Evaluation of Northern Winter Seasonal Forecasts of Major Sudden Stratospheric Warmings

Abstract

This study evaluates Northern winter seasonal forecasts initialized in early November to assess both whether they can skillfully predict the occurrence or absence of one or more major sudden stratospheric warmings (MSSWs) in the subsequent December‐January‐February period and whether ensemble members provide internally consistent predictions of MSSW occurrence. These questions are important because MSSWs are known to influence anomalous tropospheric and surface weather conditions; however, the predictability of MSSWs at the seasonal time scale remains unclear. Results based on the coefficient—a special case of the correlation coefficient for binary data—from 43‐season forecast data of one primary system demonstrate that the members are internally consistent with each other but that they are not significantly consistent with observations, suggesting a lack of or only weak forecast skill. The lack of skill is supported by results using other metrics. These conclusions are broadly robust across the different MSSW definitions tested. Similar results are found in an additional analysis using shorter 24‐season forecasts from each of the six systems. Internal consistency is not found in one system with the smallest ensemble size, while some hint of skill is observed in limited cases depending on the system and target latitude for MSSWs. It is worth noting that the results, including those related to the coefficient, vary depending on both ensemble size and the number of seasons.