A Euro-Mediterranean tree-ring reconstruction of the winter NAO index since 910 C.E.

We develop a new reconstruction of the winter North Atlantic Oscillation (NAO) index using a network of 97 Euro-Mediterranean tree-ring series. The reconstruction covers the period 910–2018 C.E., making it the longest annually resolved estimate of winter NAO variability available. We use nested correlation-weighted principal components regression and the Maximum Entropy Bootstrap method to generate a 2400-member ensemble of reconstructions for estimating the final reconstruction and its quantile uncertainties. Extensive validation testing of the new reconstruction against data withheld from the calibration exercise demonstrates its skill. The skill level of the new reconstruction is also an improvement over two NAO reconstructions published earlier. Spectral analyses indicate that the new reconstruction behaves like a ‘white noise’ process with intermittent band-limited power, suggesting that the winter NAO is stochastically forced. The ‘white noise’ properties of our reconstruction are also shown to be consistent with the spectral properties of long instrumental NAO indices extending back to 1781 and NAO indices extracted from a large number of forced climate model runs covering the last millennium. In contrast, an annually resolved multi-proxy NAO reconstruction of comparable length, based in part on speleothem data, behaves more like externally forced ‘red noise’ process, which is inconsistent with our reconstruction, long observations, and forced model runs.