High dimensional model-based clustering of european georeferenced vegetation plots

An important challenge in complex vegetation systems is the classification of vegetation since it represents a useful tool for summarizing our knowledge of vegetation patterns and, consequently, for nature conservation, landscape mapping and land-use planning. It typically requires standard clustering methods that are capable of identifying groups of plots characterized by dominant and diagnostic species. When the data are high-dimensional, however, efficient clustering methods have to be considered. In this paper, we consider a robust model-based clustering, called Gaussian mixture models for high-dimensional data (HD-GMM) which takes into account for the specific subspace around which each cluster is located and, consequently, provides parsimonious modeling. Results are encouraging and deserve further discussion.