Assessing the Impact of Air-Sea Coupling on Subseasonal Predictability of Atmospheric Rivers

Atmospheric rivers (ARs), moisture filaments extending from the tropics to extratropics, have significant impacts on western North America and also account for the majority of California’s annual precipitation. Understanding factors contributing to AR intensity is key to improving the accuracy of AR forecasts.

Large-scale AR predictability sources, primarily tropical drivers such as El Niño Southern Oscillation (ENSO) and the Madden-Julian Oscillation (MJO), are highly air-sea coupled systems, setting up large-scale sea surface temperature (SST) patterns and convection sources. While we know how air-sea coupling processes are essential in many subseasonal climate phenomena, the community has yet to quantify the improvement of the subseasonal AR forecast associated with a shift from fixed SST to fully coupled models.

We take advantage of Environment and Climate Change Canada's (ECCC) global model output in the S2S project to assess the impact of air-sea coupling on subseasonal prediction. The assessment of air-sea coupling is examined by comparing a pair of global ECCC models: one with SST held constant based on the initial conditions, and another with SST varying according to forecasts from a coupled dynamical ocean model. The consistency of the atmospheric model removes the uncertainty in atmospheric physics, which is common in inter-model comparisons. Each ECCC model produces four ensemble members, with each ensemble providing a 32-day forecast every week. A total of twenty years of forecasts are used for this study.

This project focuses on storm activity in the Northern Hemisphere during the winter months from November through March. We find that neglecting air-sea coupling results in overestimating the strength of Aleutian and Icelandic Lows by 15 days, subsequently leading to forecasting stronger integrated water vapor (IVT) over the Northeast Pacific and Northeast North Atlantic. On the other hand, the air-sea coupling in ECCC does not improve IVT predictions in the southwest USA, underscoring the importance of capturing tropical IVT sources. Beyond 15 days, the differences in forecasts between the paired models are mainly a result of unrealistic stormtrack locations due to missing seasonal SST evolution.