A Compositional Approach to Micropalaeontological Ecozonation using Divisive Hierarchical Clustering

Ecozonation methodologies are widely adopted, in both continental and marine successions, to define high-resolution bio- and- chronostratigraphic schemes. The concept behind this approach is that changes in the composition of fossil assemblages are proxies of palaeoenvironmental or palaeoclimatic changes. Thus, the definition of ecozones make possible the identification of intervals characterised by relative stability or, by converse, stratigraphical levels which marks environmental changes. The definition of ecozones may be strengthened by adopting stratigraphically constrained cluster algorithms (i.e. samples can only be grouped with samples or clusters immediately preceding or following them), which allow ecozones boundaries to be objectively defined. Among them, the stratigraphically constrained method CONISS has been widely adopted in micropalaeontological studies. The application of CONISS to micropalaeontological data expressed in terms of percentages requires an approach conform to the nature of the Compositional Data (CoDa). On these bases, Di Donato et al., (2008) defined compositional intervals based on the CONISS method applied to logratio coordinates. The CONISS method is an aggregative hierarchical clustering algorithm. Here, we explore a divisive constrained approach to ecozonation developed in a field, that of micropalaeontological associations, that is well suited to a CoDa approach.