Range-wide post- and pre-breeding migratory networks of a declining neotropical-nearctic migratory bird, the blackpoll warbler

Abstract

Identifying the drivers of population declines in migratory species requires an understanding of how individuals are distributed between periods of the annual cycle. We built post- (fall) and pre-breeding (spring) migratory networks for the blackpoll warbler (Setophaga striata), a Neotropical-Nearctic songbird, using tracking data from 47 light-level geolocators deployed at 11 sites across its breeding range. During pre-breeding migration, two stopover nodes (regions) on the U.S. eastern seaboard received high scores in our network metrics (betweenness centrality and time-adjusted node weight), likely acting as key refuelling areas for most of the global blackpoll warbler population before their multi-day flights over the Atlantic Ocean. During post-breeding migration, highly ranked stopover nodes in the southeastern U.S. acted as a geographical bottleneck before birds dispersed to their boreal breeding destinations. Nodes located in northern Colombia and Venezuela were also ranked highly during both migrations and were likely used to prepare for (pre-breeding) and recover from (post-breeding) Atlantic flights. Blackpoll warblers showed a crosswise migration pattern, whereby individuals from western breeding populations tended to spend the nonbreeding season in the eastern part of the nonbreeding range and vice-versa. Despite this, the strength of migratory connectivity between the breeding and nonbreeding grounds ranged from moderate to low, largely because many individuals used more than one node during the 'stationary' nonbreeding period. Our results suggest that the number of breeding populations affected by a threat in the blackpoll warbler's range will strongly depend on where and when this threat occurs. Consequently, our migratory network should be key to inform future conservation planning and population monitoring efforts.

Keywords: Conservation; Geolocator; Migratory stopovers; Movement ecology; Setophaga striata; eBird.

Fig. 1

Structure of the post- and pre-breeding migratory networks. Panel (a,b) show the location and weight of network nodes. The blackpoll warbler breeding, migration, and nonbreeding distributions polygons, obtained from Birdlife International distribution maps, are also shown. Node size and edge size are proportional to the relative abundance contributed by blackpolls that used them. Nodes are located at the median location of their stationary locations, with the exception of nodes 5, 6, and 7, which were placed at supposed coastal departure points. These three nodes and all others in North America during post-breeding migration were clustered using longitude only. Edges connect nodes but do not represent actual migratory routes. Panels (c,d) show the colour-coded and numbered clusters of stationary locations that correspond to each stopover and nonbreeding node. Each location corresponds to the median of the posterior distribution from the geolocator analysis, and the error bars show the 95% credible interval for longitude and latitude (occasionally hidden by symbols). In the post-breeding migration network, a portion of the stationary locations in the Caribbean were placed manually based on our assessment of light-level data and lacked credible intervals (See Supplementary Methods 4). Stationary locations estimated within 2 weeks of the equinox are circled in black and their latitude should be considered highly unreliable. The maps were generated in R using package sf version 1.0-16 (https://cran.r-project.org/web/packages/sf/), ggplot2 version 3.5.1 (https://ggplot2.tidyverse.org/), and lake and country polygons obtained from the Natural Earth dataset (www.naturalearthdata.com), via the naturalearth package version 1.0.1 (https://docs.ropensci.org/rnaturalearth/).

Fig. 2

Relative node use patterns for blackpoll warblers from each of the four regions we defined in the nonbreeding grounds (Eastern, Central, Western, and Northwestern, see Fig. S1) during (a) post-breeding and (b) pre-breeding migration. Inset maps (c,d) show only edges for movements performed between nonbreeding nodes, as well as the final (after post-breeding migration) and initial (before pre-breeding migration) population-specific use and weight accounting only for individuals that stayed over 2 weeks (i.e., excluding individuals making stopovers). The communities identified by the Walktrap algorithms are shown in (e,f), with their modularity values. A modularity near 0 indicates a weak community structure, while a high value indicates more edges within rather than between communities. The maps were generated in R using package sf version 1.0-16 (https://cran.r-project.org/web/packages/sf/), ggplot2 version 3.5.1 (https://ggplot2.tidyverse.org/), and lake and country polygons obtained from the Natural Earth dataset (www.naturalearthdata.com), via the naturalearth package version 1.0.1 (https://docs.ropensci.org/rnaturalearth/).

Fig. 3

Values of betweenness centrality in the (a) post-breeding and (b) pre-breeding migratory networks for blackpoll warblers and time-adjusted node weights in the (c) post-breeding and (d) pre-breeding migratory networks. Nodes with high betweenness centrality are located at the intersection of the shortest path between other nodes in the network. Nodes with high time-adjusted weight are expected to have a higher relative abundance of blackpoll warblers and are used for longer stopovers. The maps were generated in R using package sf version 1.0-16 (https://cran.r-project.org/web/packages/sf/), ggplot2 version 3.5.1 (https://ggplot2.tidyverse.org/), and lake and country polygons obtained from the Natural Earth dataset (www.naturalearthdata.com), via the naturalearth package version 1.0.1 (https://docs.ropensci.org/rnaturalearth/).

Fig. 4

An overview of post-migratory nonbreeding movements performed between the post- and pre-breeding migrations analyzed in the migratory networks. Locations of the initial and first stationary locations used by individuals during the nonbreeding season are shown in (a). Static points show the mean stationary locations of individuals that made no detectable post-migratory nonbreeding movements. Error bars show the 95% credible interval for the longitude and latitude of each location. The direction and timing of post-migratory nonbreeding movements are shown in (b) and movements initiated within 2 weeks of the spring equinox are indicated by dashed arrows. The maps were generated in R using package sf version 1.0-16 (https://cran.r-project.org/web/packages/sf/), ggplot2 version 3.5.1 (https://ggplot2.tidyverse.org/), and country polygons obtained from the Natural Earth dataset (www.naturalearthdata.com), via the naturalearth package version 1.0.1 (https://docs.ropensci.org/rnaturalearth/).