1Aims: The General Dynamic Model of oceanic island biogeography (GDM) predicts how 2 2 biogeographical rates, species richness, and endemism vary depending on island age, area, and 2 3 isolation, based on the interplay of colonization, extinction, and speciation. Here, we used a 2 4 simulation model to test whether GDM predictions may arise from individual-and population-2 5 level processes.
6Location: Hypothetical hotspot islands. 2 7 Methods: Our model (i) considers an idealized island ontogeny, (ii) metabolic constraints, and 2 8 (iii) stochastic, spatially-explicit, and niche-based processes at the level of individuals and 2 9 populations (plant demography, dispersal, competition, mutation, and speciation). Isolation 3 0 scenarios involved varying dispersal ability and distances to mainland. 3 1 Results: Humped temporal trends were obtained for species richness, endemic richness, 3 2 proportion of cladogenetic endemic species, number of radiating lineages, number of species per 3 3 radiating lineage, and biogeographical rates. The proportion of anagenetic endemics and of all 3 4 endemics steadily increased over time. Extinction rates of endemic species peaked later than for 3 5 non-endemic species. Species richness and the number of anagenetic endemics decreased with 3 6 isolation as did rates of colonization, anagenesis, and extinction. The proportion of all endemics 3 7 and of cladogenetic endemics, the number of cladogenetic endemics, of radiating lineages, and of 3 8 species per radiating lineage, and the cladogenesis rate all increased with isolation.3 9 Main conclusions: The results confirm most GDM predictions related to island ontogeny and 4 0 isolation, but predict an increasing proportion of endemics throughout the experiment: a difference 4 1 3 attributable to diverging assumptions on late island ontogeny. New insights regarding the 4 2 extinction trends of endemics further demonstrate how simulation models focusing on low 4 3 ecological levels provide tools to test biogeographical-scale predictions and to develop more 4 4 detailed predictions for further empirical tests.4 5