Using sediment texture and geochemistry as predictors to automatically classify sub-depositional environments in a modern estuary

Sedimentary cores from the Ravenglass Estuary, NW England, lack some of the sedimentary structures which can be seen in other estuarine sands due to their unconsolidated nature, making it difficult to meaningfully interpret depositional environments using standard sedimentological facies analysis. Here we explore how sediment texture, obtained from laser particle size analysis, and bulk geochemistry, obtained from portable X-ray fluorescence, can be used independently, or in combination, within a machine learning model to automatically classify sub-depositional environment and estuarine zone at the estuary surface to create a model which can be used to classify subsurface sediment samples. Using an adapted an established machine learning workflow we select the most informative geochemical elements to be included in the training set for the classification model.

The most important geochemical elements for modelling represent major elements of the most abundant minerals in the estuary, and minor elements representing trace signals of sulfide mineral deposits present in the hinterland. Models that are trained exclusively on textural data significantly outperform those that use geochemical data when classifying sub-depositional environment but are comparable when classifying estuarine zone. However, the combination of textural and geochemical data in training sets improves model performance in all but one class when compared to separate textural and geochemical models.

Ultimately, we have applied the surface-calibrated combined textural and geochemical model to classify paleo-sub-depositional environment in geotechnical cores obtained from the Ravenglass Estuary. This allows us to interpret their environmental evolution and build scaled correlation panels spanning key areas of the estuary to suggest how the estuary has changed since its formation.