A computationally inexpensive climate model analog method yields skillful climate predictions across timescales, from seasons to multiple years, complementing existing climate prediction systems and potentially providing valuable information for sectors like agriculture and energy.

This paper presents and evaluates a set of retrospective decadal predictions with the EC-Earth3 climate model. These experiments successfully predict past changes in surface air temperature but show poor predictive capacity in the subpolar North Atlantic, a well-known source region of decadal climate variability. The poor predictive capacity is linked to an initial shock affecting the Atlantic Ocean circulation, ultimately due to a suboptimal representation of the Labrador Sea density.