Areas and algorithms: evaluating numerical approaches for the delimitation of areas of endemism in the Canary Islands archipelago

Authors

  • Mark A. Carine,

    Corresponding author
    1. Department of Botany, The Natural History Museum, London, UK
    Search for more papers by this author
  • Christopher J. Humphries,

    1. Department of Botany, The Natural History Museum, London, UK
    Search for more papers by this author
  • I. Rosana Guma,

    1. Unidad de Botánica Aplicada, Instituto Canario de Investigaciones Agrarias, Jardín de Aclimatación de La Orotava, Puerto de La Cruz, Santa Cruz de Tenerife, Spain
    Search for more papers by this author
  • J. Alfredo Reyes-Betancort,

    1. Unidad de Botánica Aplicada, Instituto Canario de Investigaciones Agrarias, Jardín de Aclimatación de La Orotava, Puerto de La Cruz, Santa Cruz de Tenerife, Spain
    Search for more papers by this author
  • Arnoldo Santos Guerra

    1. Unidad de Botánica Aplicada, Instituto Canario de Investigaciones Agrarias, Jardín de Aclimatación de La Orotava, Puerto de La Cruz, Santa Cruz de Tenerife, Spain
    Search for more papers by this author

*Mark Carine, Department of Botany, The Natural History Museum, Cromwell Road, London SW7 5BD, UK. E-mail: m.carine@nhm.ac.uk

Abstract

Aim  Areas of endemism are the fundamental units of cladistic biogeographical analysis but there is no consensus on the most appropriate method for their delimitation. In this paper, the relative performance of a number of algorithmic approaches for the delimitation of areas of endemism is investigated within the context of the Canary Islands flora, and areas of endemism within the Canary Islands archipelago are defined.

Location  The Canary Islands.

Methods  A data matrix comprising the distributions of 609 endemic spermatophyte taxa (c. 90% of the endemic flora) scored on a 10 × 10 km UTM grid was analysed using: (1) UPGMA (unweighted pair group method with arithmetic mean) clustering of Jaccard and Kulczynski similarity coefficient matrices, (2) parsimony analysis of endemicity (PAE), and (3) the program ndm (eNDeMism). The performance of each method was then determined by the extent to which the resulting areas of endemism met three criteria: (1) possession of two or more strict endemic taxa, (2) diagnosability, and (3) geographical contiguity.

Results  Each of the four methods resulted in substantially different sets of areas. ndm analysis resolved 17 areas of endemism consistent with all three criteria, and collectively these accounted for 59% of all cells. In the hierarchical analyses none of the methods recovered more than eight areas of endemism, and the total coverage of cells ranged from 13% to 33% when the results were confined to intra-island areas of endemism.

Main conclusions ndm outperforms hierarchical clustering methods in terms of both the number of intra-island areas of endemism delimited that meet the three evaluation criteria and the total coverage of those areas. ndm may also be considered preferable because it is non-hierarchical, incorporates spatial information into the delimitation of areas, and permits overlap between areas of endemism where there is evidence to support it. The results support the use of ndm as the most appropriate method currently available for the delimitation of areas of endemism. The areas of endemism identified by the ndm analysis are discussed.

Ancillary