Step-by-step process to a posteriori cluster the species in functional groups based on an ecosystem function.

<p>(A) Species are a priori clustered in functional groups (here σ = 3 as an example). (B) Each ecosystem is then described by a combination of functional groups of species, that is an assembly motif (here <i>m =</i> 2^σ-1 = 7 possible assembly motifs). (C) Ecosystems described by the same assembly motif are grouped in a given ecosystem clusters (<i>k</i> = 1 …7). (D) The goodness-of-fit of ecosystem classification is evaluated by its residual sum of squares (<i>RSS</i>_{<i>modelling</i>}), that can be expressed as a coefficient of determination <i>R</i>^<i>2</i>. The ecosystem classification is compared with others, rejected if worse but selected if better than others. (E) The process is iteratively repeated as long as a species clustering can generate an ecosystem classification better than all others. The process a posteriori selects the species clustering that best accounts for the observed ecosystem functioning.