A data-driven approach to understanding esker morphogenesis

Eskers are ubiquitous features on previously glaciated landscapes, recording the configuration and dynamics of the channelized meltwater system. Studies of esker composition and form have resulted in a variety of genetic interpretations surrounding the ice, water, and sediment characteristics under which they may develop. However, issues of apparent equifinality currently limit the usefulness of eskers for reconstructing broad-scale glacial hydrology. Although some authors have attempted to asses esker morphogenesis, previous studies are limited by their small sample size and/or use of qualitative morphometric indices.

This project aims to explore whether eskers have a distinct morphogenetic signature using data science techniques. Published research has been mined for empirical studies of esker composition and structure. These data were compiled into a database summarizing the genetic interpretations commonly invoked for eskers (e.g., depositional environment, meltwater flow regime) as well as the supporting evidence for such inferences (e.g., sedimentary logs). Semi-automated methods will be tested to map eskers from high resolution (1-2 metres) LiDAR digital terrain models and to extract their morphometry. A range of planform- and profile-scale morphometric indices will be employed and new indices that can more precisely quantify esker morphometry will be developed.

The resulting highly-dimensional dataset can be analyzed using machine learning techniques in order to assess the relationships between sedimentologic, morphometric, and genetic variables. Preliminary results from database development and analysis will be presented and methodological concerns will be discussed.