ASSESSING THE EFFICACY OF CLUSTERING TECHNIQUES FOR AIDING REGIONAL SYNTHESIS OF QUATERNARY GEOLOGY

The correlation of map units and contacts between adjacent geologic maps can impede the characterization of natural resources or the quantification of geologic processes. An objective of the National Geologic Synthesis project is to compile a Quaternary geologic map of the conterminous United States based on publicly available datasets, which necessitates the development of consistent generalized map units that can describe the variable contents of the source data.

Here we present a first-order attempt at evaluating and compiling Quaternary geologic units based on objective methods for clustering, a strategy common in data science. As inputs to these algorithms, we rely on previously classified Quaternary geology units, soils data, and elevation data and derivatives. Our workflow groups previously map units from various sources into compilation map units based on common attributes such as unit names or geologic descriptions. We then use gridded SSURGO soil data and 3DEP elevation data as inputs for an isodata clustering algorithm to determine the characteristics of natural groupings of cells within a given study region. These distinct characteristics are then passed to a maximum likelihood classification tool where a classified raster is output. We then iterate on these steps to refine assignments of synthesis map units where appropriate to minimize notable anomalies between data-driven clusters and existing observations, resulting in a semi-objective approach to identify robust regional Quaternary geologic map units.

We demonstrate this approach on a longitudinal transect of the United States spanning the transition from the Rocky Mountains to the Great Plains. State and regional geologic maps of Quaternary geology in this area have major gaps in the availability of mapping and show large discontinuities in the geologic units present in the maps (e.g., most notably in the presence and classification of older alluvium and of eolian units). However, the efforts presented here show promise in negating some of those biases for regional Quaternary geologic map units.