Estimating species richness from quadrat sampling data: a general approach

Jérôme Dupuis (1) , Goulard Michel (2)

1 LSP - Laboratoire de Statistique et Probabilités
2 DYNAFOR - Dynamiques Forestières dans l'Espace Rural

We consider the problem of estimating the number of species (denoted by S) of a biological community located in a region composed of $n$ quadrats. To address this question, different parametric approaches have been recently developed. However, they all have some limitations which reduce their use in practice: indeed, either they presuppose that the sampled quadrats are taken from a large population of quadrats (theoretically infinite), or they require an upper bound on S. Our approach is more general in that it applies without limitation on $n$ and it can be used in the presence of prior information on S, as well as in totally unknown regions. We pay attention to the prior adopted for S; in particular, different non informative priors are considered and motivated. We first consider a simple model which assumes that occurrence and detectability parameters do not depend on quadrats. It constitutes a suitable framework to clarify the links existing between current approaches and ours. We then extend this model by assuming that the region of study is spatially heterogeneous. We illustrate our approach by estimating the number of species of a birds community located in a forest.

Mots clés: Bayesian estimate; Biodiversity; Jeffreys prior; Missing data; Quadrat sampling; Species richness; Number of species; Species richness.

Domaines: Applications [stat.AP]

Jérôme Dupuis, Goulard Michel. Estimating species richness from quadrat sampling data: a general approach. 2010. ⟨hal-00503260⟩