A georeferenced implementation of weighted endemism

Abstract

<jats:title>Summary</jats:title><jats:p> <jats:list>

<jats:list-item><jats:p>Areas of endemism are conservation priorities and indicators of biogeographical processes, representing concentrations of range‐restricted biodiversity. While the term endemism describes geographical range restriction, the technical definition has often been re‐worked. Categorical definitions are increasingly replaced by weighting of species according to range sizes. In various contexts, range sizes have been estimated by the number of occupied map grid cells, the latitudinal or longitudinal range of point locations, or the area of a polygon surrounding point locations. Existing implementations of weighted endemism use the number of occupied grid cells to estimate range. However, this represents area of occupancy (<jats:styled-content style="fixed-case">AOO</jats:styled-content>), not extent of occurrence (<jats:styled-content style="fixed-case">EOO</jats:styled-content>).</jats:p></jats:list-item>

<jats:list-item><jats:p>We developed endemism calculations that include strictly geographical (georeferenced) interpretations of range (‘span’ and ‘area’), and compared them to cell frequency‐based measures. We compared species weights and endemism scores for 330 004 records of 3083 vascular plant species in 14 328 plots (Biological Survey of South Australia). We provide a self‐contained R function that calculates endemism using alternative range weights and integrates point data with automatically generated rasters. A parallel function tests for deviance from the expected (null) distribution, given observed species richness (nonparametric significance test; outlier metrics).</jats:p></jats:list-item>

<jats:list-item><jats:p>Species range weights based on grid cell frequency (<jats:styled-content style="fixed-case">AOO</jats:styled-content>) versus georeferenced ‘span’ and ‘area’ (<jats:styled-content style="fixed-case">EOO</jats:styled-content>) are conceptually different and were poorly correlated for the same data set, resulting in differences in the classification of grid cells as outliers from expected endemism. For example, the correlation (Kendall's τ) between endemism based on geographical ‘span’ and cell frequency was just 0·57, when partialling out the influence of species richness, potentially leading to contrasting conservation priorities over 8% or 46 000 km<jats:sup>2</jats:sup> of the state of South Australia.</jats:p></jats:list-item>

<jats:list-item><jats:p>Weighting endemism by georeferenced range estimates provides an alternative approach that more explicitly highlights restriction in <jats:styled-content style="fixed-case">EOO</jats:styled-content>, generates differences in endemism metrics and requires only georeferenced species records. This approach need not replace existing implementations, but provides an alternative measure of range‐restricted biodiversity. The functions provide efficient calculation of species ranges (with or without geographical outlier exclusion) and weighted endemism with novel implementation of significance tests and outlier detection, in the R environment.</jats:p></jats:list-item> </jats:list> </jats:p>