Double the dates and go for Bayes — Impacts of model choice, dating density and quality on chronologies

Maarten Blaauw, J. Andres Christen, K. D. Bennett, Paula J. Reimer

Research output: Contribution to journalArticlepeer-review

129 Citations (Scopus)
535 Downloads (Pure)

Abstract

Reliable chronologies are essential for most Quaternary studies, but little is known about how age-depth model choice, as well as dating density and quality, affect the precision and accuracy of chronologies. A meta-analysis suggests that most existing late-Quaternary studies contain fewer than one date per millennium, and provide millennial-scale precision at best. We use existing and simulated sediment cores to estimate what dating density and quality are required to obtain accurate chronologies at a desired precision. For many sites, a doubling in dating density would significantly improve chronologies and thus their value for reconstructing and interpreting past environmental changes. Commonly used classical age-depth models stop becoming more precise after a minimum dating density is reached, but the precision of Bayesian age-depth models which take advantage of chronological ordering continues to improve with more dates. Our simulations show that classical age-depth models severely underestimate uncertainty and are inaccurate at low dating densities, and perform poorly at high dating densities. On the other hand, Bayesian age-depth models provide more realistic precision estimates, including at low to average dating densities, and are much more robust against dating scatter and outliers. Indeed, Bayesian age-depth models outperform classical ones at all tested dating densities, qualities and time-scales. We recommend that chronologies should be produced using Bayesian age-depth models taking into account chronological ordering and based on a minimum of 2 dates per millennium.
Original languageEnglish
Pages (from-to)58-66
Number of pages9
JournalQuaternary Science Reviews
Volume188
Early online date06 Apr 2018
DOIs
Publication statusPublished - 15 May 2018

Fingerprint

Dive into the research topics of 'Double the dates and go for Bayes — Impacts of model choice, dating density and quality on chronologies'. Together they form a unique fingerprint.

Cite this