Low and annually variable migratory connectivity in a long‐distance migrant: Whinchats Saxicola rubetra may show a bet‐hedging strategy

Abstract

The spatial scale of non‐breeding areas used by long‐distance migrant animals can vary from specific, relatively small non‐breeding areas for each independent breeding population (high connectivity) to a distribution over a large non‐breeding area with mixing of breeding populations (low connectivity). Measuring variation in the degree of connectivity and how it arises is crucial to predict how migratory animals can respond to global habitat and climate change because low connectivity is likely to be an adaptation to environmental uncertainty. Here, we assess whether use of non‐breeding areas in a long‐distance migrant may be stochastic by measuring the degree of connectivity, and whether it is annually variable. Twenty‐nine wintering Whinchats tagged with geolocators over 2 years within 40 km2 in central Nigeria were found to be breeding over 2.55 million km2 (26% of the land area of Europe), without an asymptote being approached in the relationship between area and sample size. Ranges differed in size between years by 1.51 million km2 and only 15% of the total breeding range across both years overlapped (8% overlap between years when only first‐year birds were considered), well above the range size difference and below the proportion of overlap that would be predicted from two equivalent groups breeding at random locations within the observed range. Mean distance between breeding locations (i.e. migratory spread) differed significantly between years (604 ± 18 km in 2013 and 869 ± 33 km in 2014). The results showed very low and variable connectivity that was reasonably robust to the errors and assumptions inherent in the use of geolocators, but with the caveat of having only ranges of 2 years to compare, and the sensitivity of range to the breeding locations of a small number of individuals. However, if representative, the results suggest the scope for between‐year variation (cohort effects) to determine migrant distribution on a large scale. Furthermore, for species with similarly low connectivity, we would predict breeding population trends to reflect average conditions across large non‐breeding areas: thus, as large areas of Africa become subject to habitat loss, migrant populations throughout Europe will decline.

Weak migratory connectivity, loop migration and multiple non‐breeding site use in British breeding Whinchats Saxicola rubetra

“…Following Blackburn et al (2017), we also calculated a minimum convex polygon (MCP) encompassing all non‐breeding locations, using the ‘areaPolygon’ function from the geosphere R library to account for curvature of the Earth (Hijams 2017). This MCP of non‐breeding locations is a further measure of population spread and enables comparison with the breeding range MCP estimated for Nigerian‐tracked Whinchats from Blackburn et al (2017).…”

“…Following Blackburn et al (2017), we also calculated a minimum convex polygon (MCP) encompassing all non‐breeding locations, using the ‘areaPolygon’ function from the geosphere R library to account for curvature of the Earth (Hijams 2017). This MCP of non‐breeding locations is a further measure of population spread and enables comparison with the breeding range MCP estimated for Nigerian‐tracked Whinchats from Blackburn et al (2017). Because the size of an MCP is sensitive to sample size, with MCP area increasing with sample size until an asymptote is reached when individuals from nearly all parts of the range are part of the sample, we additionally examined the degree to which non‐breeding range size approached an asymptote.…”

“…Because the size of an MCP is sensitive to sample size, with MCP area increasing with sample size until an asymptote is reached when individuals from nearly all parts of the range are part of the sample, we additionally examined the degree to which non‐breeding range size approached an asymptote. As per Blackburn et al (2017), we tested this by randomly sub‐sampling the non‐breeding locations 1000 times for each of 5–19 individual location points, and then calculated the area for each sub‐sample. The relationship between the mean area of the 1000 sub‐sampled ranges and number of individuals was then tested with a generalized linear model (GLM) to determine the line of best fit, comparing linear and quadratic fits to determine whether an asymptote in range with sample size was reached.…”

“…We predict a spring and autumn migration route through Iberia based on British ringing data, therefore differing from the spring route over the Mediterranean taken by Whinchats breeding in the Baltic region of Europe found by a previous study that used geolocators deployed in Nigeria in the non‐breeding season (Blackburn et al 2019). Also, based on results from tracking work in Nigeria (Blackburn et al 2017), we predict that migratory connectivity will be weak and migratory spread high, in the order of 1000 km, with the non‐breeding ranges of different British breeding populations overlapping.…”

“…2019). A tracking study conducted in Nigeria reported extremely high migratory spread, with individuals from a single non‐breeding site breeding as far as 2000 km apart (Blackburn et al 2017). By extension, individuals from a single breeding site are expected to spread out over equivalently distant non‐breeding sites.…”