Hydrology is a major driver of species distribution patterns in freshwater ecosystems, but information on hydrology is generally only available for a small fraction of sites from which biological data have been collected. To try and fill this data gap, we coupled data from hydrologic gauging sites across the Murray-Darling Basin with information on climate, physiography and water infrastructure, and used these to generate predictive models of long-term runoff variability. We then used these models to extrapolate descriptors of hydrologic variability to ungauged sites, for use in subsequent modeling of fish distribution patterns. Here we discuss the modeling results and their potential applications with reference to the good, bad and ugly sides of their use in inference and prediction. In particular we consider the problems for range-retracted and rare species, extrapolation in geographic and environmental space, and temporal dynamics.