Biogeographic and anthropogenic correlates of Aleutian Islands plant diversity: A machine‐learning approach

This is the first comprehensive analysis of vascular plant diversity patterns in the Aleutian Islands to identify and quantify the impact of Aleutian Island distance dispersal barriers, geographical, ecological and anthropogenic factors. Data from public Open Access databases, printed floristic accounts, and from collections made by the primary author were used to develop an Aleutian floristic database. The most common plant distribution pattern was “an eastern origin community”, though it compared similarly to the “Western” and “Widespread” distribution pattern. We established an ecological plant community composition class for each island, based on clustering species assemblage dissimilarity measurements (Jaccard Index), and a measurement of phylogenetic dissimilarity (UniFrac). We modelled these composition classes and species richness values in non‐parametric algorithmic models and concepts (data cloning using machine learning, stochastic boosting‐ TreeNet) based on classic and Aleutians‐specific island biogeography hypotheses. Plant species richness is strongly associated with the equilibrium model variables of area and island isolation, as well as distance to the Alaska Peninsula, and island total stream length. Species composition is strongly associated with the landmass groups during the last glacial maximum, maximum island elevation, island isolation and island area. Phylogenetic composition is associated with island area, distance from the islands to the Chukotka Peninsula, maximum island elevation, island geologic age, and island isolation. This study extends the equilibrium theory of island biogeography by including additional drivers of diversity during the Anthropocene, such as the landmass during the LGM, as well as factors that may be related to anthropogenic extinction rate.