Sub-seasonal to seasonal (S2S) predictability derives from climate processes that evolve on timescales of weeks to months. S2S predictive skill derives from forecast models faithfully simulating and predicting these processes. Stratospheric variability impacts the troposphere on approximately these timescales, making the stratosphere a potentially important source of S2S prediction skill (NAS, 2016; NRC, 2010). The S2S Prediction Project (Vitart et al., 2017) seeks to promote and improve S2S predictions by identifying sources of predictability and understanding systematic model errors that are acting to limit forecast skill. Given the observed impact of the stratosphere on the troposphere, creating a sub-project focused on the stratosphere would be valuable.
In collaboration with S2S during Phase I, the Stratospheric Network for the Assessment of Predictability (SNAP), which is a WCRP/SPARC initiative, focused on the predictability of extreme stratospheric polar variability, especially stratospheric sudden warmings (SSW), and the coupling to the troposphere (Tripathi et al., 2015). Predictability of SSW was demonstrated into the sub-seasonal range but stratospheric model biases and predictability of tropospheric planetary wave were found to limit predictive skill of SSW. In order to fully tap into the predictability that the stratosphere may provide for the troposphere in the sub-seasonal range and several key questions need to be addressed:
  • Quantification and understanding of coupling. What are the mechanisms of vertical coupling that act to promote tropospheric variability at S2S timescales (both in polar regions and in tropics)? How well are these mechanisms depicted in S2S models? What is their contribution to predictability of the troposphere at S2S timescales? What is the role of model resolution?
  • Model biases. What are key systematic errors in the stratosphere in the S2S models and where are these common across the ensemble? How do model biases impact coupling to the troposphere? How are they acting to limit predictive skill?
  • Initial conditions and ensemble generation. What is the quality of the stratospheric initial conditions in S2S models and how is it impacting S2S skill?
  • Whole atmosphere diagnostics. What additional data output is required in order to better diagnose the stratosphere and its impact on the troposphere in the S2S models?
The Stratospheric Network for the Assessment of Predictability (SNAP) is a WCRP/SPARC initiative which seeks to understand the role of the stratosphere in tropospheric predictability. In its first phase, SNAP used targeted model simulations to investigate the predictability of the stratosphere on sub-seasonal timescales (Tripathi et al., 2015). The second and current phase of SNAP is focused on evaluating model ensemble forecasts from the S2S project to better determine the role of stratospheric processes on improving tropospheric forecasting skill. A sub-project focused on the stratosphere would therefore strengthen the link between S2S and SNAP, and promote research and collaboration between these two groups.
SNAP is already working to coalesce S2S projects focusing on the role of the stratosphere in an overview paper led by Ms Daniela Domeisen. There are 13 research projects from 20 researchers around the world focused on various aspects of both upward and downward coupling between the troposphere and stratosphere. Each project uses S2S output to evaluate topics such as the role of the Quasi-Biennial Oscillation on Madden-Julian Oscillation predictability, the predictability of surface climate following stratospheric final warming events, or the role of weak or strong polar vortex events in tropospheric predictability. While many of the projects will also produce individual publications, the SNAP S2S paper will provide a broad overview on how both the tropical and extratropical stratosphere may contribute to prediction skill in the S2S models.
Figure 7. Anomaly Correlation Coefficient for NAM1000 at week 3 or week 4 lead time following (top row) neutral stratospheric vortex or (bottom row) weak stratospheric vortex conditions. Here the state of vortex is based on NAM100.
For example, Figure 7 shows the prediction skill of the 1000 hPa Northern Annular Mode for week 3 or week 4 in the S2S models, for either neutral stratospheric vortex conditions (top row), or weak vortex conditions (bottom row). It is clear that for most models, skill is higher following weak vortex conditions. Similar results are found following strong vortex conditions. In other words, extremes in the polar stratosphere can improve tropospheric winter climate prediction, during weeks notoriously difficult to skilfully predict. The figure also highlights that there is considerable difference in the size of this gain in skill between the models and a clear priority of the SNAP work will be to quantify and understand the origin of these differences.
Proposed studies
  • Continue and complete initial community paper on stratosphere-troposphere coupling in S2S models.
  • Collaboration with the SPARC Data Assimilation Working Group, the WWRP Probability, Dynamics and Ensemble Forecasting (PDEF) group, and the S2S Teleconnections Project on this topic.
  • Develop proposed set of diagnostic output/additional levels for the stratosphere and stratosphere-troposphere coupling which would help further analysis of S2S models through a community survey.
  • A session at the forthcoming SPARC General Assembly (September 2018) which focuses on the impact of the stratosphere on predictability on the S2S timescale.
  • A community meeting at the SPARC General Assembly (September 2018) to solicit ideas for additional S2S experiments that could yield insight into the questions posed above. One possible line of experiments might look at nudged model runs similar to those used by Hitchcock and Simpson (2014).