Urban versus forest ecotypes are not explained by divergent reproductive selection

Increasing urbanization offers a unique opportunity to study adaptive responses to rapid environmental change. Numerous studies have demonstrated phenotypic divergence between urban and rural organisms. However, comparing the direction and magnitude of natural selection between these environments has rarely been attempted. Using seven years of monitoring of great tits (Parus major) breeding in nest-boxes across the city of Montpellier and in a nearby oak forest, we find phenotypic divergence in four morphological and two life-history traits between urban and forest birds. We then measure reproductive selection on these traits, and compare selection between the habitats. Urban birds had significantly smaller morphological features than their rural counterparts, with a shorter tarsus, lower body mass, and smaller wing and tail lengths relative to their overall body size. While urban female tarsus length was under stabilizing selection, and forest males show positive selection for tarsus length and negative selection for body mass, selection gradients were significantly divergent between habitats only for body mass. Urban great tits also had earlier laying dates and smaller clutches. Surprisingly, we found selection for earlier laying date in the forest but not in the city. Conversely, we detected no linear selection on clutch size in the forest, but positive selection on clutch size in the urban habitat. Overall, these results do not support the hypothesis that contemporary reproductive selection explains differences in morphology and life history between urban- and forest-breeding great tits. We discuss how further experimental approaches will help confirm whether the observed divergence is maladaptive while identifying the environmental drivers behind it.

Appendix S1: Details on the field sites and monitoring protocol Monitoring of great tit populations took place in the city of Montpellier, France (43°36'N, 3°52'E) and in the forest of la Rouvière near Montarnaud, France (43°40'N, 3°40'E). Montpellier is the seventh most populated city in France with a total of 277 639 inhabitants in 2015 and a density of 4 881 inhabitants per square kilometre. Its urban area spreads over 57 square kilometres with 11 square kilometres of green spaces. In autumn 2010, 283 nest-boxes were placed across the city, but at the beginning of the breading season in 2011, 74 of them had already been stolen or vandalised. Since then, depending on our replacement effort and number of thefts, the number of nest-boxes fluctuated between 203 and 223 among years. Nest-boxes with 32 mm entrance holes were nailed at 2m height or higher, on a variety of trees such as: holm and downy oaks (Quercus ilex and Q. pubescens), platanus (Platanus × acerifolia), European nettle tree (Celtis australis), black locust (Robinia pseudoacacia), pines (Pinus halepensis, P. pinaster and P. pinea) and olive trees (Olea europea).
The forest of La Rouvière is a historical study site where blue tit and great tit populations have been monitored since 1991, with between 51 and 92 great tit nest-boxes placed in cages hanged on top of posts to avoid nest predation. On both sites, great tit nest-boxes were spaced by 100m to avoid intraspecific competition.
Great tits are known to explore an area of 50m radius around their nest-box (Naef-Daenzer 2000). Hence, proportions of green cover were measured for each nest-box in a circle of 100m diameter around the nest-box, giving a measure of habitat artificialisation level. Measures of green cover (versus impervious surfaces) around each nest-box were obtained by analysing satellite pictures with QGIS (2018). In the forest, Great tit territories were on average covered at 98.9±2.3% (mean±SD) by vegetation whereas this proportion droped to 61.9±26.9% in the city, the remaining 38.1% being artificial surfaces (roads, sidewalks, buildings).
Each week, all nest-boxes were visited to follow nest and brood development, providing information on laying date, clutch size, incubation period, number of hatchlings, age of chicks and number of fledglings. Parents were captured inside nest-boxes with mechanical traps when chicks were at least 9 days old, measured for morphological and behavioural traits and ringed with unique metal rings provided by the CRBPO (Museum National d'Histoire Naturelle, Paris). In particular, we measured body mass to the nearest 0.1g with a Pesola spring scale, tarsus length to the nearest 0.01mm with a digital calliper, and wing length and tail length to the nearest 0.5mm with rulers. All these morphological measures, apart from tail length which is rarely measured, have been shown highly repeatable and heritable (Gosler & Harper 2000& Postma 2014).

Figure S1
: Satellite images of the study sites. (a) One section of the urban city site of Montpellier, France. Each dot represents a nest-box. (b) The forest site, in the forest of La Rouvière, Montarnaud, France. Each number represents a nest-box. Note that for the forest site, numerous nest-boxes are designed for blue tits (entrance whole of 28 mm). © Google Earth. Figure S2: Variation in great tit body mass across years and between habitats (green = forest, purple = urban). Boxplots of predicted data from the best linear mixed model, representing minimum, 1 st quartile, median, 3 rd quartile and maximum. Figure S3: Variation in great tit wing length across years and between habitats (green = forest, purple = urban). Boxplots of predicted data from the best linear mixed model, representing minimum, 1 st quartile, median, 3 rd quartile and maximum. Figure S4: Variation in great tit tail length across years and between habitats (green = forest, purple = urban). Boxplots of predicted data from the best linear mixed model, representing minimum, 1 st quartile, median, 3 rd quartile and maximum.