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

with a recently well-resolved taxonomy. We did not source other museum databases except for the KwaZulu-Natal Museum, since it represents by far the largest collection of land snails in SA. Furthermore, the KwaZulu-Natal Museum database also includes the locality data from the monograph of South African non-marine Mollusca [98] and thus overlaps extensively with the holdings of the South African Museum for the selected genera. There is also a considerable overlap with the East London Museum collection through a collaborative field work pro- gramme spanning the last 20 years. Hence, the inclusion of other smaller collections is unlikely to change the species incidence matrix for the selected genera at the coarse scale of OGUs used for this study. Although less likely, any errors in the boundaries of proposed land snail biogeographical entities, owing to qualitatively derived vertebrate-based boundaries of OGUs, could only be corrected once the land snail incidence data are available at an appropriate level of complete- ness to facilitate an analysis at a finer equal-area grain, preferably at the QDS scale. Unfortu- nately, land snail species have not been sampled intensively at QDS scale in SA [15], and poor or spatially heterogeneous sampling can result in unreliable biogeographical regionalisation patterns [99]. Consensus across methods on patterns of endemism The primary objective of our analyses was to attempt a land snail-based bio-regionalisation of the area and thence to see if that would support the results of a similar analysis conducted on

vertebrates [22, 24]. Standard methods endorsed by Kreft & Jetz [39] for biogeographical regionalisation, which have been followed by the majority of subsequent studies throughout the world were used for the main analysis of our research, i.e. agglomerative hierarchical clus- tering with UPGMA algorithm. Using this method, we regionalised the area and further iden- tified the geographic clusters of the dendrogram with an aggregation of range overlapping endemic species as COEs. However, we further attempted establishing the AOEs using PAE described by Morrone [59], being the most widely used method for the identification of AOEs (see Morrone [100] for a detailed discussion). PAE is designed for incidence data matrices of equal area grid cells and it performs best using such OGUs. Our data matrix uses pre-defined eco-geographical units as OGUs, and hence several equal area grid cells with ecological similarity are already combined in a single OGU, possibly lowering the power of PAE to define the complete hierarchy of AOEs, as it makes it difficult to numerically establish the strict sympatry of taxa, however still being able to define AOEs based on the clades in area cladogram. Perera et al. [22] described the delimitation of our OGUs through a qualitative overlaying of range maps in identification of endemic vertebrate distributions (= biochoria [70] for endemic vertebrates), which are bio- geographical area units other than AOEs [101, 102]. Therefore, using such units here for iden- tification of AOEs of land snails is not circular and warranted. The determination of AOEs

was the prime objective of carrying out a PAE in our study, while those identified AOEs are perfectly congruent to COEs identified from the UPGMA clustering. PAE has exclusively been used in regionalisation attempts to provide hierarchical area cladograms of geographic units in some global to sub-continental scale studies on entire faunas [24, 60, 61], in which more endemics are captured, increasing the power of the analysis. However, we believe the power of PAE in doing so depends on several factors such as (a) higher numbers of endemic taxa in the matrix, (b) higher number of OGUs being occupied by more than one species, and (c) whether the study uses equal area OGUs defined by a grid of squares. Our OGUs were not equal area units nor were they square in shape, while the matrix had only 73 species with 15 out of 40 OGUs being occupied only by a single or no species (parsimony uninformative), thence 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 18 / 26