Nests in the cities: adaptive and non-adaptive phenotypic plasticity and convergence in an urban bird

Abstract

Phenotypic plasticity plays a critical role in adaptation to novel environments. Behavioural plasticity enables more rapid responses to unfamiliar conditions than evolution by natural selection. Urban ecosystems are one such novel environment in which behavioural plasticity has been documented. However, whether such plasticity is adaptive, and if plasticity is convergent among urban populations, is poorly understood. We studied the nesting biology of an ‘urban-adapter’ species, the dark-eyed junco (Junco hyemalis), to understand the role of plasticity in adapting to city life. We examined (i) whether novel nesting behaviours are adaptive, (ii) whether pairs modify nest characteristics in response to prior outcomes, and (iii) whether two urban populations exhibit similar nesting behaviour. We monitored 170 junco nests in urban Los Angeles and compared our results with prior research on 579 nests from urban San Diego. We found that nests placed in ecologically novel locations (off-ground and on artificial surfaces) increased fitness, and that pairs practiced informed re-nesting in site selection. The Los Angeles population more frequently nested off-ground than the San Diego population and exhibited a higher success rate. Our findings suggest that plasticity facilitates adaptation to urban environments, and that the drivers behind novel nesting behaviours are complex and multifaceted.

1. Background

Phenotypic plasticity can manifest in an individual's development and behaviour, and may help facilitate adaptation to new environments [1,2]. Behavioural plasticity has been extensively documented in a wide variety of animals, and in behaviours ranging from communication [3–6] to nest-site selection [7,8] to breeding phenology [9,10]. Where organisms encounter unfamiliar conditions due to environmental change or by colonizing a novel environment, plasticity may allow for more rapid adaptation to new conditions than slower mechanisms such as evolution by natural selection [11,12].

While plasticity can facilitate local adaptation, plasticity is not always adaptive. Recent work has shown that non-adaptive plasticity is more common than typically assumed [13]. Plasticity in novel environments can be maladaptive when individuals react with unfamiliar conditions in ways that lower fitness [11]. These mismatches can lead to evolutionary traps, in which maladaptive behaviours contribute to local population decline or extirpation [14]. Examples of this include preferentially establishing territories in suboptimal habitats or building nests in substrates with higher predation pressure [15]. Thus, it is important to evaluate the effect of plasticity on individual fitness to determine whether it is adaptive, non-adaptive or maladaptive.

For either adaptive or non-adaptive plasticity to be expressed, individuals rely on environmental cues [2]. These cues may be public information accessible to all individuals [16] or private information available only to a given individual [17]. Private and public cues may influence the same behaviour. For example, individuals might select nest sites based on public information on predator distribution [8,18,19] along with private information in the form of previous nesting experience [7,20]. One of the simplest forms of plasticity based on private information is the ‘win–stay/lose–switch’ (WSLS) strategy. In this strategy, individuals repeat a behaviour if it led to a successful outcome, but choose a different behaviour otherwise [17,21,22]. By allowing individuals to explore alternative behaviours and learn which behaviours maximize fitness, private information strategies such as WSLS can facilitate local adaptation to a novel environment.

Because cities are evolutionarily novel environments with unfamiliar challenges for wildlife [23,24], urban landscapes provide ideal systems for understanding how plasticity might promote or hinder adaptation to new environments [25]. Urban birds are particularly excellent systems for studying the effects of behavioural plasticity on urban adaptation as they are frequently found in close proximity to humans, are generally diurnal and easy to observe, and their various behaviours, such as foraging, nesting and singing, are impacted by urbanization [26,27]. Unsurprisingly, behavioural plasticity in urban birds has been extensively documented, and the extent of plasticity may play a role in allowing species to colonize or persist in cities. For example, plasticity has been implicated in changes in vocalization [6], fear response [28–30], breeding season length [31,32] and nest-site selection [33,34] in response to urbanization. However, whether plasticity is adaptive for urban populations is less well understood [25]. A still outstanding question is to what extent plasticity in urban wildlife populations impacts fitness outcomes, and whether urban ecosystems might selectively filter out species that either do not exhibit behavioural plasticity or show maladaptive plasticity in their new environment [35].

Behaviours associated with nesting and reproduction are critically important for population persistence. As a result, birds that thrive in urban areas frequently exhibit novel nesting behaviours. Many urban populations of raptors, swifts and swallows have adapted to use buildings and other artificial sites for nests, in lieu of natural cliffs and riverbanks [36]. Tawny frogmouths (Podargus strigoides) [31], dark-eyed juncos (Junco hyemalis) [32] and Eurasian magpies (Pica pica) [37] exhibit a protracted breeding season in urban areas, possibly as a result of mild climates and/or greater food abundance compared with surrounding wildlands. Some urban species, such as house finches (Haemorhous mexicanus), incorporate anthropogenic materials such as twine, plastic or even cigarette butts into the construction of their nests [36], where such novel nest material may reduce parasite load in nestlings [38].

While much has been learned about urban nesting behaviour in recent years, relatively little is known about the predictability of nesting behavioural shifts in urban areas. Cities do not share identical ecological characteristics; instead, they can differ in key traits such as canopy cover, dominant vegetation and per cent impervious surfaces [39]. Consequently, do different urban populations adapt similarly to these novel conditions, or do they adapt differently depending on the unique conditions each population faces? One study has demonstrated that great tits (Parus major) modify song characteristics in similar ways in at least 10 different cities [6], but our understanding of how plasticity in nesting behaviour is expressed in isolated urban populations remains poorly understood.

The dark-eyed junco (hereafter ‘junco’) is an ideal species to study plasticity in urban nesting behaviour and convergence across multiple urban populations. Traditionally a breeding resident of mixed-pine forests in North America, this sparrow has successfully colonized cities throughout southern California since the late twentieth century [40]. Urban junco populations have diverged rapidly from their wildland counterparts through behaviours including a protracted breeding season and sedentarism [32]. While this species typically builds cup-shaped nests on the ground and under vegetation, prior research on an urban population near San Diego, California demonstrated plasticity in nesting sites, including the use of off-ground locations and the placement of nests on artificial substrates [34]. However, we do not know whether this plasticity is replicated across urban populations, or what consequences arise from nesting off the ground or on artificial substrates. Understanding the role of plasticity in nesting success and its predictability across cities is essential to establishing what makes this urban colonizer successful.

Here, we examine plasticity in nest-site selection in an established population of juncos in urban Los Angeles. We ask three questions. (i) Is plasticity in nest-site selection adaptive or non-adaptive? (ii) Do juncos practice ‘informed re-nesting’ by changing nest-site characteristics based on prior nest outcomes? (iii) Are nesting plasticity and nest outcomes convergent between two urban junco populations? We predicted that behavioural plasticity in nest-site selection would be adaptive, and that use of off-ground, artificial nest sites would increase nest success, similar to findings from the San Diego population [34]. Furthermore, we predicted that breeding pairs would be more likely to change nest heights and substrate types after a failure than after a success. Finally, we predicted that the Los Angeles population would show a similar proportion of nests above the ground as the San Diego population, with a corresponding similarity of success.

2. Methods

(a). Study sites

Field work was conducted on the campus of the University of California, Los Angeles (UCLA). The campus is located in the foothills of the Santa Monica Mountains and is surrounded by an urban matrix including commercial development, residential communities and urban green space. The nearest significant natural habitat lies 2 km away from the Santa Monica Mountains. The campus is heavily urbanized and is characterized by impervious surface cover (buildings and walkways) and extensive grassy lawns interspersed with ornamental trees (particularly Pinus canariensis and Eucalyptus sp.). While the native groundcovers and shrubs that juncos typically nest in are absent from this site, they are known to readily use ornamental groundcover species such as ivy (Hedera sp.) that are abundantly planted as landscaping. Juncos did not breed in urban Los Angeles before 2000 [41], but commonly wintered throughout west Los Angeles [40]. Juncos were first recorded summering in nearby neighbourhoods sometime between 2004 and 2008 (Jared Diamond 2019, personal communication). Fledglings were seen on the UCLA campus by summer 2008 (P. Yeh 2008, personal observation). They have expanded their breeding range throughout much of west Los Angeles. The UCLA campus junco population consists of approximately 100–120 breeding pairs.

To test for behavioural convergence in southern California juncos, we used previously published data collected from an urban junco population at the University of California, San Diego (UCSD) in San Diego County, California. Juncos colonized this suburban campus in the early 1980s and rapidly established a resident population on and around the campus [32]. Data on San Diego juncos were collected between 1998 and 2002; so, data were collected between 15 and 20 years after colonization at both sites. The San Diego population is sedentary and is isolated from the Los Angeles population by over 170 km, so there is likely negligible intermixing between these two populations, although genetic connectivity has not been tested. Owing to the similarity in land use between the two sites, we predicted that juncos would nest off-ground at similar rates.

(b). Field methods

Adult juncos were captured and banded in Los Angeles from January to July in 2018 and 2019. Juncos were captured with mist nets using audio playback of regional junco songs. The majority of juncos captured in this way were territorial males, but some females were also captured. Once captured, juncos were immediately extracted from the net and fitted with one aluminium USGS leg band along with a unique combination of three coloured plastic leg bands, allowing individuals to be reidentified in the field after release. Individuals were sexed using primary sexual traits and plumage characteristics. Blood, cloacal swabs and morphometrics were taken, after which the birds were released.

Beginning in March 2019, pairs of juncos were monitored for nesting activity. Nest building began in early to mid-March and was carried out by both sexes by bringing nesting material (primarily pine needles, dead grass and hair) to a concealed spot on the ground, in a shrub or tree, or on an artificial surface. When a nest was found in the nest-building phase, it was monitored once every second day until the first egg was laid, and the first egg date determined. Subsequently, nests were checked after two weeks, close to the age of hatching. Other nests were found during the incubation by following females exhibiting bouts of rapid feeding followed by disappearance. Nests found at this stage were monitored once per week until hatching. Nests were also found at the nestling stage by monitoring parents bringing food back to the nesting site. Nestlings were aged in situ, and were banded, measured and sampled for blood 7 days after hatching. To avoid disturbance, measurements were taken when parents were away from the nest, and the process was carried out in 5 min or less. After banding, nests were monitored every 2–3 days until fledging occurred or the nest failed. Some nests were located in inaccessible areas (e.g. high in trees or buildings). These were monitored every 5–7 days until fledging. Approximately 10 nests were found when the nestlings were older than 8 days; to minimize the probability of pre-fledging, we did not band these chicks, but simply monitored their nests every 2–3 days until fledging occurred. When we did not observe the first egg date, we were able to estimate the first egg date for most nests by back-calculating from the observed age of nestlings, with the exception of inaccessible nests and nests that were predated prior to hatching. Pairs produced between two and four nests per year. Females initiated new nesting attempts soon after a prior nest was fledged or predated (within a few days to two weeks). The latest nests were banded at the end of July, and the nesting season was complete by the second week of August, at which point captured adults had begun molting wing and/or tail feathers.

Multiple characteristics of nests were measured. Nest height above the ground was estimated visually by a single trained observer. After nesting concluded, we revisited sites to identify the vegetation of nesting substrate. Juncos nested in a wide variety of vegetation species. To facilitate analysis, these substrates were lumped into nine different functional types: debris, bunchgrass, herbaceous/non-woody perennial, prostrate shrub, erect shrub, vine, tree and artificial substrate. Artificial substrates encompassed all unnatural surfaces on which nests were directly placed, primarily on buildings but also on concrete surfaces.

Nesting attempts were considered successful if they produced at least one fledgling. Fledglings were located by careful monitoring of the territory surrounding the nest to watch for parents delivering food to concealed young. Nests with chicks that disappeared well before fledging (less than 10 days since hatching) and those abandoned or predated during incubation were considered failed nests. Owing to greater than 900 h of cumulative monitoring, we have very high confidence of the recorded nest outcomes. Only individuals for whom all nests were found and their outcome determined were used in our analysis. All parental pairs included at least one banded individual. While it is possible that individuals might have swapped mates over the course of the breeding season, we consider this unlikely due to observed intra-season pair fidelity by pairs where both males and females were banded. We did not measure success by the number or proportion of nestlings that successfully left the nest due to the lack of resources to extensively monitor young post-fledging. In addition, we could not determine the number of nestlings in nests placed in inaccessible locations, such as in trees or high on buildings.

(c). Statistical analysis

Multiple analyses were conducted to test for behavioural plasticity and its impacts on breeding populations of juncos in Los Angeles and San Diego. In each analysis, we included 130 nests produced by 48 breeding pairs during the 2019 breeding season in Los Angeles and 579 nests monitored between 1998 and 2002 in San Diego. Models were built using a combination of generalized linear models (GLMs) and generalized linear mixed models (GLMMs) and run in R v. 3.6.1 [42] with the lme4 package [43].

(i). Adaptive or non-adaptive plasticity

We examined the binary response of nest success as a function of two plastic traits—nest height and nest substrate—using a GLM with a logit link. To increase the power of our statistical analysis, we consolidated nest-substrate data into two categories of a binary variable: nests placed directly on an artificial surface or on vegetation, such as vines, growing on an artificial surface, were categorized as ‘artificial’. All other nests were categorized as ‘non-artificial’. We tested three different models: one that included nest height but not artificial substrate, one that included artificial substrate but not nest height and one that included both variables, as well as their interaction coefficient, to evaluate the explanatory power of each variable separately as well as together, and to account for covariance. Because many off-ground nests were difficult to monitor, it is possible that nests above the ground might have been consistently found at a later stage than on-ground nests. Consequently, each model also included age of nest at discovery and first egg date as fixed effects. As pairs produced two to four nests during the breeding season (median = 3), we initially included pair identity as a random effect. However, this term was dropped from the model as it did not contribute to explaining total variance. We were unable to determine the first egg date and discovery age for some nests; to avoid discarding samples, we imputed values for these nests using the multiple imputation package mice [44]. Similar results were obtained when data were not imputed. We generated values using a stochastic regression method, with vegetation type, discovery age, first egg date and nest order as predictors. Five imputations were generated five times, and the model p-values were subsequently pooled. All continuous covariates—nest height, first egg date and discovery age—were scaled to a mean of 0 and standard deviation of 1 prior to modelling. Variables were considered to have predictive power if their p-value in the best candidate model was less than 0.05. We expected that nests located on artificial substrates to be highly correlated to off-ground nests, owing to the location of most off-ground nests. To determine if one of these two correlated variables was a primary predictor of success, a variance inflation factor (VIF) was determined using the package car [45], with a VIF over 5 indicating that one of the variables was a dominant predictor of nest success.

(ii). Informed re-nesting

We defined informed re-nesting as individuals selecting nest sites based on the WSLS model [7]. We used Pearson's χ² test of independence to examine whether parent juncos changed their nest height (off-ground versus on-ground), or substrate preference (switching from one to another of the nine vegetation functional types, or switching between artificial and non-artificial substrates) at a greater frequency after a nest failure than after a success, in accordance with WSLS. To determine if WSLS was adaptive, we examined nests that were built after the previous nest failed, and used Pearson's χ² test of independence to evaluate whether nest success differed by either substrate reuse or nest height reuse.

(iii). Behavioural convergence

Behavioural convergence between the Los Angeles and San Diego populations was tested by comparing all of our Los Angeles nest records with data collected from 579 nests by Yeh et al. [34] from 1999 to 2002. We used Pearson's χ² test of homogeneity to test for a difference in the proportion of off-ground nests between the sites. To determine if success varied between the two sites and by nest height, we used a binomial GLMM with nest success as a response variable, population location and nest height (off-ground versus on-ground), as well as their interaction, as fixed effects, and year of data collection as a random effect.

3. Results

(a). Nest characteristics of the Los Angeles population

In total, 170 nests were found in the Los Angeles population during the 2019 breeding season. Of these, 130 nests belonged to 48 pairs that were intensively monitored throughout the breeding season in Los Angeles and were used for statistical analysis. The 130 nests were found in 37 taxa of vegetation. Hedera sp. was the most frequently used nest substrate, with 28 nests (21.5% of the total), followed by Lantana sp. (lantana) and Parthenocissus tricuspidata (Boston ivy) with eight and six (6.2% and 4.6%) nests, respectively. A plurality of nests was found in herbaceous cover (39; 30%); the next most abundant functional types used were: grass/bunchgrass (18; 14.3%), artificial (18; 13.8%) and prostrate shrub (16; 15.1%). Erect shrubs (12; 9.5%), vines (12; 9.5%), trees (8; 6.2%) and debris (7; 5.6%) were used less frequently as substrates (see electronic supplementary material, table S1).

Eighteen nests (13.8%) were located directly on artificial structures, along with 12 (9.5%) that were placed on vines growing on the walls of buildings, for a total of 30 nests (23.1%). Generally, nests located directly on buildings were placed on horizontal ledges on the sides of buildings. For example, one pair nested four times in two abandoned black phoebe (Sayornis nigricans) nests 7 m above the ground in the inner eave of a building. Another pair nested twice on top of an external window shutter. One pair nested twice in the stairwell of a building, connected via skylights to the outside. One pair nested four times in two unusual locations, alternating between nesting underneath a cardboard box in a trench and 5 m below ground in a gutter (see electronic supplementary material, photographs).

Forty-nine nests (38%) were located off the ground—on artificial structures, vines, trees or shrubs. Nest heights of off-ground nests were highly variable, ranging from less than 0.25 m off the ground (near the base of a Nandina bush) to more than 8 m off the ground. Off-ground nests were found in both artificial substrates, such as on horizontal ledges and eaves over 10 m high, as well as on non-artificial substrates, including a nest placed over 12 m high in a palm tree.

(b). Adaptive or non-adaptive plasticity

Overall, 95 of 130 nests (71.5%) fledged at least one offspring. We found that nests placed on artificial surfaces or on vines on artificial surfaces were significantly more successful than nests on other substrates (slope ± s.e. = 1.44 ± 0.77; p = 0.02). We found that height above the ground by itself had a marginally positive impact on nest success, but this effect was not significant (slope ± s.e. = 0.45 ± 0.26; p = 0.09). Including both variables in the same model diluted the effect (artificial: slope ± s.e. = 0.18 ± 0.13, p = 0.17; height: 0.02 ± 0.06, p = 0.82), and the interaction coefficient was not significant (slope ± s.e. = 0.02 ± 0.10; p = 0.85). Nest height above the ground and nest substrate were highly correlated (r² = 0.57); however, the VIF test was inconclusive (VIF = 1.32). Neither first egg date (slope ± s.e. = −0.22 ± 0.22; p = 0.32) nor discovery age (slope ± s.e. = 0.23 ± 0.21; p = 0.27) showed a strong influence on nest success based on our full model (table 1).

Table 1.

Summary of GLMs of adaptive plasticity at Los Angeles (n = 130).

parameter	model
	full model		artificial nest only		nest height only
	slope ± s.e.	p-value	slope ± s.e.	p-value	slope ± s.e.	p-value
intercept	0.75 ± 0.24	0.002	0.70 ± 0.22	<0.001	0.99 ± 0.21	<0.001
nest height	0.15 ± 0.30	0.610			0.45 ± 0.27	0.094
date of first egg	−0.24 ± 0.22	0.281	−0.22 ± 0.22	0.316	−0.24 ± 0.21	0.266
nest age at discovery	0.20 ± 0.22	0.349	0.23 ± 0.21	0.268	0.19 ± 0.21	0.356
artificial nest	1.25 ± 0.75	0.097	1.44 ± 0.66	0.022
AIC	153.35		151.66		154.52
McFadden's R2	0.066		0.065		0.045
Nagelkerke R2	0.107		0.104		0.074

(c). Informed re-nesting

WSLS was practiced by at least some members of this population. Juncos were more likely to choose a different functional type of vegetation after a failure than after a success (χ² = 4.93, d.f. = 1, p = 0.03; figure 1). Pairs reused the same substrate 68% of the time if their previous nest was successful, compared with only 40% of the time if their previous nest had failed. Prior success also showed a marginal correlation with re-nesting at a different height, though this effect was not significant (χ² = 2.85, d.f. = 1, p = 0.09). Pairs re-nested at a similar position, either on-ground or off-ground, 79% of the time following a successful nesting attempt, compared with only 59% of the time following a nesting failure. A post hoc analysis to further test whether failure shifted the direction or magnitude of change in nest height was not significant (difference in the direction of shift: Pearson's χ² test of independence, χ² = 0.17, d.f. = 3, p = 0.92; difference in the magnitude of shift: Wilcoxon rank sum test, W = 635.5; p = 0.96). Pairs reused artificial or non-artificial substrates 95% of the time after a prior failure, compared with only 90% of the time after a prior success, though this difference was not significant (χ² = 0.49, d.f. = 1, p = 0.48). Finally, pairs that changed either nest substrate or off-ground status after a prior failure were not more successful than those that used the same substrate and height, although the sample size was small (n = 21; substrate: χ² = 1.66, d.f. = 1, p = 0.20; height: χ² = 0.42, d.f. = 1, p = 0.52). We conducted further post hoc analysis to determine whether pairs that followed WSLS more frequently, or that shifted nest height more after nest failure, exhibited greater nesting success, but did not find any significant effect (see electronic supplementary material).

Figure 1.

Substrate reuse following a nest failure (n = 21) and success (n = 62) in Los Angeles. Substrate types are described in the Field Methods section. Difference in the substrate reuse rate was significant (p = 0.03).

(d). Behavioural convergence

Both the Los Angeles and the San Diego populations had a sizable proportion of off-ground nests. However, off-ground nests were significantly more frequent in the Los Angeles population, 35%, than the San Diego population, 13% (χ² = 40.0, d.f. = 1, p < 0.001; figure 2). Controlling for nest height, nest success was also significantly higher in the Los Angeles population (slope ± s.e. = 0.88 ± 0.41, p = 0.03), where 71% of nests in Los Angeles were successful compared with only 50% of San Diego nests (table 2). In both locations, nest success was much higher in off-ground nests (slope ± s.e. = 1.24 ± 0.27, p < 0.001). This difference is potentially due to the differential success of on-ground nests between the two areas: in Los Angeles, 67% of on-ground nests were successful compared with only 45% in San Diego, whereas 82% of off-ground nests in Los Angeles and 77% of off-ground nests in San Diego were successful. However, the model of nest success did not support this interaction between location and nest height (slope ± s.e. = −0.47 ± 0.52, p = 0.39).

Figure 2.

Off-ground nesting rates at Los Angeles (n = 130) and San Diego (n = 579). The difference in the proportion of off-ground nests was significant (p < 0.001).

Table 2.

Outcomes of off-ground and on-ground nesting attempts at each study site (n = 709).

study site	nest height	no. successful nests	no. failed nests
San Diego	off-ground	57	23
	on-ground	223	276
Los Angeles	off-ground	40	9
	on-ground	54	27

4. Discussion

We found evidence that nest-site selection by dark-eyed juncos that have recently colonized urban Los Angeles was adaptively plastic and identified the WSLS strategy as a potential mechanism. Plasticity was demonstrated by the use of a wide variety of artificial nesting sites—such as on buildings and in artificial trenches—that are not available in wildland sites, along with a high proportion of off-ground nests. Nests built on artificial and off-ground sites were more successful than those built in vegetation or directly on the ground, as predicted by prior research on the San Diego population. As expected, juncos were more likely to use a different nesting substrate if their previous nest failed than if their previous nest was successful. However, contrary to our prediction, this strategy did not increase nest success, although our analysis may have lacked power due to the low rate of nest failure. Our findings generally support the hypothesis that plasticity plays a role in adaptation to urban environments, although not always in the ways we predicted.

Pre-existing plasticity may predispose juncos to make use of novel nest sites in urban areas. Juncos in non-urban populations have long been known to occasionally nest in atypical locations when such opportunities have been available, such as in bait cans [46], and discarded food cans [47]. When the expression of traits is constrained by environmental limitations, cryptic phenotypic variation may persist, only to be expressed when the environment changes [11]. Interestingly, juncos have been observed to use natural embankments in wildland areas as nesting sites [48]. Artificial surfaces such as buildings and walls might mimic these similar natural surfaces. Peluc et al. [8] experimentally demonstrated that orange-crowned warblers (Leiothlypis celata) on California Channel Islands free of avian nest predators altered nest site and provisioning rate when exposed to those predators' calls [8]. Similarly, juncos may retain the ability to recognize vertical surfaces with horizontal ledges or crevices as suitable sites for nesting, even when the availability of such sites is sharply constrained by their environment. This plasticity could then be expressed in novel environments, which may in turn benefit species that harbour greater plasticity.

Another possible explanation for high off-ground nesting rates in cities is a lower quality of on-ground nesting sites. Most on-ground nests in our population were in small patches of ivy or similar vegetation. These may be less appealing to juncos than larger groundcover patches present in non-urban sites as they may be more susceptible to human disturbance or predation. This may explain the significant number of nests placed in trees in Los Angeles, which are readily available yet rarely used in non-urban populations. Further research should quantify both nest-site preference and success at finer scales, such as groundcover patch size, shape, isolation and distance to human traffic to better understand the factors impacting nest-site success in urban areas.

As the vast majority (more than 85%) of nest failures for which a cause could be determined were as a result of predation (such as broken eggs or missing nestlings), it seems likely that the higher success rate of off-ground and artificial nests was due to reduced predation pressure. The likely explanation appears to be that nests placed on buildings were either less visible or less accessible to predators. For example, many nests were placed on inward-facing ledges or under eaves, probably too high for potential predators such as domestic cats (Felis cattus) to climb, and difficult for avian predators such as American crows (Corvus brachyrhynchos) to detect or access. This may also explain why nests built in trees exhibited predation rates similar to on-ground nests. Further studies should determine the fine-scale nest-site characteristics that increase nest success at these sites.

The use of novel nesting sites in the Los Angeles population may contribute to relatively high rates of nest success in the population. We observed a nest failure rate of 28.5% in the Los Angeles population. In comparison, Ketterson et al. [49] observed annual nest failure rates from 20 to 75% in rural Virginia, while Martin [50] found a nest failure rate of approximately 55%. High nesting success in urban bird populations has been attributed to factors such as differing predator composition and alternative food sources for predators. Nevertheless, further research in both urban and non-urban populations is necessary, as nest failure rates are both spatially and temporally variable.

Identifying mechanisms of plasticity is important for understanding how plasticity impacts adaptation [2]. We found that informed re-nesting through WSLS explains some plasticity in junco nesting behaviour. Informed re-nesting may facilitate local adaptation to unfamiliar environments by allowing individuals to learn from prior experience [51,52]. By switching nest substrates after failure, juncos may sample a wider variety of nest-site characteristics, allowing them to discover and use safer sites. In comparing two species, an ‘urban adapter’ and an ‘urban avoider’, Kearns & Rodewald [53] found that the urban adapter altered nest characteristics after failure, while the urban avoider did not. At the same time, juncos in our study did not switch between artificial and non-artificial sites more often after failure than after success, suggesting that they may not evaluate nest-site success using this criterion.

WSLS in nest-substrate choice did not increase nesting success. This runs counter to our expectations based on prior work by Chalfoun & Martin [7], which demonstrated that WSLS in nest location and characteristics increased nesting success in Brewer's sparrows (Spizella breweri) [7]. One explanation for why informed re-nesting may be non-adaptive is that after a failure, an individual may choose between many different substrates, only some of which might increase nesting success. As a result, different individuals may perform informed re-nesting in different ways, with only a small proportion gaining an adaptive benefit. Additionally, plasticity in nest-substrate selection might be accompanied by associated costs [54]. For example, while birds that remain in similar substrates may not happen upon the ideal nest substrate, by changing within-substrate nest characteristics, they may still be able to increase nest success. Finally, nesting success could be more heavily dependent on other factors, such as a parent's ability to obtain enough food for its offspring, or to be stealthy enough to keep predators from finding the nest.

One caveat is that due to a low nest failure rate, our sample size for this analysis was very small, and more data should be collected to better understand the impact of informed re-nesting on success. Juncos could be using private information gathered across multiple years to determine ideal nesting sites. Longer-term, longitudinal studies will allow for more sophisticated models to test the factors that play a role in nest-site selection, such as the role of alternating patterns of success and failures on site preferences.

We found that this behavioural plasticity acted similarly across both the Los Angeles and San Diego populations, suggesting that there is behavioural convergence in these two urban areas. Both populations showed far higher levels of off-ground nesting than has been observed for this subspecies in non-urban locations; prior research on non-urban juncos in western North America found off-ground nesting rates from 0 to 6% [32,55], compared with rates of 13% in San Diego and 37% in Los Angeles. This suggests that plasticity can facilitate behavioural convergence in response to urbanization [6]. Nevertheless, the two populations differed in the magnitude of this shift, with Los Angeles juncos nesting off-ground three times more often than San Diego juncos, suggesting behavioural divergence between the two populations. One possible explanation is that the use of off-ground nest sites is human density-dependent and the Los Angeles site is a more densely built and populated than its San Diego counterpart [56]. Thus, if juncos harbour a preference for off-ground nesting sites, these sites may be more readily available to the Los Angeles population. One method to test this would be to place artificial nest sites such as walls with ledges and eaves in non-urban populations and determine whether juncos use those sites. This would allow the testing of both junco preference for artificial nest sites and their impact on nest success, and further determine how relative abundance of potential substrates impacts nest placement.

Additionally, increased urbanization may lead to more frequent off-ground nesting as a result of increased human disturbance and traffic. Human disturbance can prevent organisms from using urban spaces [57]. However, we found few cases of abandoned on-ground nests, even in highly trafficked parts of our study area, suggesting that juncos may be habituated to human disturbances. Our findings suggest that while conditions in the two cities cause populations' nesting behaviour to change in the same direction, the magnitude of the plastic response differs between populations.

Los Angeles juncos showed higher nesting success rates than San Diego juncos. Increased nesting success may have contributed to the rapid colonization of west Los Angeles by juncos, compared with the more modest range expansion documented in San Diego, where juncos were restricted to the UCSD campus at the time of data collection [34,40]. Interestingly, the difference in success persisted even after accounting for a higher frequency of off-ground nests at the Los Angeles site. Possible explanations include a lower predator density at the Los Angeles site, or the use of more secure nesting locations, though such possibilities remain hypothetical until further data are collected. That on-ground nests in Los Angeles fare better is unexpected, considering the high levels of off-ground nesting; this suggests that use of off-ground nesting sites may not necessarily be a response to predation. However, it is possible that by increasing the number of nest sites available and reducing the density of on-ground nests, off-ground nesting may reduce frequency-dependent predation [58]. The role of novel nesting sites as a driver rather than an outcome of predation rates should be further studied in the context of urban ecosystems.

Ultimately, we found that urban birds demonstrated plasticity and that this plasticity was adaptive, but the outcome of that plasticity varies across populations. For urban adapters, there may be no ‘one-size-fits-all’ way to live in our cities. Understanding the importance of plasticity to urban adaptation will help us understand the patterns of both species and behavioural diversity in cities. Furthermore, understanding how and if convergence occurs within one species is necessary to understand the predictability of adaptation in cities [59]. Findings from one population might not be generalizable to other populations due to variation in fitness trade-offs across populations and cities [60]. Future research should take advantage of cities as independent ecological experiments to better understand how populations adapt to unique urban pressures.

Ethics

All procedures were approved by the UCLA IACUC Ethics Committee and in accordance with IACUC protocol ARC-2018-007. Banding was conducted under United States Geographical Survey Banding Permit 23809 and California Department of Fish and Wildlife Specific Collection Permits S-183270002-18337-001 and S-183040004-18313-001.

Data accessibility

The datasets supporting this article have been uploaded as part of the electronic supplementary material.

Authors' contributions

P.J.Y. designed the study. S.A.B. and E.S.D. conducted fieldwork. S.A.B. and M.W.T. conducted statistical analyses. S.A.B., E.S.D., M.W.T. and P.J.Y. wrote and revised the manuscript.

Competing interests

We declare we have no competing interests.