Comparing migratory connectivity across species: The importance of considering the pattern of sampling and the processes that lead to connectivity

Abstract

Measuring the degree of migratory connectivity – how much and where different populations of species mix as they migrate over their annual cycle – is important because it informs the understanding of the evolution of migration, how populations will be affected by both habitat and climate change, and which areas to prioritize for conservation. But existing measures of connectivity may be difficult to compare because they measure different things and are confounded by sampling bias. Here we use tagging data from all available published landbird tracks up to July 2019 (224 populations, 86 species and 1524 individuals tracked in the three main global flyways) to identify robust measures to compare migratory connectivity across species. We consider two widely used descriptive measures: (1) degree of breeding population overlap on the non‐breeding grounds and (2) Mantel correlation, which tests the degree of spatial autocorrelation between the breeding and non‐breeding individuals; as well as one causative measure of the main process that leads to connectivity patterns: migratory spread of individuals from the same breeding population across the non‐breeding area. We investigated the sensitivity of these three measures to the distance between breeding locations of sampled populations (breeding distance) and their sample size. We also considered the confounding effects of migration distance because longer migrations decreased overlap and increased Mantel correlations and migratory spread. We found that the degree of overlap between breeding populations on the non‐breeding grounds decreased with increasing breeding distance and increased with increasing sample size. Mantel correlation coefficients also increased significantly with increasing breeding distance; sample size did not affect accuracy, but precision was greatly improved above a sample size of about 15 individuals. Migratory spread, however, was independent of breeding distance; sample size had only small effects on accuracy and precision, with no significant effects when more than four individuals per population were included. Furthermore, migratory spread was highly positively correlated with the maximum non‐breeding range. Overlap and Mantel correlations were highly confounded by the spatial pattern and amount of sampling, whereas migratory spread was relatively unconfounded, even by migratory distance. Although any descriptive migratory connectivity measure can help set priorities by determining current areas for conservation on the non‐breeding grounds, migratory spread, which leads to these patterns, needs fewer data, is comparable, and gives information on evolutionary flexibility and so ability to deal with changing habitat and climate.