Spatial regression methods capture prediction uncertainty in species distribution model projections through time

Author

Alan Swanson, The University of MontanaFollow

Year of Award

2012

Document Type

Thesis

Degree Type

Master of Science (MS)

Degree Name

Forestry

Department or School/College

College of Forestry and Conservation

Committee Chair

Solomon Dobrowski

Commitee Members

Dave Affleck, John Kimball

Keywords

climate change, conservation planning, GLM, GLMM, historic data, species distribution models, transferability, uncertainty

Abstract

Species distribution models (SDMs) relate observed locations of a species to climate, and are used for projecting the fate of a species under climate change scenarios. To be useful in a decision-making context, the uncertainty associated with these projections must be known. However, the uncertainty associated with SDM projections is largely ignored, perhaps because many current methods have been shown to produce biased estimates. Failure to account for spatial autocorrelation (SAC) of residual error explains much of this bias. Generalized linear mixed models (GLMM) have the ability to account for SAC through the inclusion of a spatially structured random intercept, interpreted to account for the effect of missing predictors. This framework promises a more realistic representation of parameter and prediction uncertainty. My work assesses the ability of GLMMs and a conventional SDM approach, based on generalized linear models (GLM), to produce accurate projections and estimates of prediction uncertainty. Bayesian methods were used to fit models to historical (1928-1940) observations for 99 woody plant species in California, USA, and assessed using modern "temporally independent" validation data (2000-2005). A set of climatic water balance metrics were calculated to inform the models. GLMMs provided a closer fit to historic data, had fewer significant covariates, were better able to nearly eliminate spatial autocorrelation of residual error, and had larger credible intervals for projections than GLMs. The accuracy of projections was similar between methods but the GLMMs better quantified projection uncertainty. Additionally, the GLMMs produced more conservative estimates of species range size and range size change than the GLMs. I conclude that the GLMM error structure allows for a more realistic characterization of SDM uncertainty. This is critical for conservation applications that rely on robust assessments of projection uncertainty.

Recommended Citation

Swanson, Alan, "Spatial regression methods capture prediction uncertainty in species distribution model projections through time" (2012). Graduate Student Theses, Dissertations, & Professional Papers. 905.
https://scholarworks.umt.edu/etd/905