Incorporating animal behavior into landscape-level models of wildlife movements

Robin E Russell, Purdue University

Abstract

I developed a computer simulation to demonstrate the effects of common assumptions regarding animal movements on the patch occupancy rates and density distributions of populations in complex landscapes. Animals in these simulations either always took the least risky pathway (least-cost rule), moved in random directions (random rule), or choose a pathway in proportion to the level of risk represented by the pathway (probabilistic rule). Least-cost rules resulted in the highest per patch densities, but the lowest patch occupancies, while random and probabilistic rules resulted in similar per patch densities and patch occupancies, however, animals utilizing probabilistic rules were able to persist in higher risk landscapes than animals following random rules. In 2002, I designed an experimental landscape, consisting of habitat patches connected by corridors of different vegetation types designed to represent different levels of risk to voles. Movements were tracked by recording the telemetry locations of radio collared voles. In 2003, I repeated the experiment but modified the corridors to represent matrix or non-habitat landscape patches. For both years, I demonstrated two methods of assessing uncertainty in habitat transition estimates from the observed data, bootstrapping and Bayesian techniques. Bootstrapping and Bayesian techniques resulted in similar mean parameter estimations, but different distributions. Accurate distributions of parameters estimates are important for reflecting individual behavioral heterogeneity within a population. I also designed simulations that combined least-cost, random, and probabilistic movements based on imperfect information (to represent dispersal) and movements based on perfect information (to represent foraging). In general animals moving with perfect information persisted longer, had higher patch occupancy rates, and greater numbers of animals on the landscape than animals with imperfect information. Least-cost foraging resulted in the greatest numbers of animals on the landscape, but when combined with least-cost dispersal resulted in lower patch occupancies than when combined with random or probabilistic dispersal. Overall, assumptions in landscape-level models regarding the behavioral decisions of animals can lead to significant effects on model outcome.

Degree

Ph.D.

Advisors

Swihart, Purdue University.

Subject Area

Ecology

COinS