Interpolating proxy observations to high-resolution XRF data

Proxy observations are the fundament for many insights in geosciences, such as the reconstruction of past environmental conditions from sediment cores. However, the labor-intensive nature of producing proxy data often renders high-resolution records prohibitively expensive. In contrast, state-of-the-art XRF analyses offer an efficient means of generating high-quality and very high-resolution elemental concentration data. Although these data initially provide only qualitative information about the bulk sample composition, their high-resolution renders them invaluable as a foundation for subsequent studies.

Here we explore the potential utility of high-resolution XRF data as a foundation for interpolating more scarce sedimentological, geochemical, and environmental analyses to the same high-resolution. We base the investigation on a long sediment core obtained from the Crozet Plateau in the Indian sector of the Southern Ocean, where changes in surface productivity, dust influx, and delivery of detritus from the nearby island dominate the sedimentological and geochemical signals.

We test different statistical methods for sediment classification and the interpolation of proxy data and discuss their reliability and limitations. For instance, application of a random forest model for the interpolation of carbonate and opal concentration conspicuously reveals the presence of distinct detrital layers. These layers are not readily discernible in the original XRF data or the sparse original opal and carbonate measurements.

We propose these straight-forward statistical methods have potential to be used on many other sediment records for which high-quality and high-resolution XRF data and certain proxy observations are available. Such an approach can streamline the interpretation of XRF data and enhance the investigation of specific environmental proxy observations at a high-resolution.