Abstract/Summary

Quantifying signals and uncertainties in climate models is essential for climate change detection, attribution, prediction and projection. Although inter-model agreement is high for large-scale temperature signals, dynamical changes in atmospheric circulation are very uncertain, leading to low confidence in regional projections especially for precipitation over the coming decades. Furthermore, model simulations with tiny differences in initial conditions suggest that uncertainties may be largely irreducible due to the chaotic nature of the climate system. However, climate projections are difficult to verify until further observations become available. Here we assess retrospective climate predictions of the last six decades and show that decadal variations in north Atlantic winter atmospheric circulation are highly predictable. Crucially, climate models underestimate the predictable signal by an order of magnitude and skill is achieved despite a lack of agreement between individual model simulations. Consequently, skilful climate predictions of European and eastern North American winters are possible but require 100 times more ensemble members than would perfect models and post-processing to overcome underestimated teleconnections. Our results highlight the pressing need to understand why the signal-to-noise ratio is too small in climate models and the extent to which correcting this model error would reduce uncertainties in regional climate change on timescales beyond a decade.