Collecting ‘Big’-data in sedimentary provenance analysis: An optimized workflow from sample preparation to analysis

For a robust interpretation in sedimentary provenance analysis studies (SPA) a combination of multiple methods is usually applied to a selected number of samples. To circumvent effects that perturb the provenance signal (e.g. hydraulic sorting) information on radiometric age, chemical composition and mineralogy is collected for mineral varieties usually by means of laser ablation inductively coupled mass spectrometry (LA-ICPMS), electron probe microanalyzer (EPMA), Raman micro-spectroscopy and polarized optical microscopy.

These methods have become increasingly efficient and allow for rapid analysis of statistically relevant numbers of samples which is fundamental to SPA. However, routine combination of these methods on the same grains is rarely realized and sample preparation quickly becomes a bottleneck when sample numbers are significantly increased. The latter is especially important to detect subtle variations in deposits due to processes operating on centennial to millennial time scales such as rapid climatic variability.

Here we present a workflow that is optimized for high throughput of silt to sand-sized sedimentary samples, which allows routine combination of optical microscopy, Raman micro-spectroscopy, EPMA and LA-ICPMS by means of machine learning methods. Due to the high degree of automatization our workflow enables to access sedimentary archives at high spatial and/or temporal resolution and will provide, depending on combined methods, single-grain datasets that contain information on grain-size, shape, roundness, color, mineralogy, degree of metamictization, chemical composition, trace element composition and radiometric age. We demonstrate innovative approaches in the relevant sample preparation steps and showcase data of several loess profiles highlighting the feasibility of our workflow.