Aim: The small-island effect (SIE), as an exception to the species-area relationship, has received much attention in true island systems. However, the prevalence and related patterns of the SIE have not been well evaluated in habitat island systems. Here, we aimed to identify the existence of SIE for habitat islands in fragmented landscapes and determine the key factors influencing species richness on small habitat islands.

Location: Inner Mongolia Autonomous Region, China.

Taxon: Vascular plants.

Methods: Based on 78 grassland fragments in fragmented landscapes of the agro-pastoral ecotone of northern China, we used piecewise regression, path analysis, and null models to investigate the SIE of the species-area relationship. We then used a multi-model selection to evaluate the impacts of four influencing factors (instability, isolation, habitat heterogeneity, and surrounding productivity) on species richness (including habitat specialists and generalists) on small habitat islands within the range of SIE.

Results: We found an obvious threshold of 5.1 ha in the species-area relationship, below which habitat island area had no direct and indirect effects on species richness. Small habitat islands (< 5.1 ha) host a lower percentage of habitat specialists and a higher percentage of generalists. On small habitat islands, species richness was positively affected by habitat heterogeneity while negatively affected by instability and isolation. Habitat heterogeneity had the strongest effect on species richness, positively affecting specialist richness while negatively affecting generalist richness.

Main conclusions: There is a SIE in fragmented landscapes of the agro-pastoral ecotone of northern China, which should be considered in biodiversity conservation. Habitat heterogeneity had a key role in determining the pattern of species richness, especially for small islands. Habitat specialists and generalists had different SIE-related patterns. Our study highlights the importance of considering different ecological groups of species to improve our understanding of the SIE in fragmented habitats.