Testing environmental DNA sampling and predictive modeling as means to investigate wood frog (Rana sylvatica) distribution in Alaska and Northern Canada

• Main Author: Spangler, Mark Anthony
• Other Authors: Huettmann, Falk, López, J. Andrés, Barnes, Brian M.
• Format: Thesis
• Language: English
• Published: 2017
• Subjects: Wood frog; Alaska; Genetics; Canada; Northern; Geographical distribution; Models; Elliot; Fairbanks; North Star; Yukon
• Online Access: http://hdl.handle.net/11122/8142

Thesis (M.S.) University of Alaska Fairbanks, 2017 Global amphibian declines over the past 30+ years have led to a greater awareness of amphibian conservation issues. Few amphibian species occur in northern landscapes, however, and the species that do occur are widely dispersed and at the northern extent of their range. Accordingly, amphibian research is not prioritized in northern landscapes. Deficient monitoring practices have resulted in incomplete distribution knowledge that impedes the management of wood frogs (Rana sylvatica) in Alaska and northern Canada. I developed an environmental DNA detection assay to complement monitoring practices at the northern extent of the wood frog's range. This assay was tested to be species-specific, allowing it to be implemented in areas where wood frogs may co-occur with other amphibian species. It can detect wood frog DNA in environmental samples to a concentration of 1.83 x 10⁻³ pg/μL. I further demonstrate that environmental DNA occurrence data can be used to predict wood frog distribution in the Fairbanks North Star Borough. I combined environmental DNA occurrence data with environmental GIS data and analyzed the resulting dataset with machine learning algorithms to define an ecological niche for the wood frog. This niche, when extrapolated to the landscape, results in a species distribution model that attains 74% predictive accuracy. Lastly, I conducted an environmental DNA mega-transect survey along the Elliot/Dalton Highway corridor in Alaska. I combined the results of this survey with citizen science occurrence data from past and current monitoring projects to create a set of alternative occurrence data. This alternative data was combined with environmental GIS data and analyzed with machine learning algorithms to create a species distribution model that achieves 92% predictive accuracy across Alaska and the Yukon Territory, Canada. These results improve upon prior species distribution models developed for wood frogs in Alaska. They provide deeper insights into ...