NEW UNSUPERVISED CLASSIFICATION OF THE MARTIAN SURFACE [poster presentation]

Eriita Jones
Research School of Astronomy and Astrophysics, Australian Na

Graziella Caprarelli
Department of Environmental Sciences, University of Technology Sydney

Franklin Mills
Fenner School of Environment and Society, Australian National University

Bruce Doran
Fenner School of Environment and Society, Australian National University

Jonathan Clarke
Mars Society Australia Inc.

     Full text: Not available
     Last modified: May 18, 2011

Abstract
*poster presentation
Maps of thermal inertia – albedo units provide information on the distribution of surface materials on Mars. Previous work has used deterministic methods to threshold the dominant values in thermal inertia and albedo from the Thermal Emission Spectrometer (TES), producing a map of 7 thermo-physical units. Unsupervised classification algorithms define clusters based on groups of pixels with similar values in multiple variables. The classification algorithms are not biased towards the most distinctive thermal inertia and albedo information and hence, they reveal substantial structure in medium - high thermal inertia materials that were not seen in previous works. In this work we identified 10 thermo-physical clusters. We interpreted the distribution of thermal inertia and albedo within the clusters, and constrained the relative fractions of key surface component – dust, sand and duricrust, bedrock and ice – within each cluster. We also constrained grain sizes and layering within the top ~ 2 cm of each cluster. Our mapping of surface materials is consistent with the locations of high latitude ice and the boundaries of some geologic units, and suggests the presence of low-latitude ice-cemented terrain.