Predicting Continental-Scale Bird Migration Routes from Landscape Parameters

William Winters, one of my GIST students, successfully filed his MS thesis last fall, in which he created proof-of-concept models for bird migration routes at the continental scale.  We developed the topic as a way to fill in a missing scale in the efforts to reduce mortality of birds at towers and buildings.  At the scale of the building and tower, mitigations are now available, including lights-out during migration, and changing the lighting scheme on towers.  But what locations are most likely to have high levels of birds during migration, outside those already known as migratory hotspots (e.g., Cape May)?

Put another way, what places on the landscape are especially bad for tall towers and buildings because they are likely to kill more birds?  Are there really fewer birds killed at obstructions in the western part of the Great Plains as the records in our 2012 PLoS ONE paper might suggest?

Residuals in tower height-mortality regression at communication towers (Longcore et al. 2012).

Residuals in tower height-mortality regression at communication towers (Longcore et al. 2012). Note that the towers along the Front Range of the Rockies killed fewer birds than expected.

I posed the question of whether one could model probable migratory routes using least-cost path analysis and a simple set of physical parameters for sets of known wintering and breeding grounds for Neotropical migrant birds.  William ran with it.

A least-cost corridor raster for Red-eyed Vireos migrating from the Northern Atlantic forest to South America.

A least-cost corridor raster for Red-eyed Vireos migrating from the Northern Atlantic forest to South America (Winters 2015).

After some careful thinking about resolution and projection, William settled upon topography (slope) as a metric of landscape resistance, along with wind and an additional resistance value for crossing the Gulf of Mexico. Then, using pairs of wintering and breeding grounds for Red-eyed Vireo, Kirtland’s Warbler, and Golden-cheeked Warbler, he developed models to predict northbound and southbound migration routes.

The resulting maps, which were the result of some experimentation and comparison with existing observations compiled for the species in the literature and on eBird provide proof-of-concept that one could develop maps for whole species by linking together known wintering and breeding grounds.

Fall and Spring migration routes predicted for Kirtland's Warbler from Michigan only.

Fall migration route predicted for Kirtland’s Warbler from Michigan only (Winters 2015).

The maps show that the least-cost paths for the spring and fall migrations might be different.  For example, the models for Red-eyed Vireos southbound from New England funnel through the Florida Peninsula in the Fall, and return via the Gulf Coast of Florida in the Spring.

To see the full thesis, including many more maps, visit the USC Library website at this link: Identifying areas of high risk for avian mortality by performing a least accumulated-cost analysis.