Mapping near surface global marine ecosystems through cluster analysis of environmental data

Abstract

Almost all classifications of the world ocean are based on expert opinion or ad hoc management areas. A quantitative analysis of environmental variables may provide a more objective basis for mapping and classifying the oceans to support data management, reporting, and conservation efforts. Here, we used long-term averages of 20 ocean variables to classify the ocean surface waters using PCA and k-means clustering. We identified seven distinct areas that fit the definition of “ecosystems,” that is, enduring regions demarcated by environmental characteristics. Of all the variables, temperature had the greatest importance and correlated with many other variables. However, some variables had uniquely significant effects on the classification, namely slope, surface current, pH, and wave height. Thus, while the present classification is robust for available data, future analyses with variables not presently available may improve it. How the ecosystems correlate with species richness, endemicity, or abundance will inform on the factors that most influence species' abundance, and thus support global modeling of the effects of climate change, for example, with regard to biological carbon fluxes.