Comparing Resource Selection and Demographic Models for Predicting Animal Density

• Main Authors: Street, Garrett M., Rodgers, Arthur R., Avgar, Tal, Vennen, Lucas M. Vander, Fryxell, John M.
• Other Authors: Wiley
• Format: Text
• Language: unknown
• Published: Hosted by Utah State University Libraries 2016
• Subjects: abundance; aerial survey; Alces; carrying capacity; demography; habitat; predictive model; resource selection function; Environmental Sciences; Canada; Alces alces
• Online Access: https://digitalcommons.usu.edu/wild_facpub/2765

Habitat-based prediction of population density relies on relationships between landscape configuration (i.e., abundance of land-cover types) and equilibrium density. This may be accomplished by estimating resource selection probability functions (RSPFs) based on presence–absence data, or by relating carrying capacity to landscape covariates. We used RSPFs for moose (Alces alces) from 2 study sites and carrying capacities from 34 wildlife management units across northern Ontario, Canada, to create 2 estimators of moose density. We compared the predictions of both models to moose density in a novel site obtained via aerial census. We also projected the RSPF across 34 management units and compared predicted density to estimated carrying capacity of each unit. The RSPF and carrying capacity models predicted moose equilibrium densities that were statistically indistinguishable from the estimated density of moose at the novel study site, but the carrying capacity model generated uninformatively broad prediction intervals. The RSPF failed to predict carrying capacities in management units across Ontario; however, the differences between RSPF estimates and carrying capacities varied predictably with differences in covariates related to forage availability, suggesting habitat selection strength and RSPF transferability vary with landscape quality. Estimating densities using RSPFs relies on a consistent relationship between habitat selection and animal density; thus, RSPF applicability across space will depend heavily on similarity between the novel and original sites. Demographic projection benefits from broad spatiotemporal datasets that improve reliability but that are relatively rare and subject to broad error. Our findings suggest that selection-based population estimation is preferable to demographically based models because of increased precision of estimates, the immediacy of available data (e.g., single survey or radio-telemetry in multiple sites vs. many generations of population estimates in a time series across space), and the capacity to predict fine-scale patterns of distribution and abundance