Weighted averaging of species indicator values: Its efficiency in environmental calibration (Q1073529)

In this paper the properties of weighted averaging of indicator values to estimate the value of a continuous environmental variable at a site are studied. The authors do this by seeking conditions under which weighted averaging compares favorably with methods based on explicit response curves. Assumptions that idealize the real world are used, among others that a single environmental variable determines the species composition at a site and that the response curves of the species with respect to this variable are already known. The following two questions are considered: (1) How should indicator values of species be defined in terms of response curves to ensure that the weighted average is a consistent estimator? (2) What should the response curves look like to ensure that the weighted average is an efficient estimator? Conditions are derived under which the weighted average is a consistent and efficient estimator for the value of an environmental variable at a site. This paper shows that a method proposed and used by community ecologists, namely weighted averaging, performs well under a model advocated by evolutionary ecologists, namely the species packing model. The results of this paper can be extended to provide a theoretical basis of the model that is implicitly invoked when reciprocal averaging is used.