ABSTRACT
Shifts in the timing of animal migration are widespread and well-documented; however, the mechanism underlying these changes is largely unknown. In this study, we test the hypothesis that systematic changes in stopover duration-the time that individuals spend resting and refueling at a site-are driving shifts in songbird migration timing. Specifically, we predicted that increases in stopover duration at our study site could generate increases in passage duration-the number of days that a study site is occupied by a particular species-by changing the temporal breadth of observations and vise versa. We analyzed an uninterrupted 46-year bird banding dataset from Massachusetts, USA using quantile regression, which allowed us to detect changes in early-and late-arriving birds, as well as changes in passage duration. We found that median spring migration had advanced by 1.04 days per decade; that these advances had strengthened over the last 13 years; and that early-and late-arriving birds were advancing in parallel, leading to negligible changes in the duration of spring passage at our site (+0.07 days per decade). In contrast, changes in fall migration were less consistent. Across species, we found that median fall migration had delayed by 0.80 days per decade, and that changes were stronger in late-arriving birds, leading to an average increase in passage duration of 0.45 days per decade. Trends in stopover duration, however, were weak and negative and, as a result, could not explain any changes in passage duration. We discuss, and provide some evidence, that changes in population age-structure, cryptic geographic variation, or shifts in resource availability are consistent with increases in fall passage duration. Moreover, we demonstrate the importance of evaluating changes across the entire phenological distribution, rather than just the mean, and stress this as an important consideration for future studies.
ABSTRACT
We analyzed data from across five decades of passerine bird banding at Manomet in Plymouth, Massachusetts, USA. This included 172,609 captures during spring migration and 253,265 during fall migration, from 1969 to 2015. Migration counts are prone to large interannual variation and trends are often difficult to interpret, but have the advantage of sampling many breeding populations in a single locale. We employed a Bayesian state-space modeling approach to estimate patterns in abundance over time while accounting for observation error, and a hierarchical clustering method to identify species groups with similar trends over time. Although continent-wide there has been an overall decrease in landbird populations over the past 40 years, we found a variety of patterns in abundance over time. Consistent with other studies, we found an overall decline in numbers of birds in the aggregate, with most species showing significant net declines in migratory cohort size in spring, fall, or both (49/73 species evaluated). Other species, however, exhibited different patterns, including abundance increases (10 species). Even among increasing and declining species, the specific trends varied in shape over time, forming seven distinct clusters in fall and ten in spring. The remaining species followed largely independent and irregular pathways. Overall, life-history traits (dependence on open habitat, nesting on or near the ground, migratory strategy, human commensal, spruce budworm specialists) did a poor job of predicting species groupings of abundance patterns in both spring and fall, but median date of passage was a good predictor of abundance trends during spring (but not fall) migration. This suggests that some species with very similar patterns of abundance were unlikely to be responding to the same environmental forces. Changes in abundance at this banding station were generally consistent with BBS trend data for the same geographic region.
