Skip to main content

Summer predictions of Arctic sea ice edge in multi-model seasonal re-forecasts


In this study, the forecast quality of 1993–2014 summer seasonal predictions of five global coupled models, of which three are operational seasonal forecasting systems contributing to the Copernicus Climate Change Service (C3S), is assessed for Arctic sea ice. Beyond the Pan-Arctic sea ice concentration and extent deterministic re-forecast assessments, we use sea ice edge error metrics such as the Integrated Ice Edge Error (IIEE) and Spatial Probability Score (SPS) to evaluate the advantages of a multi-model approach. Skill in forecasting the September sea ice minimum from late April to early May start dates is very limited, and only one model shows significant correlation skill over the period when removing the linear trend in total sea ice extent. After bias and trend-adjusting the sea ice concentration data, we find quite similar results between the different systems in terms of ice edge forecast errors. The highest values of September ice edge error in the 1993–2014 period are found for the sea ice minima years (2007 and 2012), mainly due to a clear overestimation of the total extent. Further analyses of deterministic and probabilistic skill over the Barents–Kara, Laptev–East Siberian and Beaufort–Chukchi regions provide insight on differences in model performance. For all skill metrics considered, the multi-model ensemble, whether grouping all five systems or only the three operational C3S systems, performs among the best models for each forecast time, therefore confirming the interest of multi-system initiatives building on model diversity for providing the best forecasts.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10

Data availability

Copernicus C3S seasonal re-forecast data for ECMWF SEAS5, Met Office GloSea5 and Météo-France System 6 were retrieved using the Mars API. The reforecasts with the EC-Earth 3.2 and CNRM-CM6-1 models were run as part of the H2020-APPLICATE project and data is available on the APPLICATE data portal or upon request.


Download references


The authors wish to acknowledge F. Massonnet who developed the sea ice assimilation technique used to initialize EC-Earth 3.2, and S. Tietsche for pointing to the relevant SEAS5 data. All plots and graphs were realized using R, and some skill evaluations and data detrending were computed using the s2dverification package available on CRAN (Manubens et al. 2018). The authors would also like to thank two anonymous reviewers, whose feedback helped substantially improve this article.


This study was partly funded by the H2020-APPLICATE project, EU grant number 727862. JCAN acknowledges the Spanish Ministry of Science, Innovation and Universities for the personal grant Juan de la Cierva FJCI-2017-34027, PRACE for awarding access to MareNostrum at Barcelona Supercomputing Center (BSC), and ESA/CMUG-CCI3 for financial support. PO work was funded by the Ramon y Cajal grant RYC-2017-22772.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Lauriane Batté.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Batté, L., Välisuo, I., Chevallier, M. et al. Summer predictions of Arctic sea ice edge in multi-model seasonal re-forecasts. Clim Dyn 54, 5013–5029 (2020).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:


  • Seasonal forecasting
  • Sea ice
  • Arctic
  • Climate prediction