We developed logistic regression models from data on biotic and abiotic variables for 172 sites on Banks Peninsula, New Zealand, to predict the probability of occurrence of two diadromous fish, banded kokopu (Galaxias fasciatus) and koaro (G. brevipinnis). Banded kokopu occurrence was positively associated with small streams and low-intensity land uses (e.g., sheep grazing or forested), whereas intensive land uses (e.g., mixed sheep and cattle farming) and lack of riparian forest cover impacted negatively on occurrence at sampled sites. Also, if forests were positioned predominantly in lowland areas, banded kokopu occurrence declined with increasing distance to stream mouth. Koaro occurrence was positively influenced by catchment forest cover, high stream altitudes, and areas of no farming activity or mixed land uses. Intensive land uses, distance to stream mouth, and presence of banded kokopu negatively influenced koaro occupancy of stream reaches. Banded kokopu and koaro presence was predicted in 86.0% and 83.7% agreement, respectively, with field observations. We used the models to quantify the amount of stream reaches that would be of good, moderate, and poor quality, based on the probability of occurrences of the fish being greater than 0.75, between 0.75 and 0.5, or less than 0.5, respectively. Hindcasting using historical data on vegetation cover undertaken for one catchment, Pigeon Bay, showed they would have occupied most of the waterway before anthropogenic modification. We also modeled potential future scenarios to project potential fish distribution.