On the reliability of the Elements of Metacommunity Structure framework for separating idealized metacommunity patterns

The Elements of Metacommunity Structure (EMS) framework originally suggested by Leibold and Mikkelson (2002) in Oikos is a popular approach to identify idealized metacommunity patterns (i.e. checkerboard, nested, evenly spaced, Clementsian, Gleasonian), and hereby to infer the existence of structuring processes in metacommunities. Essentially, the EMS framework consists of the rearrangement of the sites-by-species incidence matrix followed by a series of tests for coherence, turnover and boundary clumping in species distributions. Here, we give a critical evaluation of the EMS framework based on theoretical considerations and simulations. We found that user defined site ordering may influence the coherence test (number of embedded absences) depending also on the ordering of species, and therefore we argue that the application of user-defined matrix rearrangement has strong limitations. The recommended ordering by correspondence analysis is sensitive to matrix structure and may even include arbitrary decisions in special cases. Further, we revealed different meanings of the checkerboard pattern and showed that negative coherence is not necessarily associated with this as assumed in the EMS framework. Also, the turnover test cannot always detect nested pattern, because turnover and nestedness are not necessarily the opposite endpoints of a continuum. We argue that the boundary clumping test can only be used for separating Clementsian, Gleasonian and evenly spaced patterns if sites are ordered along a real environmental gradient rather than a latent one identified by correspondence analysis. We found that the series of tests in the EMS framework are burdened by anomalies and that the detection of some metacommunity patterns is sensitive to type II error. In sum, our findings suggest that the analytical methodology of the EMS framework, as well as the conclusions drawn from its application to metacommunity studies require careful reconsideration.