An algorithm for clustering profile data and its application to near-surface ice content data from wet coastal tundra soils near Barrow, Alaska

Abstract

An algorithm to cluster profile data into groups that minimize the sum of the intra-group variances was applied to near-surface soil ice content data collected near Barrow, Alaska, in wet tundra terrain. When the algorithm was requested to produce 2–5 groups and group mean profiles, the results were consistant with the modern theory of ice segregation. This process produces much of the variability of near surface soil ice stratigraphy in nature. These results strengthen the case for employing the algorithm on other profile data sets as an aid in hypothesis formulation.