Articles | Volume 13, issue 9
Clim. Past, 13, 1169–1180, 2017
Clim. Past, 13, 1169–1180, 2017

Research article 19 Sep 2017

Research article | 19 Sep 2017

A complete representation of uncertainties in layer-counted paleoclimatic archives

Niklas Boers1,2,5, Bedartha Goswami3, and Michael Ghil1,4 Niklas Boers et al.
  • 1Geosciences Department and Laboratoire de Météorologie Dynamique (CNRS and IPSL), École Normale Supérieure, Paris, France
  • 2Potsdam Institute for Climate Impact Research, Potsdam, Brandenburg, Germany
  • 3Institute of Earth and Environmental Science, University of Potsdam, Potsdam, Germany
  • 4Department of Atmospheric and Oceanic Sciences and Institute of Geophysics and Planetary Physics, University of California, Los Angeles, USA
  • 5PSL Research University, Paris, France

Abstract. Accurate time series representation of paleoclimatic proxy records is challenging because such records involve dating errors in addition to proxy measurement errors. Rigorous attention is rarely given to age uncertainties in paleoclimatic research, although the latter can severely bias the results of proxy record analysis. Here, we introduce a Bayesian approach to represent layer-counted proxy records – such as ice cores, sediments, corals, or tree rings – as sequences of probability distributions on absolute, error-free time axes. The method accounts for both proxy measurement errors and uncertainties arising from layer-counting-based dating of the records. An application to oxygen isotope ratios from the North Greenland Ice Core Project (NGRIP) record reveals that the counting errors, although seemingly small, lead to substantial uncertainties in the final representation of the oxygen isotope ratios. In particular, for the older parts of the NGRIP record, our results show that the total uncertainty originating from dating errors has been seriously underestimated. Our method is next applied to deriving the overall uncertainties of the Suigetsu radiocarbon comparison curve, which was recently obtained from varved sediment cores at Lake Suigetsu, Japan. This curve provides the only terrestrial radiocarbon comparison for the time interval 12.5–52.8 kyr BP. The uncertainties derived here can be readily employed to obtain complete error estimates for arbitrary radiometrically dated proxy records of this recent part of the last glacial interval.

Short summary
We introduce a Bayesian framework to represent layer-counted proxy records as probability distributions on error-free time axes, accounting for both proxy and dating errors. Our method is applied to NGRIP δ18O data, revealing that the cumulative dating errors lead to substantial uncertainties for the older parts of the record. Applying our method to the widely used radiocarbon comparison curve derived from varved sediments of Lake Suigetsu provides the complete uncertainties of this curve.