Although the widespread effects of global change impact almost all ecosystems, we lack a detailed understanding of how wildlife that thrive in human-dominated environments are able to adjust their life history to modifications in land use of their natural habitat. In particular, spatial variation in environmental conditions is predicted to influence development during the crucial early life phase, with marked impacts on individual performance and population dynamics for long-lived species. Large herbivores such as roe deer (Capreolus capreolus), a synanthropic species, have increased substantially in number and distribution over the last half century across Europe. Roe deer have been particularly successful, gradually colonizing agricultural landscapes to cope with a global warming-driven phenological mismatch in their natural forest habitat. However, to date, little is known about how habitat heterogeneity impacts their demographic performance in this heavily human-impacted environment. Specifically, we predicted that fawns born in predominantly cultivated local habitats would achieve faster early development due to the food subsidies obtained by their mothers from agricultural crops. Contrary to our expectations, fawns in semi-natural forest habitats were around 10% heavier at birth than those born in more mixed (by 0.163 ± 0.058 kg) and open (by 0.169 ± 0.006 kg) agricultural habitats. However, all fawns subsequently grew at a similar average rate (0.148 ± 0.058 kg/day), irrespective of their habitat. This habitat-dependent variation in birth mass appeared to be driven by reproductive phenology, as i) early-born fawns were heavier than late-born fawns, and ii) mothers living in the forest gave birth around 10 days earlier than those living in the mixed and open habitats. As natural habitats become increasingly scarce and fragmented due to the activities of humans, the prospects for many wild populations will depend on their ability to subsist in the heavily modified habitats of anthropogenic landscapes.

Females GPS Data acquisition

Each winter, from 2007 to 2022, we caught roe deer (at least 8 months old) between January and March using drive netting on seven capture sites, two located in forest habitat, two in mixed habitat, and three in open habitat. At capture, the deer are immediately tranquilized and transferred to a wooden holding box until the end of the drive. Subsequently, each animal is equipped with ear tags and a GPS collar (Lotek 3300S revision 2 or GPS PLUS-1C Store on Board, Vectronic Aerospace) (see data CEFS 2021). Based on tooth wear, we distinguished juveniles (8-10 months old) from adults (>20 months old) and then released the animal on site. Each deer was monitored by GPS from capture until the end of the calendar year (11 months). Density in the mixed and open habitat of this population appears quite consistent over time and likely much lower than in forest habitat (Hewison et al. 2007) at between 5.07 (± 3.32) to 6.99 (± 4.24) individuals per 100 ha.

Fawn capture and monitoring

From mid-April to mid-June, 2 or 3 pairs of 2 observers searched for fawns in the three habitat types. Fawns were often located by opportunistic observation of the mother during semi-systematic surveys of the whole study site, but we also conducted regular targeted observations of all adult females that had been previously equipped with a GPS or VHF collar during winter captures. As roe deer fawns adopt a hider strategy to avoid predation, their mobility is low during their first days of life, and thus, they can be hand-caught at this stage. Each caught fawn was ear-tagged, sexed, and age was estimated based on appearance and behaviour according to Jullien et al. (1992). Body mass was measured using a spring balance (Testut) which was calibrated at the beginning and during the season. Fawns were assigned to a type of habitat (open, mixed, forest) based on their capture site. Finally, they were equipped with a very high frequency (VHF) tag fixed to an expandable collar (Biotrack-Lotek UK). Their location and status were recorded every weekday from capture until death, or until the collar failed. Thus, in the following analysis, we considered age-dependent body mass at first capture as our proxy for early body development for the whole fawn sample. In total we obtained data on 359 fawns captured between 2007 and 2022 for which we knew if they were dead or alive at the start of September, when weaning is almost complete.

Birth date estimation

For fawns caught during their first few days of life (1-20 days old), birth date was calculated by simply subtracting their estimated age from the date of capture. In addition, to increase sample size for the analysis of birth phenology only, we estimated parturition date of GPS monitored adult females for which we were unable to capture their fawn. To do so, we used a modified version of the method described by Marchand et al. (2021) to estimate parturition date for 33 females: 13 in forest habitat, 9 in mixed habitat, and 11 in open habitat.