Testing the accuracy of single-grain OSL dating on Eemian quartz samples

Single-grain OSL dating of quartz is a popular approach to OSL dating, even when incomplete bleaching is not likely to be significant. However, little testing of the accuracy of single-grain dating has been published; particularly for samples older than 50 ka. In this study, we investigate the accuracy of single-grain quartz OSL dating, when a significant number of individual grains are no longer able to accurately measure the burial dose because of saturation effects. We compare standard multi-grain OSL results with those obtained from single-grain OSL measurements for five OIS substage 5e (Eemian) samples (similar to 128 ka). We show that for these samples, standard multi-grain quartz dose estimation results in dose estimates in good agreement with the predicted doses (four of the five samples recover age control), but that standard frequentist single-grain dating procedures significantly underestimate the age controls, i.e. the measured to predicted dose range between 0.42 +/- 0.03 and 0.84 +/- 0.06, where the underestimation increases with increasing relative number of grains in saturation. Attempting to remove the inevitable bias in the dose estimation resulting from a significant number of saturated grains (by using the D-c criterion) reduced the underestimation, i.e. the measured to predicted dose ratio range between 0.63 +/- 0.05 and 0.94 +/- 0.08, but only the sample with the smallest absorbed dose is consistent with the age control. Using Bayesian analysis ("BayLum") the ratio of measured to predicted dose range between 0.75 +/- 0.07 and 1.14 +/- 0.08, but only two of the five samples agree with the independent age control. Our results have implications for the evaluation of single-grain OSL dating of quartz in the 100-200 Gy natural dose range.