Why Do Extended‐Range Forecasts Underpredict the Extreme Negative Pacific/North American Pattern in February 2018?

Abstract The Pacific/North American (PNA) pattern is a dominant mode of mid‐latitude climate variability with major impacts on North American weather. Its extended‐range prediction remains challenging, influenced by complex drivers including upstream wave dynamics, tropical forcing, and the stratosphere. This study investigates the notable failure of extended‐range forecasts to capture the extreme negative PNA event of February 2018 using the novel Stratospheric Nudging and Predictable Surface Impacts (SNAPSI) multi‐model ensemble designed to isolate stratospheric sources of predictability. The results reveal that the forecast error stems from an underestimation of several key processes, including a chain of misrepresented factors. The primary bottleneck is the poor simulation of the East Asian trough, a well‐established upstream precursor; models that better represent it produce more accurate PNA forecasts. Biases in both tropical (La Niña) and extratropical North Pacific sea surface temperatures also contribute to the PNA forecast bias. While the Madden–Julian Oscillation has limited direct influence within the SNAPSI forecasts, a supplementary analysis shows that its accurate simulation in earlier weeks is critical for EAT development based on models from the Subseasonal‐to‐Seasonal prediction project database. The stratospheric contribution is initialization‐dependent: for a late‐January forecast, the tropospheric response to the February 2018 sudden stratospheric warming varies widely, limiting its net effect, while for a February initialization, most models produce an excessive negative Arctic Oscillation response, which amplifies a positive PNA bias. The findings highlight priorities for improving predictions: better representing mid‐latitude wave dynamics, refining the representation of stratosphere‐troposphere coupling, and integrating tropical‐extratropical linkages.