An empirical study on the variability of luminescence ages for coeval sediment samples

Abstract

Optically stimulated luminescence (OSL) dating is a widely used numerical-age technique in Quaternary research. OSL age datasets at high vertical resolution are increasingly generated in order to quantify rates of processes or to identify gaps in the sediment record. However, incorporation of luminescence ages in age-depth models or age comparisons is not straightforward due to the variety and complex combination of uncertainties associated with luminescence dating. Moreover, previous studies reported on a variability in age results for coeval loess samples that is significantly larger than expected and remains to be understood.

In this study we aim to provide the first empirical assessment of the variability in the OSL ages in coeval sediment samples. Multiple samples have been collected at high horizontal resolution, from loess and an aeolian sand layers that bracket Campanian Ignimbrite/Y-5 tephra (⁴⁰Ar/³⁹Ar dated to 39.9 ± 0.1 ka), in southern Romania, southeastern Europe. All individual 63–90 μm quartz OSL ages are stratigraphically consistent with this independent age control within 1 σ or 2 σ uncertainty. There is a 1.7 % scatter in the individual ages in the sand layer and 4.6 % in the loess layer.

The various sources of uncertainty are detailed and we observe that the random sources of uncertainty have a limited contribution to the total uncertainty of the age and the systematic sources of uncertainty dominate. Weighted average ages of 46 ka and 39 ka are calculated for the sediment layers below and above the tephra, respectively. The overall random uncertainty of the weighted OSL ages (1.1 % sands, 1.6 % loess) is similar to the observed random uncertainty among individual OSL ages (0.8 % sands, 0.9 % loess) which indicates that the sources of variability in the OSL ages are properly accounted.

By averaging ages on multiple coeval samples, the precision (random uncertainty) of the age of the sedimentary context was improved with up to 1.9 % and 3.8 % in the sand and loess layers, respectively, compared to that on individual ages. Such an increase in precision of the age may lead to improvements in the precision of relative age-frameworks based on luminescence dating. However, averaging ages of coeval samples has a small effect on the overall uncertainty associated to the average weighted age, owing to a less than 10 % fractional contribution of overall random uncertainties to the total age uncertainty budget. Our empirical results show that for the samples investigated here it is not possible to obtain a luminescence age with an overall relative uncertainty of less than 6 %.