Climate-driven extreme events are fundamentally reshaping riverine ecosystems, with floods, droughts, and heatwaves increasing in severity and frequency. River ecosystems, structured by their connected dendritic networks, are highly vulnerable to the propagation of localised disturbances. Both isolated and compounding extreme climatic events events affect biodiversity across scales, from the erosion of genetic diversity to the loss of ecosystem functioning, and these impacts are often amplified by other underlying stressors. However, extreme events are rare by their very nature, making them challenging to understand due to limited opportunities to learn. This presents a conundrum for predictive modelling of their impacts. Faced with insufficient information about how ecosystems might respond to such events over both short and long timescales, managers may be forced to make decisions that lack scientific credibility. Traditional ecohydrological models, often based on empirical correlations within historical ranges of variability and assumptions of hydroclimatic stationarity, may be limited when anticipating the impacts of ECEs. First, extremes are rare events located in the tails of statistical distributions, thereby challenging conventional methods that focus on central tendencies, and providing questionable insight into, and opportunities to learn from, the dynamics of extreme values. Second, the effects of extremes can propagate across levels of ecological organisation and among components of a species’ life history. Third, anomalous events are often treated as statistical anomalies that are scrubbed from the data prior to analyses. In light of these and other challenges, I discuss promising avenues for modelling frameworks that can be employed to predict the impacts of ECEs ranging from near-term forecasts to mechanistic approaches particularly suited to long-term scenario-based projections such as community-wide matrix population models. Distributional regression approaches, such as Generalised Additive Models for Location, Scale and Shape offer a particularly useful set of tools to understand the full range of ecological responses to extremes. Increasing risks of extreme events in freshwater ecosystems means new approaches are needed to transform the field from restoration thinking to resilience thinking.