Effective conservation and management of natural resources requires accurate predictions of ecosystem responses to future climate change, but environmental science has largely failed to produce these reliable forecasts.  The future response of Pacific salmon (Oncorhynchus spp.) to a changing environment and continued anthropogenic disturbance is of particular interest to the public because of their high economic, social, and cultural value.  While numerous retrospective analyses show a strong correlation between past changes in the ocean environment and salmon production within the north Pacific, these correlations rarely make good predictions.  Using a Bayesian time–series model to make successive 1–yr–ahead forecasts, we predicted changes in the ocean survival of Snake River spring/summer chinook salmon (O. tshawytscha) from indices of coastal ocean upwelling with a high degree of certainty (R² = 0.71).  Furthermore, another form of the dynamic times-series model that used all of the available data indicated an even stronger coupling between smolt–to–adult survival and ocean upwelling in the spring and fall (R² = 0.96).  This suggests that management policies directed at conserving this threatened stock of salmon need to explicitly address the important role of the ocean in driving future salmon survival.