Dynamical Precursors for Statistical Prediction of Stratospheric Sudden Warming Events

This work explores dynamical arguments for statistical prediction of stratospheric sudden warming events (SSWs). Based on climate model output, it focuses on two predictors, upward wave activity in the lower stratosphere and meridional potential vorticity gradient in the upper stratosphere, and detects large values of these predictors. Then it quantifies how many SSWs are preceded by predictor events and, inversely, how many events are followed by SSWs. This allows to compute conditional probabilities of future SSW occurrence. It is found that upward wave activity leads to important increases in SSW probability within the following 3 weeks but is less important thereafter. A weak potential vorticity gradient is associated with increased SSW probability at short lags and, perhaps more importantly, decreased SSW probability at long lags. Finally, when both predictors are considered in combination, the information gain is large on the weekly and small but significant on the intraseasonal time scale. Plain Language Summary After a breakdown of the polar vortex in the winter hemisphere stratosphere, Northern Hemisphere weather can be unusual for several months or even weeks. Therefore, there is great interest in being able to predict these breakdowns, which are known as stratospheric sudden warming events. The traditional way of forecasting these events is to run full climate models forward in time similar to everyday weather forecasting. However, this is computationally expensive and model skill is very low beyond about 2 weeks. This study explores a new, probabilistic way of prediction based on dynamical arguments. It uses the past evolution of the stratosphere to construct probabilities of occurrence as a function of lead time. It shows that meaningful information can be obtained for an entire extended winter season, which is an order of magnitude longer than traditional model forecasting.