Quantitative ecologist uses virtual animals to explore how changes in the environment influence their real-world counterparts

  • December 23, 2013
  • Science Highlights

Pat Zollner and his students have golden eagles, butterflies, bats, herons, flying squirrels, chipmunks, martens and raccoons wandering around at Purdue but you needn’t worry about the University being overrun.

The insects, birds and mammals are safely contained and roaming a landscape in silicon. Zollner’s research group uses computer modeling to study, for example, how a bit more foliage along the path can make it easier for egg-laying Karner blue butterflies to live harmoniously with hikers on recreational trails at the Indiana Dunes National Lakeshore, or how traffic volume on roads near their roosts can influence endangered Indiana bats foraging to feed their young.

Simulations with the level of complexity Zollner is able to incorporate, enabled by ITaP’s Envision Center and Purdue’s community clusters, can provide useful information for policymakers and planners. For instance, it helped in designing new multi-use trails at Fort Harrison State Park in Indianapolis that mitigate the effects on the bird population there, a project on which Zollner’s group collaborated with the Indiana Department of Natural Resources.

“Once we build these virtual animals it's not just an academic exercise, then we start to look at how do we use these to address real-world issues,” says Zollner, an associate professor in Purdue's Department of Forestry and Natural Resources and a quantitative ecologist.

Issues like how to best manage forests in Alaska so flying squirrels thrive; or where to put viewing stands at the Indian Ridge Marsh in Illinois to avoid crowding black-crowned night heron nestlings; or optimal habitat for successfully reintroducing martens in northern Wisconsin; or techniques for minimizing bird collisions with aircraft taking off and landing at airports.

“If we can get a better idea of how raccoons disperse, or how they colonize areas, that may have real implications for disease control programs in the future, like deploying vaccines to slow or even stop the spread of rabies,” Zollner says.

Zollner's group uses two main models in its research, SODA (for Simulation of Disturbance Activities), which simulates how animals respond to things going on in and around their normal home ranges, and SEARCH (Spatially Explicit Animal Response to Composition of Habitat), which simulates animal dispersal, or movement, over an area in response to various factors. The models are agent based, meaning they simulate the actions of individual animals in a scenario.

The computerized agents act according to an array of rules. They might, say, shy away from an area because they perceive a greater risk of being eaten by a predator or be more likely to move because it is summer, food has been abundant and they’ve been feeding well and have energy stores to fuel their travels. The rules governing the behavior of the virtual animals are built on observations of real animals in field studies by Zollner’s group and others. That involves such things as monitoring bats tagged with radio transmitters or backtracking martens by snowshoe, portable GPS unit in hand.

“We don't just make it up,” Zollner says. “There's a lot of sweat and field work that go into creating these models.”

Likewise, the maps built into the simulations are GIS-based, accurate renditions of real-world terrain, which the researchers can render at a variety of resolutions and modify dynamically, removing a stand of trees here, adding a road there.

“Not only do we need a realistic animal, but we have to have a realistic map that captures the features the animals are using,“ Zollner says.

Simulating numerous virtual animals in motion and interacting with the world around them, each possessing a virtual personality, so to speak, is computationally demanding. Add to that the complexity required in the maps of the terrain over which the agents are moving. Take the marten habitat study, which found that detail down to the level of small stands of hemlock and cedar are important to martens. The simulations also must account virtually for significant time periods of interaction between animal and environment.

Desktop computers the researchers employ limit the complexity of the simulations they can run. But Envision Center staff helped them migrate their work to the Windows HPC cluster operated by ITaP Research Computing (RCAC), a supercomputer in Purdue's Community Cluster Program that runs the high-performance-computing version of Microsoft Windows.

“Now we have the capability to ask about more variations on what trail designs should be like, what scenarios are for different volumes of vehicular traffic, and things like that,” Zollner says. “Beyond that, it lets us run a bigger map, it lets us include more habitat features in our maps, and follow more animals, give animals more complicated rules for how they respond. It provides feasibility in numerous ways for questions we just couldn't ask otherwise.”

More Information

Originally posted: July 1, 2014 4:27pm EDT