Wayfinding for deer: new approach predicts migration corridors

We did this using cost-distance modeling, which is commonly used for mapping landscape connectivity. However, we avoided the standard approach to movement, which uses educated guesswork to turn habitat suitability scores into a map of movement costs. Instead, we used maximum likelihood and optimization to directly estimate how much each environmental factor increased or decreased the costs of moving through landscapes, based on the animals’ migration tracks. We tested our methodology with three different migratory mule deer herds in Idaho and Wyoming.

Our models act just like a wayfinding app on smartphones: they identify the best route to navigate between two points. For mule deer, however, the predicted corridor that best matched spring migrations wasn’t a straight line, or shortest distance, between their summer and winter range. Instead, the deer generally preferred routes with hilly terrain, shrubby vegetation, and less human development.

We were surprised at how well we were able to predict most deer movements. This suggests that instead of moving randomly, migrating mule deer appear to follow rules that balance the costs of moving with the benefits of forage access. Our approach empowers managers to train cost-distance models with their own migration data in order to predict migration routes for data-deficient areas. It also lets them assess how different environmental factors affect movement routes.

The predictive migration models are an alternative to costly and labor-intensive tracking of uncollared animals. For example, instead of deploying more GPS tags, managers can use existing tracking data and knowledge of summer and winter range locations to model the locations of migratory routes for uncollared herds. To restore a route disrupted by a barrier such as a highway, planners can use the previously fitted cost-distance models to link seasonal ranges on either side of the highway and identify the best location for a wildlife crossing structure.

This predictive ability also lets managers and planners assess the effects of land use or climate change scenarios on corridors by testing different scenarios on migration routes. Our approach assigns probabilities of use to different areas of a corridor based on the GPS training data, helping managers communicate to stakeholders the level of importance and uncertainty associated with different predictions. In many cases, this approach can be used to build corridor predictions for non-migratory animals as well.

The code and a tutorial for this approach is available as open-access. This work was made possible through the Corridor Mapping Team, a partnership created in 2018 by the USGS and consisting of wildlife agencies, tribes, and federal researchers who use GPS movement data to map migrations and support regional big-game management and conservation.

Related Posts