PROCESSING DATA AND INCORPORATING UNCERTAINTIES IN LARGE GEOCHEMICAL COMPILATIONS

Large geochemical datasets---which comprise both published and contributed observations---can provide insights into surficial and crustal processes on Earth. However, extracting meaningful trends from these datasets is not always straightforward, due in part to the fact that each observation has temporal, spatial, and analytical uncertainties. The methods by which researchers choose to filter, resample, and statistically evaluate datasets can impact the results and any subsequent interpretations. As such, it is important for scientists to consider how data are compiled, processed, and analyzed in large geochemical compilations.

Here, we present a workflow using the Sedimentary Geochemistry and Paleoenvironments Project (SGP) database. In its current state, the SGP database contains 82,579 samples, each with associated geochemical analytes and metadata. Compiled from three main sources, the SGP database is globally distributed and covers sedimentary deposits across most of Earth’s history. Using our workflow, we examine Al₂O₃ and U, two geochemical components that behave differently in sedimentary environments. In doing so, we demonstrate how filtering and weighted resampling impact the resulting global trends. Through our application, we present several challenges associated with using data compilations and provide suggestions for how to deal with these issues. We propose that, with slight modifications to our workflow, researchers can confidently use large geochemical datasets to gain a better understanding of Earth history.