In order to improve seasonal-to-interannual precipitation forecasts and their application by decision makers, there is a clear need to understand when, where, and to what extent seasonal precipitation anomalies are driven by potentially predictable surface–atmosphere interactions versus to chaotic interannual atmospheric dynamics. Using a simple Monte Carlo approach, interannual variability and linear trends in the SST-forced signal and potential predictability of boreal winter precipitation anomalies is examined in an ensemble of twentieth century AGCM simulations. Signal and potential predictability are shown to be non-stationary over more than 80% of the globe, while chaotic noise is shown to be stationary over most of the globe. Correlation analysis with respect to magnitudes of the four leading modes of global SST variability suggests that interannual variability and trends in signal and potential predictability over 35% of the globe is associated with ENSO-related SST variability; signal and potential predictability are not significantly associated with SST modes characterized by a global SST trend, North Atlantic SST variability, and North Pacific SST variability, respectively. Results suggest that mechanisms other than SST variability contribute to the non-stationarity of signal and noise characteristics of hydroclimatic variability over mid- and high-latitude regions.