An exploration of the utility of speleothem age distributions for palaeoclimate assessment

Abstract

Abstract Age distributions of speleothem growth frequency are often employed as proxies for past climate variation. However, their interpretation can be influenced by many variables such as the type of probability density estimator used, the number of samples studied, and indeed the nature of the sampling itself. In order to explore the utility of speleothem age distributions in this context, we constructed a synthetic speleothem growth dataset by randomly generating U-Th ages, based on a known climatic forcing to evaluate whether the most basic climatic information could be recovered by subsequent sampling of this dataset under differing conditions. This modelling approach allows us to separately examine the influences of sample size, the type of probability density estimator, and its parameterisation on the shape of the resulting age distribution. Our model demonstrates that periodic Quaternary climate fluctuations can be recovered from growth frequency distributions derived from as few as 120–150 radiometric ages. To construct such distributions, a kernel density estimator may more accurately represent speleothem growth frequency than the commonly employed probability density plot, because peaks in the latter often reflect data precision rather than data density. For U-Th ages, randomizations of the activity ratios can be used to reintroduce chronometric or analytical uncertainty into the Kernel Density Estimator in a Monte-Carlo simulation.