Temporal shifts and temperature sensitivity of avian spring migratory phenology: a phylogenetic meta-analysis

Abstract

There are wide reports of advances in the timing of spring migration of birds over time and in relation to rising temperatures, though phenological responses vary substantially within and among species. An understanding of the ecological, life-history and geographic variables that predict this intra- and interspecific variation can guide our projections of how populations and species are likely to respond to future climate change. Here, we conduct phylogenetic meta-analyses addressing slope estimates of the timing of avian spring migration regressed on (i) year and (ii) temperature, representing a total of 413 species across five continents. We take into account slope estimation error and examine phylogenetic, ecological and geographic predictors of intra- and interspecific variation. We confirm earlier findings that on average birds have significantly advanced their spring migration time by 2·1 days per decade and 1·2 days °C^-1 . We find that over time and in response to warmer spring conditions, short-distance migrants have advanced spring migratory phenology by more than long-distance migrants. We also find that larger bodied species show greater advance over time compared to smaller bodied species. Our results did not reveal any evidence that interspecific variation in migration response is predictable on the basis of species' habitat or diet. We detected a substantial phylogenetic signal in migration time in response to both year and temperature, suggesting that some of the shifts in migratory phenological response to climate are predictable on the basis of phylogeny. However, we estimate high levels of species and spatial variance relative to phylogenetic variance, which is consistent with plasticity in response to climate evolving fairly rapidly and being more influenced by adaptation to current local climate than by common descent. On average, avian spring migration times have advanced over time and as spring has become warmer. While we are able to identify predictors that explain some of the true among-species variation in response, substantial intra- and interspecific variation in migratory response remains to be explained.

Keywords: arrival date; bird migration timing; climate change; migratory phenology; plasticity.

Figure 1

Funnel plots of slope estimates plotted against the inverse of standard errors (SE) obtained (a) over time (days year⁻¹) and (b) with respect to temperature (day °C⁻¹). Vertical lines represent the average effect size (solid) for the slope of spring migration timing and the associated lower and upper 95% CIs (dashed), as estimated using a mixed model meta‐analysis that included phylogeny, species, study and location as random effects and the grand mean as the sole fixed effect.

Figure 2

Posterior median advances (and associated 95% CIs) in spring migration timing (a) over time (days year⁻¹) and (b) with respect to temperature (day °C⁻¹) for different metrics of monitoring migration timing, as estimated under the basic model. Estimates are for arrival to the breeding ground as reported by standardized field studies, with year response estimates representing advances in the decade 1980. Note that although CIs of the FAD and MAD slope estimates overlap each other, the 95% CI for the difference in slope between FADs and MADs as estimated directly from the basic models (plotted to the right of the grey vertical line) does not overlap zero and is significant.

Figure 3

Predicted effects of latitude on changes in mean/median arrival dates over time in the Northern and Southern Hemispheres. Grey circles represent temporal slope estimates. Black lines represent the latitudinal predictions in the Northern and Southern Hemispheres, as estimated under the ecological model with latitude as a predictor. Estimates are for short‐distance migrants; migrants that do not rely on forest habitats during breeding and passage; migrants with a predominantly invertebrate diet; habitat and diet specialists; body size of 10 g; arrival data as reported by standardized field studies; and the decade 1980.

Figure 4

Posterior median advances (and associated 95% CIs) in mean/median arrival dates for different migration distance classes (a) over time (day year⁻¹) and (b) with temperature (days °C⁻¹) at the mean latitude of the data set in the Northern Hemisphere (46·1°N), as estimated under our ecological models. Unclassified migrants refer to migrants that were not assigned migration distances in the original studies. Estimates are for migrants that do not rely on forest habitats during breeding and passage; migrants with a predominantly invertebrate diet; habitat and diet specialists; body size of 10 g; arrival data as reported by standardized field studies; and the decade 1980 for year slope estimates. The difference in slope between short‐ and long‐distance migrants as estimated directly from our ecological models (plotted to the right of the grey vertical line) does not overlap zero and is significant.