Discriminating luminescence age uncertainty composition for a robust Bayesian modelling

• Verfasst von: Zeeden, Christian [VerfasserIn]
• Verfasst von: Dietze, Michael [VerfasserIn]
• Verfasst von: Kreutzer, Sebastian [VerfasserIn]
• Titel: Discriminating luminescence age uncertainty composition for a robust Bayesian modelling
• Verf.angabe: Christian Zeeden, Michael Dietze, Sebastian Kreutzer
• E-Jahr: 2018
• Jahr: February 2018
• Umfang: 10 S.
• Fussnoten: Online veröffentlicht: 10. Oktober 2017, Artikelversion: 16. Oktober 2017 ; Gesehen am 26.03.2025
• Titel Quelle: Enthalten in: Quaternary geochronology
• Ort Quelle: Amsterdam [u.a.] : Elsevier, 2006
• Jahr Quelle: 2018
• Band/Heft Quelle: 43(2018) vom: Feb., Seite 30-39
• ISSN Quelle: 1871-1014
• DOI: doi:10.1016/j.quageo.2017.10.001
• Datenträger: Online-Ressource
• Sprache: eng
• Sach-SW: Bayesian age-depth model
• Sach-SW: Bayesian statistics
• Sach-SW: Loess
• Sach-SW: Luminescence dating
• Sach-SW: Quartz
• Sach-SW: R
• Sach-SW: Uncertainty discrimination
• K10plus-PPN: 1920605797

Abstract

Age-depth-relationships are essential to understand expressions of Earth history. Age-depth-relationships reveal the environmental significance of (terrestrial) sediment deposits and relate them to other paleoenvironmental archives. Although luminescence-based ages are of paramount importance, their incorporation in age-depth-models is limited due to the convolved uncertainties with unknown proportions of systematic and random parts. Here, we use an inverse modelling approach for fine-grained quartz OSL ages of loess deposits to (i) generate a probability density function of the random uncertainty part and (ii) use this function to recalculate and improve stratigraphically related OSL ages. Our approach can be applied to other luminescence dating techniques. Applying the algorithm to synthetic data sets generates valid results, despite an average underestimation of the relative random uncertainty by ∼18%. All tests show that precision of the inverse modelling of uncertainty parts depends only on the number of samples per dataset. By using the output of the approach on published data the precision of a late Pleistocene loess sequence chronology was improved by 21-30%.