Abstract 183

Abstract ID: 183

Probabilistic causal network modelling of Southern Hemisphere jet stream long-range predictability in spring-to-summer

Lead Author: Elena Saggioro
University of Reading, United Kingdom

Keywords: predictability, causality, stratosphere, long-range, Southern Hemisphere

Abstract: The Southern Hemisphere eddy-driven jet is an important component for skilful predictions on sub seasonal to seasonal (S2S) timescales in the mid-to-high latitudes. The jet’s variability in early summer is strongly coupled with the springtime stratospheric polar vortex variability. Since the vortex is known to be influenced by a number of long-lead drivers in the troposphere (El Niño southern oscillation) and in the stratosphere (quasi-biennial oscillation, ozone concentration, the polar night jet oscillation), it is believed that the vortex could provide a pathway for improving S2S tropospheric forecast in the region. Yet a quantification of the predictability arising from each of these drivers, and from their combinations, has been lacking. Most modelling studies address one driver at the time, and fail to detect the potential effect of their combinations. Statistical analyses of observations often rely on correlations or composites, which are not able to quantify predictability but only association.

Here a simplified model of the coupled stratospheric-tropospheric variability on a monthly timescale is constructed to generate synthetic predictions of the jet and quantify the skill arising from the above-mentioned drivers. The model is a probabilistic Causal Network which incorporates the drivers of interest, and is parametrised with the ECMWF 51-members ensemble hindcast initialised in late winter. The vortex state in spring is confirmed to be determinant for skilful predictions of jet’s variability in early summer, contrasted with a small role played by ENSO. This is true for both the jet’s seasonal poleward and equatorward shifts individually, confirming a hypothesis present in the literature. However, while the analysed long-lead drivers have the expected association with the vortex state, they only provide moderate prediction skill. A possible explanation is a dominant role of internal stratospheric variability on time scales shorter (e.g. weeks) than the ones considered here (months) in determining the vortex state. This suggests that long-range jet predictability may be ultimately constrained by how well models capture internal stratospheric variability in spring.

Theodore G. Shepherd (University of Reading)