SLOSS-based inferences in a fragmented landscape depend on fragment area and species–area slope

Aims: Whether a Single Large Or Several Small (SLOSS) habitat patches contain more species is central to the debate about how habitat fragmentation threatens species diversity. However, the geographical and biogeographical variables that affect emergent SLOSS patterns remain poorly understood. Here, we quantified SLOSS-based diversity patterns of woody plant, bird and spider assemblages in a subtropical archipelago of land bridge islands. Location: Thousand Island Lake, Zhejiang Province, China. Taxon: Woody plants, birds and spiders. Methods: We analysed species accumulation curves and species–area relationships (SARs) to quantify SLOSS-based patterns of all, common and rare species for each taxon across different groupings of islands. Differences in the number of species between a single large island and sets of several small islands were measured in these analyses using a Saturation index (SI) and SLOSS index. Generalized additive models were used to assess the relationships between SI and SLOSS index values and the maximum area of islands included in the analysis, the slope of SARs (i.e. z-score) and the degree of nestedness. Results: SI values increased with the maximum area of islands for three taxa, while SLOSS index values only increased with the maximum island area for woody plants. SI values increased as z-scores of woody plants decreased, and showed a quadratic relationship for birds and spiders. SLOSS index values decreased as z-scores of three taxa increased. The degree of nestedness and SI values or SLOSS index values were weakly correlated with all, rare and common species in the three taxa. Main conclusions: SLOSS-based inferences of fragmentation effects are contingent on variation in the maximum area of patches included in analyses and the slope of SARs in fragmentated landscapes. Whether conservation efforts should prioritize a single large or several small patches depends on the geographical (e.g. maximum fragment area) and biogeographical (e.g. slope of SAR) attributes of a fragmentated landscape.