Land snail biogeography and endemism in south-eastern Africa: Implications for the Maputaland-Pondoland-Albany biodiversity hotspot

resulting in a low level of hierarchical structuring in the PAE area cladogram. Hence, with a restricted spatial extent and taxonomic coverage in our study, we did not attempt a regionali- sation exclusively based on PAE as done in previously mentioned studies. However, less struc- turing in our PAE area cladogram does not necessarily negate the hierarchical structure suggested by the UPGMA clustering, as it depends only on the similarity coefficients between each of the OGUs based on entire species (either endemics or not) assemblages. Consequently, we acknowledge that the present dataset indicates limitations in matching the hierarchical out- puts from PAE and UPGMA clustering approaches. However, in order to test the robustness of endemicity findings, we performed a BEA [63], including a parametric bootstrap test for clustering of distribution areas [64]. Similar to PAE, the BEA has also been designed to be used in grid-based equal area OGUs and hence have a lower power in detecting the structuring in our OGUs, as OGUs with varied shapes result in a neighbourhood matrix with uneven numbers of neighbouring units. However, the BEA indi- cated our data also show a certain degree of clustering (with a test statistic lower than expected by the null model, although with a lower level of statistical significance (p = 0.33)). In evidence to the diminished power of BEA when using non-grid based OGUs, our statistical significance was increased to (p = 0.21), by adjusting our neighbourhood data considering only four neigh-

bouring OGUs for a given unit. Our argument can be further supported by Hausdorf & Hen- nig’s [63] observation that the same dataset when used at different scales clustered well at the 2˚ grid scale, but not at a finer 1˚ grid scale. This illustrates one of the biggest issues in numerical biogeographical analysis, i.e. although the use of equal area grid cells is recommended as the standard practice, it is accompanied by trade-offs of data completeness at fine scales, and distortion of natural boundaries and con- glomeration of habitat heterogeneity (especially crucial if patchy habitats are important for the study lineage) at large scales of equal area grid cells. However, Hennig & Hausdorf [103] pro- poses an approach to be used under such issues of data incompleteness at an appropriate scale of equal area grid cells, which is unsuitable for our data matrix with uneven OGUs. Our study marks (a) the first ever analysis on the biogeography of land snails in south-east- ern Africa and (b) the first attempt to study the biogeography of an invertebrate group of the area using UPGMA based phenetic hierarchical clustering of OGUs. However, an incidence data matrix for the weevil genus Sciobius, at 2˚ grid squares for the same study area has been used by Morrone [59] to introduce and substantiate the application of PAE. Several more stud- ies have used the same dataset to conduct various numerical biogeographic analyses and com- pare their results, viz. Szumik et al. [104] introducing an optimality criterion for AOE

identification; Mast & Nyffeler [105] proposing the use of a null model to recognise significant co-occurrence of taxa; Hausdorf & Hennig [63] introducing BEA also using a null model to test for clustering; Hennig & Hausdorf [103] to account for incomplete sampling in BEA and Escalante [106] comparing PAE with Analysis of Endemicity. Therefore, we not only analysed our dataset using the PAE and BEA in addition to the UPGMA clustering of OGUs, but also checked for cross taxon congruence of our land snail AOEs with those for Sciobius weevils. Our biotic elements support all five COEs of land snails in south-eastern Africa including those six AOEs derived from PAE for the same data set. Furthermore, our land snail COEs and AOEs correspond well with the AOEs and biotic elements of other studies [59, 63, 104– 106] derived for a 2˚ equal area OGU based dataset of Sciobius weevils in south-eastern Africa as summarised in the Table 4. Additionally, two of our biotic elements (elements 7 and 2) sup- port the Knysna extension of the proposed greater MPA region, while two more (elements 9 and 3) supports the argument of the Knysna transition zone between the greater MPA prov- ince proposed here and the greater CFR delimited by Born et al. [70]. PLOS ONE Land snail endemism in the Maputaland-Pondoland-Albany biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0248040 March 4, 2021 19 / 26