A Comparative Analysis of Evenness Index Sensitivity

Authors

  • Jean-Nicolas Beisel,

    1. UPRES “Ecotoxicité, Biodiversité et Santé Environnementale”, Equipe de Démoécologie, Université de Metz, UFR SciFA – Campus Bridoux, Av. du Général Delestraint, 57070 METZ, France
    Search for more papers by this author
  • Philippe Usseglio-Polatera,

    1. UPRES “Ecotoxicité, Biodiversité et Santé Environnementale”, Equipe de Démoécologie, Université de Metz, UFR SciFA – Campus Bridoux, Av. du Général Delestraint, 57070 METZ, France
    Search for more papers by this author
  • Vincent Bachmann,

    1. UPRES “Ecotoxicité, Biodiversité et Santé Environnementale”, Equipe de Démoécologie, Université de Metz, UFR SciFA – Campus Bridoux, Av. du Général Delestraint, 57070 METZ, France
    Search for more papers by this author
  • Jean-Claude Moreteau

    1. UPRES “Ecotoxicité, Biodiversité et Santé Environnementale”, Equipe de Démoécologie, Université de Metz, UFR SciFA – Campus Bridoux, Av. du Général Delestraint, 57070 METZ, France
    Search for more papers by this author

Abstract

Evenness indices are numerous but the lack of knowledge of their properties is a limitation to their biological usefulness. 15 evenness indices, two of them being recently proposed, were studied in this work. We investigated the sensitivity of each index using (1) 189 macroinvertebrate communities sampled in the field, and (2) a set of communities modified in a controlled way.

There is no single way to measure evenness. We demonstrated that a measure should be chosen considering (1) the kind of data analysed and, (2) the index properties wanted by users. Depending on the ecological data set the index should be more sensitive to variations in rare, median or abundant species. For most of macroinvertebrate community analyses, a convenient evenness index requirs to be symmetric, unsensitive to variation on rare taxa, with a large range of variation and can be compared with a diversity index. Depending on diversity measure used, five indices correspond to these criteria: EPielou, EHurlbert, E–ln(D), E1–D, EMI. Our results were summarized in a table which may help users to select a convenient evenness measure according to their specific data. Concerning index properties, three main features that an evenness index should meet are briefly discussed: dependence with richness, symmetry criteria and variation range. A revue of this controversial subject allowed a best understanding of values obtained with evenness measures depending on their own features.

Ancillary