Forecasting population persistence in environments subjected to periodic disturbances represents a general challenge for ecologists. In arid and semiarid regions, climate change and human water use pose significant threats to the future persistence of aquatic biota whose populations typically depend on permanent refuge waterholes for their viability. As such, habitats are increasingly being lost as a result of decreasing runoff and increasing water extraction. We constructed a spatially explicit population model for golden perch Macquaria ambigua (Richardson), a native freshwater fish in the Murray-Darling Basin in eastern Australia. We then used the model to examine the effects of increased aridity, increased drought frequency, and localized human water extraction on population viability. Consistent with current observations, the model predicted golden perch population persistence under the current climate and levels of water use. Modeled increases in local water extraction greatly increased the risk of population decline, while scenarios of increasing aridity and drought frequency were associated with only minor increases in this risk. We conclude that natural variability in abundances and high turnover rates (extinction/recolonization) of local populations dictate the importance of spatial connectivity and periodic cycles of population growth. Our study also demonstrates an effective way to examine population persistence in intermittent and ephemeral river systems by integrating spatial and temporal dynamics of waterhole persistence with demographic processes (survival, recruitment, and dispersal) within a stochastic modeling framework. The approach can be used to help understand the impacts of natural and anthropogenic drivers, including water resource development, on the viability of biota inhabiting highly dynamic environments.