Climate change is affecting coastal ecosystems worldwide as water temperatures increase, hydrologic regimes change, and sea levels rise. Consequently, estuaries risk declines in ecosystem functioning due to increasing temperatures and other hydrologic factors. Characterizing and predicting estuarine water temperature are challenging because these systems are highly dynamic. Statistical models have been used to accurately assess air temperature-water temperature relationships in lakes and streams but have not been effectively applied to tidally influenced ecosystems like estuaries. We used 6 years of continuous monitoring data from the Nisqually River Delta in Puget Sound, Washington, U.S.A., to parameterize and run a non-linear statistical model and generate spatially explicit model predictions. Our goal was to examine spatiotemporal patterns in estuarine water temperature and thermal refugia given current estimates of climactic change. The performance of the parameterized model was similar to that of non-linear stream temperature models (NSE = 0.76; RMSE = 2.34 °C). Scenarios incorporating forecasted high-emission air temperatures through the year 2100 (+ 7 °C) predicted a corresponding 3.55 ± 0.63 °C increase in average water temperatures; however, moderate and high rates of sea-level rise offset temperature increases by 3–20% and substantially reduced the amount of time temperatures exceeded the thermal stress threshold of 20 °C for juvenile salmon. These findings demonstrate how the effects of one climate stressor (sea-level rise) may offset another (temperature increases) to maintain thermal refugia for coldwater fishes. Similar exercises may allow managers to explore mitigation options like the planting of riparian vegetation or modified flooding regimes to further offset rising water temperatures.