Influences of species, latitudes and methodologies on estimates of phenological response to global warming

Abstract New analyses are presented addressing the global impacts of recent climate change on phenology of plant and animal species. A meta‐analysis spanning 203 species was conducted on published datasets from the northern hemisphere. Phenological response was examined with respect to two factors: distribution of species across latitudes and taxonomic affiliation or functional grouping of target species. Amphibians had a significantly stronger shift toward earlier breeding than all other taxonomic/functional groups, advancing more than twice as fast as trees, birds and butterflies. In turn, butterfly emergence or migratory arrival showed three times stronger advancement than the first flowering of herbs, perhaps portending increasing asynchrony in insect–plant interactions. Response was significantly stronger at higher latitudes where warming has been stronger, but latitude explained < 4% of the variation. Despite expectation, latitude was not yet an important predictor of climate change impacts on phenology. The only two previously published estimates of the magnitude of global response are quite different: 2.3 and 5.1 days decade −1 advancement. The scientific community has assumed this difference to be real and has attempted to explain it in terms of biologically relevant phenomena: specifically, differences in distribution of data across latitudes, taxa or time periods. Here, these and other possibilities are explored. All analyses indicate that the difference in estimated response is primarily due to differences between the studies in criteria for incorporating data. It is a clear and automatic consequence of the exclusion by one study of data on ‘stable’ (nonresponsive) species. Once this is accounted for, the two studies support each other, generating similar conclusions despite analyzing substantially nonoverlapping datasets. Analyses here on a new expanded dataset estimate an overall spring advancement across the northern hemisphere of 2.8 days decade −1 . This is the first quantitative analysis showing that data‐sampling methodologies significantly impact global (synthetic) estimates of magnitude of global warming response.