A modified procedure for mixture-model clustering of regional geochemical data

Applied Geochemistry
By: , and 

Links

Abstract

A modified procedure is proposed for mixture-model clustering of regional-scale geochemical data. The key modification is the robust principal component transformation of the isometric log-ratio transforms of the element concentrations. This principal component transformation and the associated dimension reduction are applied before the data are clustered. The principal advantage of this modification is that it significantly improves the stability of the clustering. The principal disadvantage is that it requires subjective selection of the number of clusters and the number of principal components. To evaluate the efficacy of this modified procedure, it is applied to soil geochemical data that comprise 959 samples from the state of Colorado (USA) for which the concentrations of 44 elements are measured. The distributions of element concentrations that are derived from the mixture model and from the field samples are similar, indicating that the mixture model is a suitable representation of the transformed geochemical data. Each cluster and the associated distributions of the element concentrations are related to specific geologic and anthropogenic features. In this way, mixture model clustering facilitates interpretation of the regional geochemical data.

Study Area

Additional publication details

Publication type Article
Publication Subtype Journal Article
Title A modified procedure for mixture-model clustering of regional geochemical data
Series title Applied Geochemistry
DOI 10.1016/j.apgeochem.2014.10.011
Volume 51
Year Published 2014
Language English
Publisher Elsevier
Contributing office(s) Crustal Geophysics and Geochemistry Science Center
Description 12 p.
First page 315
Last page 326
Country United States
State Colorado
Online Only (Y/N) N
Additional Online Files (Y/N) N