Evaluating predictability limits of North American winter precipitation

Given that seasonal precipitation predictions over North America are not particularly skillful, assessing whether forecast system refinements can enhance skill and societal usefulness of seasonal forecasts is important. We investigate by using average predictability time (APT) analysis to filter wintertime, seasonal precipitation hindcasts provided by the Seamless System for Prediction and Earth System Research. Using this method, which decomposes forecasts into predictable modes, we find limited potential to improve seasonal precipitation forecasts over North America owing to the subseasonal predictability timescales of most APT modes. Nevertheless, more skillful forecasts of APT mode 2, which is tied to equatorial Pacific convection and has a predictability timescale of about 220 days, may improve seasonal precipitation forecasts over North America. We demonstrate that predictions for the winters of 2015–2016 and 2021–2022, which featured notable forecast errors over western North America, may have been improved with better predictions of this second APT mode.