ChronRater: A simple approach to assessing the accuracy of age models from Holocene sediment cores

We have assembled a database of previously published Holocene proxy climate records from the Arctic, with the goal of reconstructing the spatial-temporal pattern of past climate changes. The focus is on well-dated, highly resolved, continuous records that extend to at least 6 ka, most of which (90%) are from sedimentary sequences sampled in cores from lakes and oceans. The database includes the original geochronological data (radiocarbon ages) for each record so that the accuracy of the underlying age models can be assessed uniformly. Determining the accuracy of age control for sedimentary sequences is difficult because it depends on many factors, some of which are difficult to quantify. Nevertheless, the geochronological accuracy of each time series in the database must be assessed systematically to objectively identify those that are appropriate to address a particular level of temporal inquiry. We have therefore devised a scoring scheme to rate the accuracy of age models that focuses on the most important factors and uses just the most commonly published information to determine the overall geochronological accuracy. The algorithm, "ChronRater" is written in the open-source statistical package, R. It relies on three characteristics of dated materials and their downcore trends: (1) The delineation of the downcore trend, which is quantified based on three attributes, namely: (a) the frequency of ages, (b) the regularity of their spacing, and (c) the uniformity of the sedimentation rate. (2) The quality of the dated materials, as determined by: (a) the proportion of outliers and downcore reversals, and (b) the type of materials analyzed and the extent to which their ages are verified by independent information as judged by a five-point scale for the entire sequence of ages. And (3) the overall uncertainty in the calibrated ages, which includes the analytical precision and the associated calibrated age ranges. Although our geochronological accuracy score is numerical, we recognize that judging the quality of material and weighting the various factors that influence accuracy can be subjective. We applied the scoring scheme to more than 110 different downcore age models to assess the distribution of the score and its dependency on the input variables. While no scoring scheme will be perfect, ours can be used to assign reasonable numerical ratings to the reliability of downcore age models based on a simple, reproducible, and customizable procedure that focuses on the most important factors that determine the overall geochronological accuracy.