We examine the multiscale influence of environmental and hydrological features of the riverine landscape on spatial and temporal variation in fish assemblages in eastern Australia. Multiresponse artificial neural network models provided accurate predictions of fish assemblages in the Mary River based on species presence–absence data (mean Bray–Curtis similarity between predicted and observed composition = 84%) but were less accurate when based on species relative abundance or biomass (mean similarity = 62% and 59%, respectively). Landscape- and local-scale habitat variables (e.g., catchment area and riparian canopy cover) and characteristics of the long-term flow regime (e.g., variability and predictability of flows) were more important predictors of fish assemblages than variables describing the short-term history of hydrological events. The relative importance of these variables was broadly similar for predicting species occurrence, relative abundance, or biomass. The transferability of the Mary River predictive models to the nearby Albert River was high for species presence–absence (i.e., closer match between predicted and observed data) compared with species abundances or biomass. This suggests that the same landscape-scale features are important determinants of distribution of individual species in both rivers but that interactions between landscape, hydrology, and local habitat features that collectively determine abundance and biomass may differ. Yes Yes