Preprints
https://doi.org/10.5194/wcd-2022-55
https://doi.org/10.5194/wcd-2022-55
 
06 Oct 2022
06 Oct 2022
Status: this preprint is currently under review for the journal WCD.

Improved Extended-Range Prediction of Persistent Stratospheric Perturbations using Machine Learning

Raphaël de Fondeville1, Zheng Wu2, Enikő Székely1, Guillaume Obozinski1, and Daniela I. V. Domeisen3,2 Raphaël de Fondeville et al.
  • 1Swiss Data Science Center, ETH Zurich and EPFL, Lausanne, Switzerland
  • 2ETH Zurich, Zurich, Switzerland
  • 3Université de Lausanne, Lausanne, Switzerland

Abstract. On average every two years, the stratospheric polar vortex exhibits extreme perturbations known as Sudden Stratospheric Warmings (SSWs). The impact of these events is not limited to the stratosphere, but they can also influence the weather at the surface of the Earth for up to three months after their occurrence. This downward effect is observed in particular for SSW events with extended recovery timescales. This long-lasting stratospheric impact on surface weather can be leveraged to significantly improve the performance of weather forecasts on timescales of weeks to months. In this paper, we present a fully data-driven procedure to improve the performance of long-range forecasts of the stratosphere around SSW events with an extended recovery. We first use unsupervised machine learning algorithms to capture the spatio-temporal dynamics of SSWs and to create a continuous scale index measuring both the frequency and the strength of persistent stratospheric perturbations. We then uncover three-dimensional spatial patterns maximizing the correlation with positive index values allowing us to assess when and where statistically significant early signals of SSW occurrence can be found. Finally, we propose two machine learning (ML) forecasting models as competitors for the state-of-the-art sub-seasonal numerical prediction model ECMWF S2S: while the numerical model performs better for lead times up to 25 days, the ML models offer better predictive performance for greater lead times. We leverage our best performing ML forecasting model to successfully post-process numerical ensemble forecasts and increase their performance by up to 20 %.

Raphaël de Fondeville et al.

Status: final response (author comments only)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on wcd-2022-55', Anonymous Referee #1, 08 Nov 2022
  • RC2: 'Comment on wcd-2022-55', Anonymous Referee #2, 23 Nov 2022
    • RC3: 'Added Comment by RC2', Anonymous Referee #2, 25 Nov 2022

Raphaël de Fondeville et al.

Raphaël de Fondeville et al.

Viewed

Total article views: 275 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
196 68 11 275 2 1
  • HTML: 196
  • PDF: 68
  • XML: 11
  • Total: 275
  • BibTeX: 2
  • EndNote: 1
Views and downloads (calculated since 06 Oct 2022)
Cumulative views and downloads (calculated since 06 Oct 2022)

Viewed (geographical distribution)

Total article views: 268 (including HTML, PDF, and XML) Thereof 268 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 04 Dec 2022
Download
Short summary
We propose a fully data driven, interpretable, and computationally scalable framework to characterize sudden stratospheric warmings (SSWs), extract statistically significant precursors, and produce machine learning (ML) forecasts. By successfully leveraging the long lasting impact of SSWs, the ML predictions outperform sub-seasonal numerical forecasts for lead times beyond 25 days. Post-processing numerical predictions using their ML counterparts yields an up to 20 % performance increase.