Identifying island safe havens to prevent the extinction of the World's largest lizard from global warming.

Abstract

The Komodo dragon (Varanus komodoensis) is an endangered, island‐endemic species with a naturally restricted distribution. Despite this, no previous studies have attempted to predict the effects of climate change on this iconic species. We used extensive Komodo dragon monitoring data, climate, and sea‐level change projections to build spatially explicit demographic models for the Komodo dragon. These models project the species’ future range and abundance under multiple climate change scenarios. We ran over one million model simulations with varying model parameters, enabling us to incorporate uncertainty introduced from three main sources: (a) structure of global climate models, (b) choice of greenhouse gas emission trajectories, and (c) estimates of Komodo dragon demographic parameters. Our models predict a reduction in range‐wide Komodo dragon habitat of 8%–87% by 2050, leading to a decrease in habitat patch occupancy of 25%–97% and declines of 27%–99% in abundance across the species' range. We show that the risk of extirpation on the two largest protected islands in Komodo National Park (Rinca and Komodo) was lower than other island populations, providing important safe havens for Komodo dragons under global warming. Given the severity and rate of the predicted changes to Komodo dragon habitat patch occupancy (a proxy for area of occupancy) and abundance, urgent conservation actions are required to avoid risk of extinction. These should, as a priority, be focused on managing habitat on the islands of Komodo and Rinca, reflecting these islands’ status as important refuges for the species in a warming world. Variability in our model projections highlights the importance of accounting for uncertainties in demographic and environmental parameters, structural assumptions of global climate models, and greenhouse gas emission scenarios when simulating species metapopulation dynamics under climate change.