Testing the accuracy of a Bayesian central-dose model for single-grain OSL,using known-age samples

While reviews of comparisons between multi-grain OSL ages and independent chronological information are available in the literature, there is hardly any such performance test for single-grain OSL ages. And yet, this is all the more needed as the interpretation of single-grain dose distributions remains a difficult task, given the typically considerable dispersion in equivalent dose values measured by OSL – and the numerous sources of such dispersion in measurements. Here, we present the study of 19 samples for which independent age control is available, and whose ages range from 2 to 46 ka. Based on multi-grain OSL age estimates, these samples are presumed to have been both well-bleached at burial, and unaffected by mixing after deposition. Two ways of estimating single-grain ages are then compared: the standard approach on the one hand, consisting of applying the Central Age Model to D_e values determined with the Analyst software; on the other hand, the central dose model recently proposed by Combès et al. (Combès, B., Philippe, A., Lanos, P., Mercier, N., Tribolo, C., Guerin, G., Guibert, P., Lahaye, C., in press. Quaternary Geochronology). The median of the relative discrepancy between single-grain OSL and reference ages is about twice as large for the standard approach (12%) as with the Bayesian model (7%). Statistical tests show that, based on our (limited) dataset, the difference between the two models seems to be significant for samples in the age range 4–46 ka. Finally, the influence of various factors on the (in-)accuracy of single grain OSL ages is discussed; it appears that the accuracy of ages estimated in a standard way decreases when age is increased, while the Bayesian model seems more robust. This study also shows that (i) there is no 20% limit on the CAM overdispersion parameter for well-bleached samples; (ii) dose recovery experiments do not seem to be a very reliable tool to estimate the accuracy of a SAR measurement protocol for age determination.