Importance of initial conditions in seasonal predictions of Arctic sea ice extent

R. Msadek, G. A. Vecchi, M. Winton, R. G. Gudgel

Research output: Contribution to journalArticlepeer-review

84 Scopus citations


We present seasonal predictions of Arctic sea ice extent (SIE) over the 1982-2013 period using two suites of retrospective forecasts initialized from a fully coupled ocean-atmosphere-sea ice assimilation system. High skill scores are found in predicting year-to-year fluctuations of Arctic SIE, with significant correlations up to 7 month ahead for September detrended anomalies. Predictions over the recent era, which coincides with an improved observational coverage, outperform the earlier period for most target months. We find, however, a degradation of skill in September during the last decade, a period of sea ice thinning in observations. The two prediction models, Climate Model version 2.1 (CM2.1) and Forecast-oriented Low Ocean Resolution (FLOR), share very similar ocean and ice component and initialization but differ by their atmospheric component. FLOR has improved climatological atmospheric circulation and sea ice mean state, but its skill is overall similar to CM2.1 for most seasons, which suggests a key role for initial conditions in predicting seasonal SIE fluctuations.

Original languageEnglish (US)
Pages (from-to)5208-5215
Number of pages8
JournalGeophysical Research Letters
Issue number14
StatePublished - Jul 28 2014
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • Geophysics
  • General Earth and Planetary Sciences


  • initial conditions
  • sea ice predictions
  • seasonal forecast systems
  • state dependence predictability


Dive into the research topics of 'Importance of initial conditions in seasonal predictions of Arctic sea ice extent'. Together they form a unique fingerprint.

Cite this