Urban resources limit pair coordination over offspring provisioning

The amount of care parents provide to the offspring is complicated by an evolutionary conflict of interest (‘sexual conflict’) between the two parents. Recent theoretical models suggest that pair coordination of the provisioning may reduce this conflict and increase parent and offspring fitness. Despite empirical studies showing that pair coordination is common in avian species, it remains unclear how environmental and ecological conditions might promote or limit the ability of parents to coordinate care. We compared the level of pair coordination, measured as alternation and synchrony of the nest visits, of house wrens Troglodytes aedon pairs breeding in a rural (10 nests) and a suburban (9 nests) site and investigated how differences in parental behaviours were related to habitat composition, prey abundance and how they ultimately related to reproductive success. We found that parents alternated and synchronized their nest visits more in the rural site compared to the suburban one. The suburban site is characterized by a more fragmented habitat with more coniferous trees and less caterpillar availability. Offspring from the rural site were heavier at fledging than at the suburban site. Taken together, these results suggest that environmental conditions play an important role on the emergence of coordinated parental care and that considering environmental variables is pivotal to assess the fitness consequences of parental strategies.


Results
The suburban site and its rural counterpart greatly differ in terms of urbanization score and vegetation composition (Table 1; Supplementary Fig. S1). The suburban site is characterized by a higher density of buildings and paved surfaces and by a decreased vegetation density compared to the rural site. Furthermore, coniferous trees are most abundant in the suburban site, while deciduous trees are predominant at the rural site (Table 1).
Daily caterpillar biomass differed over the breeding season in interaction with site ( Fig. 1; Table 2). Post-hoc tests indicated that the difference in caterpillar abundance between sites occurred around the caterpillar peak and not at the beginning and end of the breeding season (see Supplementary Table S2). Furthermore, the significant interaction of site and date 2 indicate that the curvature of the peak of caterpillar abundance differed by site, i.e., the peak is more pronounced at the rural site (Fig. 1). In addition, the caterpillar peak occurred early in the suburban site compared to the rural counterpart (t 3.38 = 3.39, p = 0.035; rural site: 156.9 ± 0.46 (mean Julian day ± SE); suburban site: 150.56 ± 1.81). www.nature.com/scientificreports/ The proportion of prey delivered at the nest differed between the two sites. In rural nests, caterpillars were delivered at the nest more often than spiders and other prey types (flying insects and beetles), whereas in suburban nests other prey types were delivered more often than caterpillars and spiders ( Fig. 2; Table 3). However, rural males delivered more caterpillars and fewer spiders and other prey types compared to rural females (Table 3), whereas suburban males delivered fewer caterpillars and more spiders compared to suburban females ( Fig. 2; Table 3).
Parental provisioning rates differed between sites (Cohen's d = 0.91 [CI 0.22-1.60]), with parents provisioning at higher rates in rural nests ( Fig. 3a; Table 4a). There was no significant effect of sex in interaction with site (F 1,17 = 0.91, p = 0.353) or sex fitted as a single term, suggesting that the two parents had similar provisioning rates. Female parents were significantly more regular in their provisioning compared to male parents while controlling for brood size ( Fig. 3b Table 4c). Specifically, only parents in the rural site alternated their visits more than expected by chance (alternation score significantly differed from zero in the rural site: t 9 = 4.64, p = 0.001; but not in the suburban site: t 8 = 0.54, p = 0.603). In addition, the proportion of synchronized visits was higher in rural nests than in suburban nests (Cohen'  Shown are means ± SE for the rural (circles and solid line) and suburban (triangles and dashed line) sites. Asterisks represent significant differences in caterpillar biomass on a specific sampling date (*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001). Only significant comparisons are shown.

Discussion
By comparing house wren pairs breeding in a rural and suburban site, we found that pair coordination of the provisioning differed between the sites, with parents being more coordinated in the rural environment. This difference was associated with divergent environmental conditions between sites, such as habitat fragmentation, tree composition and caterpillar availability, and rural young were fed more caterpillars than suburban young and weighed more at fledgling. Differences in ecological conditions between rural and urban sites have an important impact on reproduction in avian species. The rural and suburban site of this study are characterized by different landscapes and composition of tree species. We found that deciduous and native tree species were predominant in the rural site, whereas the suburban site was mainly composed of coniferous species and non-native deciduous trees. This difference is likely to cause a large decrease in insect abundance 38,39 , especially for caterpillars of the Phryganidia spp which are more abundant on deciduous trees rather than on coniferous trees 40,41 . In this regard, similar to previous studies 30,31,37,42 we found that caterpillar biomass was higher in the rural site compared to the suburban counterpart. Moreover, we noticed a difference in the caterpillar phenology between sites such that the peak of caterpillar production was early by ca. seven days in the suburban site. Early caterpillar emergence and production could be the results of urban heat island effect 28 , which is known to advance vegetation phenology 43 . In this study, however, we have not explored differences in leaf emergence phenology and insect preference for native  www.nature.com/scientificreports/   www.nature.com/scientificreports/ or non-native plants between the sites and therefore we cannot draw firm conclusions about the mechanisms causing an earlier and smaller caterpillar peak in our suburban site. Lower caterpillar biomass at the suburban site was associated with lower provisioning rates at the nest and lower proportion of caterpillar fed to the offspring by house wren parents. Differences in parental behaviours such as provisioning rate and type of food item delivered at the nest between rural and urban environments are well documented 30,37 . Some studies on passerine birds have shown either an increase in provisioning rate in urban sites 30,37 or no difference with rural ones 32 , possibly because parents at least compensated for the lack of primary food source by bringing larger quantities of other prey types, such as spiders, flying insects, and beetles. We found that provisioning rate was lower in the suburban site even though parents provided a more variegated diet to the offspring. Discrepancies in provisioning rates at urban sites among studies could be due to different insect or food availability between the study locations or differences in prey selectivity by the parents 44,45 . We show that parents in our suburban site increased the proportion of alternative food items, which are of lower quality 46 and may be responsible, together with lower provisioning rates, for lower offspring mass in the suburban site than in the rural one 47 . A previous study on house wrens also found nestlings weighting less at suburban sites than rural ones with authors suggesting that it may be due to lower average quality of prey 48 . We provide support for this hypothesis with similar findings of lower offspring mass in suburban areas. Lower weight at fledging is linked to lower survival, especially for migrating bird species 49 . Although urban wrens may raise similar numbers of offspring, fewer may survive past the winter. These fitness differences have implications for population dynamics in urban areas 50,51 .
Interestingly, we also found that males provided more caterpillars than females, but only in the rural site. These sex differences in prey delivery suggest that parents may use different foraging strategies [52][53][54] or that males are better foragers of high-quality prey. Alternatively, females may be more responsive to begging behaviours of the offspring and may favour consistency over quality in their provisioning 55 . In addition, we found that at both sites, males have less regular feeding intervals compared to females despite having similar provisioning rates. This could be explained by males foraging in different locations more often than females 56 . Further studies on sex differences between rural and urban populations are needed to better understand whether urbanization poses different selective pressures on the sexes. We expected more irregular provisioning trips at the suburban site, possibly due to a more fragmented habitat, but we did not notice a difference with the rural counterpart. This could be explained by suburban parents foraging on the same number of patches but choosing less profitable prey items to maintain regular feedings to the offspring. Radio tracking studies are needed to better investigate the relationship between visit patterns at the nest (provisioning rate and regularity) and foraging behaviour between urbanized and rural environments.
Pair coordination of the provisioning differed between sites, with higher alternation and synchrony of nest visits in the rural site compared to the suburban one. Why do rural parents have more coordinated provisioning? Four non-mutually exclusive scenarios are possible. First, spatial heterogeneity (habitat fragmentation), lower food availability or human disturbance in the urban environment may decrease coordinated behaviours because parents may need to forage in different locations 24 or further away from the nest 57 . Empirical evidence in zebra finches Taeniopygia guttata showed that parents foraging independently from each other at different foraging areas decreased synchrony of nest visits 16 . Urban areas, such as our suburban site, may thus induce parents to forage independently from each other more often, resulting in lower coordination of their nest visits. A recent radio-tracking study on blue tits Cyanistes caeruleus showed differences in foraging behaviour between www.nature.com/scientificreports/ populations at an urban and rural site in that provisioning trips occur on average further from the nest for parents nesting in an urban site compared to a forest site 57 . However, to the best of our knowledge, it is not known whether urban environments also affect other aspect of the foraging behaviour, such as the number of foraging locations or the coordination between the parents (whether parents forage together or independently from each other).Further telemetry studies with both parents tracked simultaneously are necessary to shed lights on space use and pair coordination of foraging parents in urban areas. Second, urban and rural environments may have different predation risks affecting parental provisioning 58 59 . A review on predation risk in urban areas reported a consistent decrease of predation rate along rural-to-urban gradients on several continents 59 (note however that this effect seems to be only valid for natural nests) 60 . Synchrony of the nest visits has been considered as an antipredator behaviour, which minimizes conspicuous activity at the nest 15,17,18 . Therefore, parental coordination might be higher in rural areas in response to higher predation risk. Third, there could be an age difference between individuals nesting in rural and urban sites leading to more coordinated care associated with increased experience 61,62 . Studies have suggested an age-specific settlement and an habitat-dependent survival in urban or novel sites [63][64][65] . For example, if the individuals settling in urban areas are younger and less experienced, this age-related distribution could explain our results. Lastly, urban noise disrupts communication and acoustic coordination [66][67][68] . There is good evidence that songbirds sing at different frequencies in the presence of urban noise pollution 69 . Therefore, urban areas may disrupt vocal communication, both between parents and between parents and offspring that facilitate coordinated behaviours 70 . These four scenarios could be experimentally tested using existing rural and urban populations. As such, urbanization research could provide valuable insights on the extent to which pair coordination results from or is constrained by environment characteristics (e.g. habitat fragmentation, predation risk) or by pair characteristics and behaviour (e.g. experience and vocal communication). Our findings that urbanization is associated with reduced parental coordination have important potential fitness consequences for offspring. First, less alternating parents could monitor and respond to their partner's activity less often, and in a turn-taking framework, they are expected to invest at a rate which is lower than their maximum 13,71 . Therefore, urbanization could in theory strengthen sexual conflict between the parents with negative consequences for offspring growth and fitness. In this respect, we found that suburban pairs fledged lighter chicks. However, it is notoriously difficult to assess the impact of reduced alternation alone on offspring fitness in correlative studies 72,73 , especially in situations in which fledgling weight is also likely to be function of provisioning rate, food availability and habitat composition such as in this study. Experimental manipulations of one parent's investment (e.g. via handicapping or selective playback experiment) with a concomitant food supplementation (to eliminate environmental constraints on parental care) might be a reasonable approach to investigate the effect of pair coordination alone on offspring fitness. Second, reduced synchrony of the nest visits in urban environments could also strengthen offspring conflict over resource allocation. It has been shown that synchronized feedings at the nest are related to more equal division of food between offspring in a cooperative breeding bird 19 . However, in this study we did not investigate food partitioning between the offspring and cannot explore differences in food allocation between rural and suburban nests.
Our study indicates that, despite a relatively small sample size of nests, different levels of parental coordination exist between pairs breeding in a rural environment compared to a suburban one and discusses how these differences could be driven by diverse ecological and environmental conditions. We promote further studies on replicated urban and rural sites to assess the generality of our findings. Furthermore, we emphasize that comparing populations breeding along an urbanization gradient represents a valuable tool to study environmental effects on parental behaviours and advocate new studies on the behavioural mechanisms driving parental coordination.

Methods characterization of suburban and rural sites.
We conducted our study from May to July 2018 at one suburban and one rural site in Reno, Nevada, USA (Table 1), which were set up with artificial nest boxes since 2016. The distance between the two sites is 10.8 km. Our suburban site was located near Caughlin Ranch, which is a suburban park (Supplementary Fig. S1). This park is located within a suburban neighbourhood with paved walkaways and artificial ponds that fragment the green spaces in pockets of vegetation. Our rural site was the University of Nevada, Reno, Agricultural Experiment Station, which is a university owned agricultural farm with ~ 1000 acres of farmland and pastures (Supplementary Fig. S1). The nest-box population at the rural site was set up in a riparian habitat along the Truckee River, in which vegetation is condensed in tree clusters along two lines. These two sites differed in terms of urbanization score and vegetation composition (Table 1). Urbanization score was estimated as the land use of each study site, using the validated method described by Seress, et al. 74 . This approach divides an aerial image of the 1 km 2 area around each study site into 100 × 100 m cells and then scores the abundance of vegetation, buildings, and paved surfaces, such as roads and parking lots, in each cell. The suburban site has a higher urbanization score than the rural counterpart by having more cells with increased building density and paved surfaces and decreased vegetation density (Table 1). Furthermore, a complete tree census, where we marked individual trees with GPS points (handheld Garmin GPSMAP 62st), indicates that vegetation composition differs between the two sites, in that coniferous trees are predominant in the suburban site, while the rural site is mainly composed by deciduous trees (Table 1).
House wrens are secondary cavity nesters that readily make use of manmade nest-boxes. They prefer open woodland habitat, rarely nesting more than 30 m from woody vegetation but also avoiding dense wooded nest sites 75 . Both males and females feed offspring with a diverse diet of invertebrates, with adult lepidoptera and caterpillars (49%) making up the bulk of the food items brought to offspring and spiders (32%) as a second choice 76,77 . estimation of caterpillar abundance. From mid-April to mid-July, we collected caterpillar frass (n = 80) under oak trees at both sites using 1 × 1 m 2 cheesecloths (n = 4 per site per 10 sampling dates). We emptied all Scientific RepoRtS | (2020) 10:15888 | https://doi.org/10.1038/s41598-020-72951-2 www.nature.com/scientificreports/ nets every week (7.1 ± 0.15 [mean days between frass sample collection ± SE)] at the same time (0800-0900 h). We dried frass for 2 h in a 60 °C oven and then picked out the frass under a dissecting microscope (40 × zoom). We weighed the dried mass to the nearest 0.0001 g. Caterpillar biomass was estimated after correcting for temperature using the methods described in Welbers, et al. 78 .
Collection of provisioning data. From the beginning of May, we monitored house wren nests every week to determine the onset of egg laying and incubation at both field sites. We then checked active nests daily from the day before the predicted hatching to determine the exact hatch date (day 0). At day 8 of chick age, we caught the parents at the nest and banded them with a unique combination of coloured rings. At day 10, we observed parental behaviour for one hour in the morning for 19 nests (ten located in the rural site and nine in the suburban site). At day 15, we measured chick weights at fledging (between 0800 and 1000 h). For this study we only used unmanipulated nests which were not part of a cross-foster experiment 47 Table 1). For each parental visit at the nest, we noted: (1) the sex of the visiting parent (identifiable by the ring colour combination), (2) the time that the bird entered the nest-box (to the nearest second), and (3) the type of delivered prey divided into three categories: caterpillars (lepidopteran larvae), spiders and others (flying insects and beetles). Unidentifiable items represented 1% (n = 4) of 293 total provisioning trips and were excluded from the analyses. A pilot study with behavioural observations and video recordings of house wren nests in the previous year indicated a 98% accuracy of behavioural observations in identifying prey items for each visit. All of the data were collected under the appropriate state and federal permits and approved IACUC protocols.
calculation of alternation and synchrony of the nest visits. From the sequence of nest visits, we calculated pair coordination, measured as alternation and synchrony of the nest visits. We defined alternated visits as visits of one individual that followed a visit of its mate. For the calculation of alternation from a sequence of nest visits (e.g. MFFMFMFMM), visits can occur at any time, and by either parent, after the previous one. We expected different amounts of alternation to arise by chance in a sequence of visits depending on the proportion of visits by the two parents. In situations in which, for instance, one parent makes either all or none of the visits in a sequence, no alternated visits can occur. Conversely, when parents feed the offspring at similar rates, the proportion of alternated visits we expected by chance increases. To account for this effect, we used an alternation score to measure the deviation of the observed amount of alternation from that expected given the relative contributions (provisioning rates) of the two parents using the following formula from Baldan, et al. 21 : An alternation score of zero represents the amount of alternation expected by chance, a value of less than zero indicates that the observed alternation is lower than expected by chance, whereas a value of greater than zero indicates that the observed alternation is greater than expected by chance. See Baldan, et al. 21 for a detailed explanation of the calculation of the alternation score.
We also calculated the proportion of synchronized visits as the number of synchronized visits over the total number of visits. Synchronized visits were defined as a pair of visits (one by each parent), which occurred within one minute of each other. Like previous studies 17,79 , we used a 1-min window to calculate synchrony to minimize the risk that synchronized visits could occur by chance (see Supplementary Fig. S2 for distribution of the time intervals between two consecutive visits). In our dataset, males and females in the rural site visited the nest on average 10.2 and 9.9 times per hour respectively, whereas in the suburban site they visited on average 5.6 and 7.1 times per hour respectively. If parents were visiting the nest independently from each other, we would expect that parental visits occurring by chance within 1 min of each other would be less than 3% [(10.2 male visit rate/60 s) × (9.9 female visit rate/60 s)] in the rural site and 1.1% [(5.6 male visit rate/60 s) × (7.1 female visit rate/60 s)] in the suburban site.

Statistical analyses.
To investigate differences in caterpillar abundance between the two sites, we used two approaches. First, we used a linear mixed model (LMM) to test whether daily caterpillar biomass differed between sites and over the breeding season. We fitted 'daily caterpillar biomass' as response variable, 'date' and its quadratic term in interaction with 'site' (suburban and rural) as fixed effects and 'frass net ID' as a random effect to account for repeated measures. We fitted this LMM with the lme function (nlme package 80 ), allowing heterogeneous variances (heteroscedasticity) between the two sites (varIdent argument within the lme function). Second, we investigated whether the timing of the caterpillar peak differed between sites. For each frass net, we estimated the date of maximum peak using the cardidate R package 81 . This methodology fits curves to environmental time series using Weibull-Functions and estimates the beginning, maximum and end dates of ecological processes 81 , such as the phenology in caterpillar biomass. We then compared the time of caterpillar peaks between sites (expressed as Julian date) using a two-samples t test. We investigated whether the proportion of prey type delivered at the nest differed between sites. We fitted multinomial logit models 82 to model the proportion of prey delivered (divided into three categories: caterpillar, spiders and others) in relation to 'site' and 'sex' and their interaction. 'Nest ID' was included in the model as random effect, and the prey counts (from www.nature.com/scientificreports/ which the proportions are derived) were included as weight. Multinomial logit models were fitted using the function mblogit in the mclogit package 83 .
To explore whether parental behaviour differed between suburban and rural nests, we first explored parental provisioning rates and regularity. For each parent we calculated individual provisioning rate as the number of provisioning trips at the nest per hour. We fitted a LMM with 'individual provisioning rate' as the response variable, 'site' , 'sex' , and their interaction as fixed effects, while controlling for 'brood size. ' 'Nest ID' was included in the model as the random effect. Individual provisioning rate was log transformed to normalize the model residuals. We then explored male and female regularity of the inter-visit intervals (time intervals between two consecutive visits by the same parent, henceforward abbreviated to IVIs). Similarly to a previous study 21 , we expressed regularity as the coefficient of variation (CV) of the IVIs (i.e. standard deviation/mean). Low CV values indicate higher regularity of the IVIs (lower standard deviation compared to the mean), whereas high CV values indicate lower regularity of the IVIs (higher standard deviation compared to the mean value). Here we fitted a LMM with 'CV of the IVIs' as the response variable, 'site' , 'sex' , their interaction and brood size as fixed effects, and 'Nest ID' as the random effect. We then investigated whether alternation of nest visits differed between suburban and rural nests. Here we fitted a linear model with 'alternation score' as the response variable, 'site' as factor and 'brood size' as a covariate. Furthermore, we explored whether synchrony varied between sites by fitting a generalized linear model (family quasi-binomial to control for overdispersion; overdispersion parameter = 2.02) with proportion of synchronized visits as the response variable, 'site' as a fixed effect and 'brood size' as covariate. In this analysis, we also included 'total number of visits' and 'proportion of male visits' as covariates, as we expected (1) the amount of synchronized visits to decrease as the difference in proportion of feeds by the two parents increases and (2) synchrony increases at higher feeding rates, as it increases the chance that two visits can occur within 1 min from each other. Lastly, we tested whether fledging success (number of young fledged) differed between rural and suburban nests. Here we fitted (1) a generalized linear model (family quasi-Poisson to control for dispersed data; overdispersion parameter = 0.15) for fledgling number, and (2) a LMM for individual chick weight.
All the statistical analyses were performed in R environment (version 3.6; R Development Core Team, 2017). All mixed models were performed with the lmer function in the lme4 package 84 . For all models with interaction terms, we first tested whether the interactions were significant. If the interaction terms were non-significant, they were removed from the final model. Cohen's d and its 95% confidence interval were calculated as a measure of effect size for the variable site in our models 85,86 . Post-hoc tests were carried out using the emmeans function in the emmeans package 87 . Significance was taken at α = 0.05 and all model assumptions were met.

Data availability
The datasets analysed during the current study are available from the corresponding author on reasonable request.