Continental Patterns of Phenotypic Variation Along Replicated Urban Gradients: A Mega-Analysis

Abstract

Individual variation among and within natural populations can have eco-evolutionary implications by, for example, affecting species interactions or evolutionary potential. Urban systems present a unique opportunity to evaluate how environmental change shapes variation since urban phenotypic differentiation is widely documented on contemporary timescales. We introduce and test three hypotheses to determine how urbanisation affects phenotypic variation at different population levels. Combining 21 long-term datasets in a mega-analysis approach, we synthesise how urbanisation impacts variation in tarsus length and lay date among and within subpopulations of great and blue tits (Parus major, Cyanistes caeruleus ) at a continental scale. Our synthesis reveals that urbanisation is associated with increased phenotypic variation within subpopulations by 11% on average, and by as much as 25% across the species and traits examined. We also find some evidence (for tarsus length in great tits) that urbanisation increases differentiation between subpopulations. We did not, however, find that urbanisation increases differences between subpopulations in their within-subpopulation variation. Our synthesis provides novel insights into how urban contexts impact individual diversity at different spatial scales and we highlight future directions that could establish the genetic and environmental effects that underlie these continental patterns of urban phenotypic variation.

Keywords: blue tit; city; environmental heterogeneity; great tit; individual diversity; intraspecific variation; lay date; subpopulation; tarsus length; urbanisation.

FIGURE 1

Visual representation of three hypotheses concerning patterns of phenotypic variation among birds in five subpopulations (or clusters as defined in methods) along a theoretical urban gradient where urbanisation increases from left to right. Subpopulations contain groups of nonurban (green) and urban (blue) individuals that differ in body size where individuals are smaller on average in urban subpopulations. Higher phenotypic variation in cities could be driven by (1) among‐subpopulation heterogeneity: Higher differentiation between urban subpopulations since there is higher variation among urban groups in their mean traits than nonurban ones (i.e., some urban subpopulations contain large individuals and others contain medium or small individuals, while nonurban subpopulations all contain large individuals), (2) within‐subpopulation heterogeneity: Higher trait differences within urban subpopulations (i.e., small, medium and large individuals) compared to nonurban subpopulations (i.e., large individuals), (3) heterogeneity in heterogeneity: Differences between urban subpopulations drive higher variation among urban groups in the trait variation they contain (i.e., urban subpopulations contain different compositions of individual sizes).

FIGURE 2

Summary of European urban gradient datasets showing (A) list of study systems (N = 14 study systems; listed by decreasing latitude, range = 61°31′ to 41°23′ N) and their year range and number of subpopulations (=clusters adefined by clustering algorithm) in urban and forest habitats, (B) map of Europe showing the location of each study system in A and whether the dataset included great and blue tits (circles) or great tits only (squares) and (C) the three traits examined and the number of study systems, individuals and subpopulations (=urban vs. forest clusters as defined in methods) of the full combined dataset. See also Tables S1 and S2 for further information.

FIGURE 3

Results of each hypothesis and associated predictions (columns) for (A) adult tarsus length, (B) nestling tarsus length and (C) lay date traits (rows). (1) Among‐subpopulation heterogeneity hypothesis (parameter 1) shows the variance estimated among forest and urban clusters on the mean of the traits, (2) Within‐subpopulation heterogeneity hypothesis (parameter 2) shows the urban effect on the residual variance of the traits (forest versus urban; positive values indicate more residual variation in urban habitats) and (3) Heterogeneity in heterogeneity hypothesis (parameter 3) shows the variance estimated among forest and urban clusters on the residual variance of the traits. Model estimates related to each parameter are from Table 1 and their 95% (thin line) and 50% (thicker line) highest posterior density intervals (or CI = credible interval) are shown. Parameters 1 and 3 are fitted as random effects in the model and return two variance estimates (urban = blue, forest = green), whereas parameter 2 is fitted as a fixed effect and returns a slope coefficient (effect of urbanisation = grey; see also Table 1). See also Figure S8 for results of 1 and 3 presented as the difference between urban and forest variance posterior draws. Estimates for great tits are represented by squares and blue tits by triangles.