SDSC-Connect 25th May 2023

Towards Biomedical Data Science & Precision Medicine | Conference


EXtending the PrEdiCTability of the Atmosphere over Europe


Currently available weather forecasts over Europe tend to have skill up to about one week, although theoretical estimates indicate a theoretical limit of about 3 weeks (Buizza et al., 2015; Domeisen et al., 2018a). After that, predictability often decreases sharply. However, longer term forecasts are possible if information is available that provides an increased probability of a long-lasting event, or an influence from a region that has a slower variability – such as the ocean (e.g. Duchez et al., 2016) – or longer predictability – such as the tropics (e.g. Greatbatch et al., 2015; Scaife et al., 2017; Wulff et al., 2017). The upper atmosphere, i.e. the stratosphere at about 12 – 50km above the Earth’s surface, is such a region that provides increased predictability to Europe after extreme stratospheric events, so-called Sudden Stratospheric Warming (SSW) events (Scaife et al., 2016). These events can provide skill over Europe for up to several weeks to months (Sigmond et al., 2013), with persistently colder than usual weather over Northern and central Europe. One of these events occurred in February 2018 and led to persistent cold weather in Europe in late February and early March after an otherwise mild winter.

SSW events themselves are however only possible to predict several days in advance. An extended prediction of SSW events would therefore significantly benefit forecasts at the surface. While the stratosphere is not the only region with predictive potential for Europe, most other predictors exhibit a pathway through the stratosphere, such as e.g. predictability arising from the tropics (e.g. Domeisen et al., 2015; Butler et al., 2016). It is therefore crucial to understand the predictability of the stratosphere itself. However, SSW events are difficult to classify, and indeed there exists a range of methods and classifications. This project aims to find an automated classification of these events in addition to an automated analysis of predictability over Europe. The main objectives of this project are described here:

  • Only recently have predictions on sub-seasonal to seasonal timescales, i.e. weeks to months, become publicly available. This project aims at extracting novel insights from this data using data science tools.
  • A first step will be an improved classification of stratospheric events, allowing for a flexible definition that includes the predictability aspects of these events.
  • In a second step, this project aims to classify remote predictors of long-term weather variability. In particular, known predictors for stratospheric and tropospheric variability will be evaluated using data science methods and possible new predictors will be identified. This knowledge is expected to lead to an improved predictability of the weather over Europe on weekly to monthly timescales.


Starting Date

February 2020

PI / Partners

Atmospheric Predictability Group (ETHZ)

Atmospheric dynamics (UNIL)

DS3Lab: Data Sciences, Data Systems, & Data Services (ETHZ)

Geoscience & Remote Sensing (TU Delft)



From a climatological point of view, winter is characterized by the apparition in the stratosphere of a vortex of strong winds over the polar regions; this vortex is generally centered at the pole and have with a circular shape.

In average once every two years, sudden and rapid warmings of the stratosphere disturb the vortex causing displacements and deformations. The impact of such events, known as Sudden Stratospheric Warmings (SSW), are not limited to the stratosphere but also strongly influence the troposphere, i.e., weather at the sea level, during two to three months after an occurrence.

Thus a good prediction of SSW events would enable the production of more accurate weather forecasts at a monthly time scale. However, the stratosphere is a complex environment: the different states of the polar vortex and their relations with SSW events are still not well understood.


  • Use Data science methodologies to better characterize the relationship between SSWs and the different states of the polar vortex.

  • Improve the prediction of disturbed state of the stratosphere by combining data driven prediction algorithms with the knowledge extracted from the first point.


    • Contribute to a better understanding of the coupling between stratosphere and troposphere.

    • Improve the predictability of weather at a seasonal time scale. Accurate seasonal forecasts are particularly important for agriculture planning, public health and natural resources management.

    Potential Vorticity at 10hPav


    • De Fondeville, R., Z. Wu, E. Székely, G. Obozinski, and D.I.V. Domeisen: Improved extended-range prediction of persistent stratospheric perturbations using machine learning, Weather and Climate Dynamics Discussions,, under review.

    • Wu, Z., T. Beucler, E. Székely, W.T. Ball, and D.I.V. Domeisen (2022): Modeling Stratospheric Polar Vortex Variation and Identifying Vortex Extremes Using Explainable Machine Learning. Environmental Data Science, Vol. 1, e17,

    • Wu, Z., B. Jiménez-Esteve, R. de Fondeville, E. Székely, G. Obozinski, W.T. Ball, and D.I.V. Domeisen (2021): Emergence of representative signals for sudden stratospheric warmings beyond current predictable lead times. Weather and Climate Dynamics, 2, 841–865.


    • Baldwin, M. P. and Dunkerton, T. J. (2001). Stratospheric Harbingers of Anomalous
      Weather Regimes. Science, 294(5542):581-584.

    • Butler, A. H., Seidel, D. J., Hardiman, S. C., Butchart, N., Birner, T., & Match, A. (2015). Defining sudden stratospheric warming. Bull. Amer. Meteor. Soc., 1–16.

    • Blume, C., Matthes, K., and Horenko, I. (2012). Supervised Learning Approaches to Classify
      Sudden Stratospheric Warming Wvents. Journal of the Atmospheric Sciences, 69(6):1824-1840.

    • Coughlin, K. and Gray, L. J. (2009). A Continuum of Sudden Stratospheric Warmings. Journal of
      the Atmospheric Sciences, 66(2):531-540.