Multidecadal behavior of the North Atlantic Oscillation during the last millennium
Abstract. The North Atlantic Oscillation (NAO) is a major source of atmospheric variability in the Northern Hemisphere, affecting temperature, precipitation, and storm tracks across North America and Eurasia. Understanding NAO variability on multidecadal to centennial timescales requires paleo-reconstructions, but previously published reconstructions disagree on the magnitude of low-frequency NAO variability over the last millennium. Paleoclimate proxies for the oxygen and hydrogen isotope composition of meteoric waters have thus far been under-utilized in published NAO reconstructions. Here, we present a new reconstruction of the NAO over the last millennium using the Iso2k database, a collection of globally distributed water isotope-based paleoclimate proxy records. In contrast to recent NAO reconstructions, we find significant multidecadal to centennial scale variability. Critically, however, the strength of the low-frequency signal has not been consistent throughout the last millennium. Isotope-enabled model simulations did not reproduce the low-frequency signal in the NAO reconstructions and thus it may be necessary to account for low-frequency variability when projecting the impacts of the NAO on temperature and precipitation under future climate scenarios.
This paper seeks to tackle the debate about how the NAO varies at different timescales over the past millennium using a large network of water isotope records. The use of water isotope records for this purpose is new and this paper deserves to be published after due consideration of the issues I bring up.
I think the problem with the debate about high vs. low-frequency variability in the NAO is in part a matter of definition. If we restrict ourselves to the original definition of the NAO by Hurrell, van Loon, Jones, and others, the NAO is simply an atmospheric pressure difference index in the North Atlantic, with little inter-decadal or longer variability indicated as far back as 1824 in CRU instrumental pressure data and back to 1781 in a high-quality extension of the NAO index by Phil Jones. This is clearly indicated by their “flat” (no slope) power spectra shown in Fig. 5b of the Cook et al. (2019) paper. Thus, it should come as no surprise that the Cook et al. (2019) winter NAO index reconstruction, after due consideration taken to match the overall slope of the instrumental data power spectra, should have a “flat” power spectrum as well, as opposed to the substantial “redness” in some other NAO reconstructions (e.g. Ortega et al., 2015).
For this reason, I argue that the NAO reconstruction being presented in this paper (and in Ortega as well), with its substantially greater decadal-to-centennial variability, is not a reconstruction of the NAO atmospheric pressure index itself, but a reconstruction that includes NAO impacts from long-range, persistent, effects on atmospheric circulation and regional climate. The authors know this because it is explicitly stated by them on lines 64-66 of the paper (“Water isotopes provide information about the NAO on broader spatial and temporal scales … because they integrate on basin-wide to hemispheric scales …”). See also lines 79-81. As such, what is presented here is distinctly different from the reconstruction of the NAO index itself. This difference in definition was also reflected in the title of the Hurrell and van Loon (1997) NAO paper “Decadal variations in climate ASSOCIATED with the North Atlantic Oscillation” [my emphasis added]. So the water isotopes are clearly telling us something useful about NAO variability and its impacts over the NH, but the reconstruction from them ought not be considered an unbiased expression of the NAO alone.
Considerable discussion is made about the somewhat intermittent multi-decadal variability indicated in the wavelet spectra. While there is little doubt that this could reflect true variations in the strength of natural forcing at these timescales, no mention is made of the likelihood that the dating of some of the annual water isotope records is very likely to degrade back in time. (Certainly for speleothems, which are never precisely dated, and should only be expected to reflect lower frequency variability.) This is known to be a problem with ice core records in general and these make up the bulk of the longer records used in the NAO reconstruction (see Figs. 1 and 2b). The result will almost certainly be a loss of high-frequency signal and a relative increase of lower-frequency power in the composite reconstruction as the dating errors accumulate and the composite consequently smooths back in time. This is actually suggested in Fig. 2a by the visible reduction in the amplitude of variability before 1700 CE. I also note that the wavelet spectrum of the NAO reconstruction based on glacier ice only (Fig. S5a) shows an almost total loss of power a periods <5 years before 1790 whereas the reconstruction based on better dated wood cellulose records maintain it better. This could reflect a gradual loss of dating fidelity back in time in the ice cores. This said, in no way I am suggesting that the loss of precise dating invalidates the usefulness of the reconstruction at lower frequencies, but it should be acknowledged as another likely contributor to the changing pattern of variability seen in the wavelet spectra.
Given these concerns, in order to recommend publication I would need the authors to: