Modelling migratory waterfowl stopover habitat while accounting for ephemeral environmental conditions
By Natalia Rogova, Mikhail Iliev, Nikolai Petkov, Sonia Rozenfeld, Alexey Timoshenko, Didier Vangeluwe, Ivan A. Zuban, Anna M. Pidgeon and Volker C. Radeloff
Migratory species depend on ephemeral environmental conditions; thus, species distribution modelling (SDM) must incorporate phenological changes along migratory routes.
The overarching goal was to model habitats for three waterfowl species migrating through Eurasian grasslands (red-breasted goose [Branta ruficollis], taiga bean goose [Anser fabalis fabalis] and Bewick’s swan [Cygnus columbianus bewickii]) while accounting for ephemeral environmental conditions. The objectives were (a) to develop a workflow of mapping ephemeral environmental conditions, (b) model habitats for the three species and (c) evaluate the protection status of habitats in natural and agricultural landscapes. Water availability, particularly ephemeral spring waterbodies, was expected to strongly influence these species’ distributions.
Bewick’s swans (Cygnus columbianus bewickii) in flight. Photo by Ben Cherry/WWT.
MODIS data for phenological synchronization of Landsat images to create species- and season-specific metrics and land cover maps were utilised. Landsat-derived environmental variables, elevation and bird GPS locations in Maxent SDM were used. locations of modelled habitats, protected areas and Ramsar sites were compared.
The land cover maps had an overall accuracy of 0.92–0.95 and captured ephemeral water extent during these species’ migrations. All models had AUC scores of 0.89–0.94; distance to water, land cover and elevation were the most important variables. Modelled habitats were distributed unevenly and occurred in both natural and agricultural landscapes; 40%–76% fell within croplands. Although most croplands provide a rich food supply, their value as waterfowl habitat critically depended on water availability. Approximately 22% of potential habitat in the natural landscape, but only 3% in croplands, had some level of protection.
The potential of phenological synchronization to estimate fine-scale ephemeral environmental conditions crucial for migratory species was demonstrated. Modelled habitats revealed key stopover sites in both natural and agricultural landscapes. The maps showed gaps in the protected area network of Eurasian grasslands, especially in agriculture. The workflow could be easily adapted for other species and is broadly relevant for conservation.
The full article can be found here
