What Drives the Spread and Bias in the Surface Impact of Sudden Stratospheric Warmings in CMIP6 Models?

Sudden stratospheric warmings (SSWs) are characterized by the rapid breakdown of the stratospheric polar vortex. Following this initial breakdown, anomalies in the stratosphere can persist for several months. The prolonged stratospheric anomalies associated with SSWs drive an equatorward shift of the tropospheric eddy-driven jet, contributing to persistent and predictable changes at the Earth’s surface. SSW events have been strongly associated with increased likelihood and severity of cold air outbreaks in Eurasia and precipitation extremes in southwestern Europe. These surface impacts can persist for weeks to months, making the stratosphere an important source of skill for forecasting winter weather on sub-seasonal to seasonal (S2S) timescales. To what extent SSWs could be used to improve the prediction of surface weather depends on how well the surface response to SSWs is represented in state-of-the-art climate models. We, therefore, evaluate the representation of the surface response to SSWs in 28 CMIP6 models, with a focus on the SSW-composite-mean surface response in sea-level pressure (SLP), surface air temperature, and precipitation.

Most models capture the overall magnitude of the surface response in SLP, but there is a wide range of simulated amplitudes, as measured by the magnitude of the Arctic SLP response. Regarding the spatial structure of the surface response in SLP, most models exhibit a more basin-symmetric negative NAM-like response, opposite to the highly basin-asymmetric negative NAO-like response in the reanalysis. This disagreement arises mainly because most models simulate a cyclonic Pacific SLP response, whereas the reanalysis exhibits no robust response in the Pacific basin.

To explain inter-model spread in the Arctic and Pacific SLP responses, multivariate regression relationships across models are developed through a forward parameter selection procedure. This identifies the SSW-composite-mean lower-stratospheric temperature anomalies (ΔT₁₀₀), and climatological strength (u₈₅₀) and latitude (φ₈₅₀) of the lower-tropospheric zonal mean jet as major factors modulating the Arctic SLP response and identifies ΔT₁₀₀, and the SSW-composite-mean Niño-3.4 SST anomalies (ΔSST_Niño3.4) and North Pacific SST dipole anomalies (ΔSST_NPac.) as major factors modulating the Pacific SLP response. Roughly 65% and 60% of inter-model spread in the Arctic and Pacific SLP responses are explained by their corresponding factors, respectively. Amongst these factors, the stratospheric factor ΔT₁₀₀ explains the most spread across the models: it explains roughly half of the spread in the Arctic SLP response and 30% of the spread in the Pacific SLP response across the 28 models. The two oceanic factors in the Pacific sector explain the cyclonic Pacific SLP response simulated in most models: a number of the models simulate a significantly positive ΔSST_Niño3.4, whereas no robust anomaly is detected in the reanalysis; furthermore, a number of the models strongly underestimate the amplitude of ΔSST_NPac., which plays a key role in preventing the Pacific sector from responding to the stratospheric forcing in the reanalysis. Improving models representation of the oceanic factors may bring the models closer to observations in terms of the Pacific SLP response, highlighting the importance of air-sea interactions in shaping the Pacific sector response to SSWs.

Despite the fact that the ensemble mean SLP response across the CMIP6 models is in close agreement with the reanalysis in terms of the overall magnitude, the surface air temperature and precipitation responses are systematically too weak. This indicates the limited ability of climate models to forecast the intensity of surface weather extremes driven by SSWs on S2S timescales, such as cold air outbreaks in Eurasia and precipitation extremes in southwestern Europe.