Estimating biases during detection of leads and lags between climate elements across Dansgaard–Oeschger events

Abstract

Dansgaard–Oeschger (DO) events occurred throughout the last glacial period. Greenland ice cores show a rapid warming during each stadial to interstadial transition, alongside an abrupt loss of sea ice and a major reorganisation of the atmospheric circulation. Other records also indicate simultaneous abrupt changes to the oceanic circulation. Recently, an advanced Bayesian ramp-fitting method has been developed and used to investigate time lags between transitions in these different climate elements with a view to determining the relative order of these changes. Here, we critically review this method in both its original implementation and a new, extended implementation. Using ice core data, climate model output, and carefully synthesised data representing DO events, we demonstrate that both implementations of the method suffer from biases of up to 15 years. These biases mean that the method will tend to yield transition onsets that are too early. Further investigation of DO warming event records in climate models and ice core data reveals that the biases are on the same order of magnitude as potential timing differences between the abrupt transitions of different climate elements. Additionally, we find that higher-resolution records would not reduce these biases. We conclude that decadal-scale leads and lags between climate elements across DO events cannot be reliably detected, as we cannot exclude the possibility that they result solely from the biases we present here.