and recolonize newly established habitats when flow resumes. However, research on the spatio‐temporal variations of surface water extent for entire river networks is scarce. In this study, we first present a new field method for rapid surface water assessment. Next, we develop predictive models relating observed water extent to environmental attributes at a large number of surveyed stream segments (n=241) in eastern Australian coastal catchments. We use the models to predict daily variations in surface water dynamics throughout entire river networks over the period of 1911‐2017, based on available long‐term environmental attributes influencing hydrological processes. We find descriptors of surface water extent can be accurately predicted based on robust internal and external validations. Environmental predictor variables representing water gaining processes were more important in predicting surface water extent than variables representing water losses. Simulated long‐term variations in surface water extent were highly dynamic through space and time, particularly in inland streams, which were predicted to be the driest on average. Total stream length with surface water ranged from 8,974 to 13,742 km across the study period. Our study presents a novel and practical approach to quantifying and predicting variations in surface water extent, with potential applicability to other parts of the world. The simulated surface water extent through space and time can be used to identify and prioritize potential aquatic refuge areas that sustain aquatic biodiversity in river networks during extended dry periods.