In Hawaiʽi and elsewhere, research efforts have focused on two main approaches to determine the potential impacts of climate change on individual species: estimating species vulnerabilities and projecting responses of species to expected changes. We integrated these approaches by defining vulnerability as the inability of species to exhibit any of the responses necessary for persistence under climate change (i.e., tolerate projected changes, endure in microrefugia, or migrate to new climate-compatible areas, but excluding evolutionary adaptation).

To operationalize this response-based definition of species vulnerability within a landscape-based analysis, we used current and future climate envelopes for each species to define zones across the landscape: the toleration zone; the microrefugia zone; and the migration zone. Using these response zones we calculated a diverse set of factors related to habitat area, quality, and distribution for each species, including the amount of habitat protection and fragmentation and areas projected to be lost to sea-level rise. We then calculated the probabilities of each species exhibiting these responses using a Bayesian network model and determined the overall climate change vulnerability of each species by using a vulnerability index. As a first iteration of a response-based species vulnerability assessment (VA), our landscape-based analysis effectively integrates species-distribution models into a Bayesian network-based VA that can be updated with improved models and data for more refined analyses in the future.

Our results show that the species most vulnerable to climate change also tend to be species of conservation concern due to non-climatic threats (e.g., competition and predation from invasive species, land-use change). Also, many of Hawaiʽi’s taxa that are most vulnerable to climate change share characteristics with species that in the past were found to be at risk of extinction due to non-climatic threats (e.g., archipelago endemism, single-island endemism). Of particular concern are the numerous species that have no compatible-climate areas remaining by the year 2100. Species primarily associated with dry forests have higher vulnerability scores than species from any other habitat type. When examined at taxonomic levels above species, low vulnerabilities are concentrated in families and genera of generalists (e.g., ferns or sedges) and typically associated with mid-elevation wet habitats. Our results replicate findings from other regions that link higher species vulnerability with decreasing range size.

This species VA is possibly the largest in scope ever conducted in the United States with over 1000 species considered, 319 of which are listed as endangered or threatened under the U.S. Endangered Species Act, filling a critical knowledge gap for resource managers in the region. The information in this assessment can help prioritize species for special conservation actions, guide the management of conservation areas, inform the selection of research and monitoring priorities, and support adaptive management planning and implementation.