Inferring plant ecosystem organization from species occurrences
In this paper, we present an approach capable of extracting insights on ecosystem organization from merely occurrence (presence/absence) data. We extrapolate to the collective behavior by encapsulating some simplifying assumptions within a given set of constraints, and then examine their ecological implications. We show that by using the mean occurrence and co-occurrence of species as constraints, one is able to capture detailed statistics of a plant community distributed across a vast semiarid area of the United States. The approach allows us to quantify the species' effective couplings: Their frequencies exhibit a peak at zero and the minimal pairwise model is able to capture about 80% of the ecosystem structure. Our analysis reveals a relatively stronger impact of the species network on uncommon species and underscores the importance of species pairs experiencing positive couplings. Additionally, we study the associations among species and, interestingly, find that the frequencies of groups of different species, which the approach is able to capture. exhibit a power-law-like distribution. (C) 2009 Elsevier Ltd. All rights reserved.
Keywords: Maximum entropy ; Ising model ; Power law ; Ecological interactions ; Occurrence data ; Species associations ; Null Model Analysis ; Maximum-Entropy ; Statistical-Mechanics ; Information Theory ; Patterns ; Facilitation ; Communities ; Biodiversity ; Networks
Record created on 2011-01-28, modified on 2016-08-09