Experimental manipulation of photoperiod influences migration timing in a wild, long-distance migratory songbird

Abstract

Previous laboratory studies have demonstrated the role of photoperiod in cueing the migration timing of small land birds; however, how migration timing of young birds in wild environments develops in relation to these cues have rarely been investigated. Such investigations can make important contributions to our developing understanding of the phenotypic plasticity of migration timing to new conditions with climate change, where changes in the timing of nesting may expose juvenile birds to different photoperiods. We investigated the impact of manipulating photoperiod during nestling development in a long-distance migratory songbird on the timing of post-breeding movements in the wild. Using programmable lighting installed in the nest-boxes of purple martins (Progne subis), we exposed developing nestlings, from hatch to fledge date, to an extended photoperiod that matched the day length of the summer solstice in Manitoba, Canada. We found that birds with a simulated, earlier photoperiod had a longer nesting period and later fledge and autumn departure dates than control group birds. This study demonstrates the phenotypic plasticity of first-year birds to the ontogenetic effect of their hatch date in the formation of the timing of their first post-breeding movements. Further, we discuss how these results have implications for the potential use of assisted evolution approaches to alter migration timing to match new conditions with climate change.

1. Introduction

Spring advancement due to climate change has resulted in different phenological responses by migratory birds, such as shifts in their migration phenology, or the timing of other seasonal events such as breeding [1]. Among migratory birds, long-distance migrants may be more dependent on their endogenous, circannual schedules to cue their migration departure timing [2,3] as they cannot predict phenological advancement at their breeding ground from overwintering areas that may be thousands of kilometres away [4]. Yet, breeding arrival timing has been observed in many long-distance migratory species to have advanced over years or decades. However, the mechanisms for these changes have been much debated and require further exploration (e.g. [5,6]).

The migration timing of long-distance migratory birds is mainly controlled by endogenous circannual rhythms synchronized to the external cue of photoperiod [2,3,7]. With climate change and corresponding advancing springs, birds that breed earlier can be exposed during nesting to different day lengths (photoperiod), which may further influence timing. Phenotypic plasticity may be a mechanism by which individuals adaptively respond to new environmental conditions [8,9], and that could provide long-distance migrants with a shorter term mechanism for adapting to climate change. A study by Both [10] using band recapture data for European pied flycatchers (Ficedula hypoleuca) found that breeding latitude predicted the timing of median recovery date during spring migration at stopover sites. It was proposed that these patterns were driven by the different day lengths experienced by nestlings at more northern versus more southern breeding latitudes. This led to the inference that migration timing may be flexible to past experiences at the breeding grounds, based upon an ontogenetic effect of photoperiod during nestling development [10]. This is supported by a laboratory study by Coppack et al. [11] on European blackcaps (Sylvia atricapilla) showing that simulating an earlier photoperiod during the nestling phase, resulted in a longer moulting period and later onset of autumn migration. In a rare, experimental field study, delaying the hatch date of pied flycatchers through manipulation of incubation timing resulted in a later spring arrival date in the following spring [12]. However, further study is required in other systems, particularly where photoperiod is manipulated, and the actual movements of individuals are directly tracked in the wild. Indeed, new syntheses of chronobiology and ecology through a ‘wild clock’ approach in a greater variety of study systems are predicted to yield invaluable new insights into how animal timing responds to environmental change [1,13]. Such research would complement previous laboratory and field research to address the current knowledge gap as to whether birds experiencing different photoperiods with earlier nests in advanced springs provide new experiences of zeitgebers that entrain an adaptive response to climate change (ontogenetic effect) in migratory songbirds.

In this study, we build upon laboratory studies by using a wild system to investigate the direct impacts of experimentally manipulated photoperiod on the development of timing in a long-distance migratory songbird. We aimed to experimentally test the hypothesis that the photoperiod experienced by juveniles in the nest during development impacts the timing of their subsequent life cycle events. Using programmable lights installed in the nest-boxes of purple martins (Progne subis) during the nestling phase, we manipulated photoperiod to summer solstice day length to simulate an earlier calendar date, as the hatch date of many birds occurs naturally after this date and birds experience shorter day length. The other environmental factors were the same for all nestlings.

We tagged nestlings and used an automated telemetry system [14] to determine subsequent migration departure dates for experimentally manipulated nestlings and controls. We predicted that nestlings exposed to extended day lengths (photoperiod) simulating earlier calendar dates would have a longer nesting period (pre-fledge) and would spend more time at the colony post-fledge resulting in later autumn departure dates.

2. Methods

We conducted the field component of this study at three purple martin breeding colonies in southern Manitoba, Canada (site 1: 49.734° N, 97.1317° W; site 2: 49.127° N, 97.5703° W and site 3: 50.173442° N, 97.133442° W) which are located at open habitats close to water bodies. This colonial bird is dependent on human-made houses for breeding in this part of its range. The number of purple martin houses at sites 1, 2 and 3 were three units (with 14 nest cavities each), four units (with 32, 12, 14 and eight cavities each) and two units (with 14 nest cavities each), respectively. Purple martins are aerial insectivores that breed in eastern North America and journey 10–20 000 km annually to overwintering locations mostly in the Amazon Basin in Brazil [15]. Spring migration starts in mid-April for our study population and the arrival and departure date at these latitudes is early May and mid-August, respectively [16,17]. To manipulate the photoperiod to test for its impact on the timing of fledge and post-breeding movements, we mounted programmable light units inside treatment nest-boxes. We installed small, light emitting diodes (LEDs) (5 mm length) in the roof of each nest cavity so that they pointed downward toward the nest [18,19]. We selected LEDs because they incorporate the spectrum of natural light, do not produce heat [19,20] and can be programmed to emit selected amounts (lux) of light. Each light was connected by a thin wire to an external unit composed of an Arduino and a real-time clock, mounted on a circuit board. Each of the four units was connected to lights in five cavities and was mounted underneath the housing units along with a rechargeable battery. In total, we used 20 nest cavities for this manipulation. The same number of nest cavities were used as the control group at two breeding sites.

We programmed the LEDs to emit 1.5 lux, which was the average (n = 4) measured at the start of civil twilight before sunrise at our study areas (digital illuminance meter, LX1330B). To experimentally extend day length, lights were programmed to match the day length of the summer solstice (21 June 2019, 16 : 21 : 06) at our study sites. We programmed Arduino units to turn the light on 1 h before sunset and off at the end of the summer solstice civil twilight. In the morning, lights turned on again according to the summer solstice civil twilight and off 1 h after sunrise. Most hatch dates at our study sites were just after the summer solstice [12] when days were getting shorter, so we opted to ‘hold’ the photoperiod at the longest day of the year (the solstice) for experimental cavities to simulate an earlier photoperiod for all nests (electronic supplementary material, figure SF1). We simulated the photoperiod of the summer solstice (21 June), which is the longest day of the year because (i) at more northern latitudes most individuals hatched after the summer solstice, (ii) it is hypothesized that this time of the year could be a baseline to re-set internal clock time [21] and (iii) the timing of the summer solstice in relation to nesting provides the best opportunity to manipulate the photoperiod in purple martin cavities during nesting in a wild environment (i.e. extending day length for nestlings once day lengths are getting shorter).

To track fledge dates and post-breeding movements, we used the Motus Wildlife Tracking System (www.motus.org) which is an international, automated radio-telemetry array of receiving stations [14]. We randomly selected four nestlings from each of the extended day length and control nest cavities when they were near fledging (age 20–22 days). We outfitted each with individually coded, radio nanotags (NTQB-3-2, Lotek Inc.) (0.67 g, 12 × 6 × 5 mm in length, width and height, respectively) using a leg-loop, backpack style harness [22] made of polypropylene thread. The average weight of nestlings near their fledge date is 50 g [23,24], and the weight of the tags we used is, therefore, less than 3% of their total body weight. Each tag emitted a signal every 10 s and had a battery life of approximately 124 days. A total of 80 nestlings from each of the extended day length and control groups were equipped with tags. We weighed each nestling at the time of tagging and their fat content (scored 0–5, following MoSI protocol [25]) was also determined. At each site, we set up a Motus receiver within approximately 8–70 m to the purple martin houses. Each receiver tower included one omnidirectional antenna with an approximate detection range of 500 m and two nine-element Yagi antennas with a range of approximately 15 km [14].

We determined fledge date for each nestling based on variation in the signal strength of the nanotags and observations from frequent nest checks. Signal strength varied little while nestlings were within their cavities at a constant distance from the receiver but began to fluctuate widely at the time of fledging when the distance between tag and receiver shifted rapidly with first flights from the nest and continued to vary more widely thereafter (electronic supplementary material, figure SF2). We conducted nest checks every other day at each colony and could confirm from visual observations of the nest that young in a nest had fledged. To determine the timing of colony departure, we used the presence of a fading signal and subsequent end of detection of individual tag signals to indicate the day that a bird had departed (electronic supplementary material, figure SF2).

(a) . Data analysis

We used linear mixed-effects models fit by REML from the package lme4 [26] to assess the influence of experimentally extended day length on the timing of fledge date, nesting period, duration at the colony post-fledge and colony departure date. The fixed variables were first egg date, number of nest-mates in each cavity and treatment (experimentally extended day length). The first egg date factor could account for the photoperiod that nestlings experienced as well as other within-season variations, such as the amount of available food. In addition to these fixed variables, weight and fat score at the date of tagging the birds were also considered in the models of nesting period and fledge date. Nested random effects of cavity ID and colony site were set to control for their variations such as potential variability associated with the level of cavities and sites through the season. However, likelihood ratio tests showed the random effect of site was not significant for our models (χ² = 0, p = 1); therefore for parsimony, the random effect of site was removed from the analysis. The interactions between treatment and first egg date and treatment and weight were examined in the models, but preliminary analysis showed these did not have significant impacts and therefore they were removed from the global model. The normality of residual distribution of the models was tested and any statistical outliers on the basis of Cook's D were removed from the dataset. Akaike information criteria corrected for small sample size (AICc) [27] using the package ‘‘MuMIn’ [28] was used to run all possible candidate models that could be built from the full model. Among all competitive models with ΔAICc less than 2, we used the Akaike weight (w) to select the most parsimonious model [27]. All analyses were conducted in R v. 3.6.1 [29].

3. Results

Among the tagged nestlings, we could track the fledge date of 51 individuals from the control group and 68 from the treatment (extended day length) group. Of these, we determined 30 colony departure dates for control birds and 49 departure dates for extended day length birds.

(a) . Nesting period and fledge date

The average duration of the nesting period for control and extended day length groups was 28 ± 0.00 and 29.69 ± 0.22, respectively (figure 1). Nestlings in the extended day length group had a nesting period that was 1.05 ± 0.36 (95% CI, −1.76 to −0.33) days longer than birds in the control group. The number of nestlings also impacted duration in the nest, where one more nest-mate resulted in 0.37 ± 0.15 days (95% CI, 0.06 to 0.68) longer in the nest (table 1).

Figure 1.

Duration in the nest for young in the treatment group that experienced extended day length as compared to young in the control group. Boxes extend to upper and lower quartiles; the line indicates the median and the black point at the middle of the boxes indicates mean. Whiskers extend to maximum and minimum values; outliers are indicated by filled points. (Online version in colour.)

Table 1.

Top linear mixed-effects factors that explain variation in the timing of fledge date and colony departure date of juveniles, as well as nesting duration and duration at the colony post-fledge. The variables we tested include the treatment, first egg date and the number of nest-mates. The global model of fledge date and nesting period also included an individual fat score and weight at the time of radio-tagging.

model		estimate ± se.	95% CI (lower)	95% CI (upper)	AICc	W
fledge date	fixed effects
	treatment (control)*	−1.37 ± 0.66	−2.66	−0.07
	first egg date*	0.97 ± 0.07	0.82	1.12	453.1	0.33
	nest-mate numbers	0.44 ± 0.25	−0.05	0.95
	random effect	variance	s.d.	%variance
	cavity ID	3.03	1.74	70.54
duration in the nest (days)	fixed effects
	treatment (control)*	−1.05 ± 0.36	−1.76	−0.33
	nest-mates numbers*	0.37 ± 0.15	0.06	0.68	409.2	0.57
	random effect	variance	s.d.	%variance
	cavity ID	0.65	0.80	34.20
colony departure date	fixed effects
	treatment (control)*	−2.38 ± 0.88	−4.10	−0.65
	first egg date*	0.90 ± 0.09	0.70	1.09	324.2	0.72
	random effect	variance	s.d.	%variance
	cavity ID	3.92	1.98	66.91
duration at the colony (days)	fixed effects
	first egg date	−0.12 ± 0.07	−0.27	0.02
	random effect	variance	s.d.	%variance	344.5	0.38
	cavity ID	1.72	1.31	36.69

The fledge date of all juveniles was between 12 July and 1 August. The mean date of fledging for juveniles in the extended day length group was 23 July (204 ± 0.53), while birds in the control group fledged on average on 21 July (202.96 ± 0.64). The experimental, extended day length treatment had an impact on fledge date with a 1.37 ± 0.66 day (95% CI, −2.66 to −0.07) delay as compared to birds in the control group (table 1). First egg date had the largest influence on fledge date as expected, where for every 1-day delay in first egg date, fledge date was 0.97 ± 0.07 days (95% CI, 0.82 to 1.12) later (figure 2a and table 1).

Figure 2.

The influence of experimentally extended photoperiod on fledge and colony departure dates: (a) shows the correlation between first egg date and fledge date and (b) shows the correlation between first egg date and colony departure date. In (a,b), each point represents individual birds that are independent and from different nest cavities. (Online version in colour.)

(b) . Duration at colony post-fledge and departure date

The juveniles in the experimental, extended day length group spent more time at the colony post-fledge than those that experienced natural photoperiod only (average of 4 ± 0.00 versus 5 ± 0.33). However, the best model only included the first egg date as a predictor variable of colony duration (table 1 and figure 3). All tagged birds departed the colony between 16 July and 9 August. The mean departure date for juveniles in the control group was 28 July (208.8 ± 0.92) and 29 July (209.95 ± 0.51) for birds in the extended day length group. As expected, the first egg date had an important and positive relationship with colony departure date (0.90 ± 0.09, 95% CI, 0.70 to 1.09) (table 1, figure 2b). However, juveniles exposed to an experimentally extended day length in the nest-box during development had colony departure dates that were 2.38 ± 0.88 days later (95% CI, −4.10 to −0.65) than juveniles that were exposed to natural day lengths only (table 1 and figure 2b). This relationship changed over the colony departure period, where birds with a later departure date were more similar to controls (figure 2b).

Figure 3.

Duration at the colony for young in the treatment group that experienced extended day length as compared to young in the control group. Boxes extend to upper and lower quartiles; the line indicates the median and the black point at the middle of the boxes indicates mean. Whiskers extend to maximum and minimum values; outliers are indicated by filled points. (Online version in colour.)

4. Discussion

Our study experimentally investigated the potential ontogenetic role of photoperiod in the development of post-breeding movement timing of free-living, long-distance migratory songbirds (electronic supplementary material, figure SF3). We show that exposing nestlings to an extended (earlier) photoperiod in a wild environment resulted in a delay in fledge date and colony departure date as compared to birds that experienced natural photoperiod (table 1). These results demonstrate phenotypic plasticity in the timing of the post-fledge movements of nestling songbirds in response to day length experienced in the nest.

Our results are consistent with, and complement, some prior laboratory-based experiments investigating the role of photoperiod in seasonal timing [30–32]. For example, Coppack et al. [11] showed prolongation of moulting and a delay in migration timing (measured through the proxy of Zugunruhe) of European blackcap juveniles in response to a laboratory-based simulation of an earlier photoperiod during nestling development. Our field-based experiment demonstrates that the extension of photoperiod during nestling development carries over to have an impact on colony departure decisions in free-living birds, directly tracked during actual movements. Even a small shift in egg-laying date and consequently fledge date, particularly at more northern latitudes [33] can potentially have carryover effects on the survival of young of aerial insectivores [34] and their subsequent life cycle events [10].

Our results are also consistent with, and help to explain, inferences from field-based, correlational studies of Arctic terns (Sterna paradisea) and pied flycatchers, where breeding date influenced the timing of subsequent migration [10,35]. For pied flycatchers, advances in the recovery dates of banded birds during spring migration were attributed to an ontogenetic effect of advancing breeding dates at more northern breeding latitudes in the previous season [10]. In another study of pied flycatchers, natal nest timing carried over to influence arrival date and breeding in the subsequent year, for three of five years examined [12]. Similarly, advancing breeding dates in one season correlated with advanced arrival and breeding dates in the subsequent season in Arctic terns. These were attributed to a plastic (learning) effect of timing that carried over from the previous year [35]. Our experimental results align with these observations and illustrate that photoperiod during development may be an important mechanism underlying these advances in timing. Taken together, these results suggest that the timing of seasonal behaviours is generally sensitive to the day length experienced while birds are developing within the nest, but that variation may be introduced or timing can be constrained by other endogenous or exogenous factors [10,12,36]. Moreover, the manipulation of light in our experiment could be considered to mimic the day length (summer solstice) of a more northern latitude for the period of the experimental treatment, where the later timing of birds in our experiment may reflect a natural adjustment to breeding at different latitudes. According to a study by Both [10], variation in the timing of pied flycatchers originating from more northern in comparison with more southern nests was attributed to the experience of different photoperiod regimes at different latitudes [10].

Our results also indicate further flexibility in timing post-fledge, in that the experimental treatment had a differential impact on early versus later hatching birds. The steep slope of the correlation between colony departure date and first egg date of both groups showed that young of the treatment group from later nests had earlier departure dates relative to their hatching dates (figure 2b). Seasonally decreasing day lengths post-fledge can have the influence of speeding up the development of later-hatched young; a so-called ‘calendar effect’ [37], which may be an adaptation so that later hatching birds can prepare for autumn migration (e.g. [38–41]). The later hatching birds in our experimental group would have experienced the largest shift in day length between the experimentally extended day length within the nest-box (held to summer solstice) and the shorter, more rapidly decreasing natural day length they experienced once they fledged from their nest-boxes. We infer that the earlier departure of these later hatching birds relative to their hatch date may correspond to a stronger ‘calendar’ effect induced by this larger shift in photoperiod experienced by these individuals, as compared to birds in the control group. Photoperiodic cues indicating that birds may be ‘late’ may induce a stronger migratory response. Long-tailed tits (Aegithalos c. caudatus) intercepted during migration and exposed in the laboratory to a photoperiod simulating one month later had a stronger migratory response (measured via Zugunruhe) than birds exposed to natural day lengths [42].

It is not known whether the impacts of photoperiod on timing that we measured would impact timing over the rest of the calendar year, or further into adulthood. Or, if the ontogenetic effect of photoperiod experienced during nesting is swamped by other intervening influences. This is important, as a phenotypic advancement in the timing of nesting in response to climate change could translate to rapid adjustment to climate change effects if the timing is carried into later life stages [10]. In pied flycatchers, a one-week experimental delay in hatch date led to the ontogenetic effect of later spring arrival and egg-laying dates only for the first subsequent year but did not continue to influence timing in the subsequent 1–2 years [12]. Similarly, in a different study of pied flycatchers, there was no influence of hatch date on the timing of birds after the juvenile stage [43]. The fact that our experimental birds that fledged later tended to be closer in colony departure timing to controls suggests further plasticity after the experimental treatment to ambient light conditions. However, future investigation of how long the influence of photoperiod in the nest predicts the timing of adult birds is required.

Further, we found that brood size impacted the duration of the nesting period, where larger broods resulted in longer nesting time. It has been demonstrated that greater brood size results in greater competition for food in purple martins, therefore prolonging the time required to reach the appropriate condition for fledging [44].

Assisted evolution approaches are being applied in coral reef systems, where corals are pre-adapted to current and predicted increases in ocean temperature [45]. Assisted evolution approaches that address mismatches between environmental phenology and migratory bird timing may be desirable [46], in response to the precipitous declines in the North American avifauna in recent decades [47]. Our results suggest that a manipulation of day length (simulating an earlier calendar date) from hatch to fledge can delay the departure date of nestlings and sheds some light on how an assisted evolution approach could potentially be used to shift the timing of migration based on manipulations in the nest. However, once exposed to a natural photoperiod after fledging, it appeared that birds in the experimental group continued to shift, suggesting that timing may not be fixed and will continue to change post-fledge. Future studies could focus on how long timing shaped in the nest may carry post-autumn migration and/or whether there is a period after which the timing routines of young birds become more ‘fixed’. Such studies could help to further reveal whether an assisted evolution approach to phenological mismatch, where birds raised in captivity and released in the wild are ‘instilled’ with more adaptive timing, may be a viable method.

5. Conclusion

Our study demonstrates the ontogenetic effect of day length experienced in the nest on the subsequent post-breeding movement timing of juvenile, migratory songbirds in the wild. Climate change is rapidly altering environmental phenology resulting in earlier springs and correspondingly earlier nest dates [10] which can expose nestling migratory birds to different photoperiods. Our results demonstrate that the manipulation of photoperiod experienced by nestlings in the wild influences the subsequent timing of their movements. Future research could further investigate the efficacy of similar manipulations as part of an assisted evolution approach to timing mismatches in wild songbirds.

Ethics

All data collection procedures and experiments were conducted in accordance with the guidelines of the University of Manitoba's Animal Care Committee who have approved this project (Animal Care Protocol no. F18-031/1(AC11388)).

Data accessibility

The data and codes are provided as the electronic supplementary material, [48].

Authors' Contributions

S.BA.: data curation, formal analysis, investigation, methodology, software, validation, visualization, writing-original draft, writing-review and editing; K.C.F.: funding acquisition, methodology, project administration, resources, supervision, writing-review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

Funding was provided by a John R. Evans Leaders Fund from the Canadian Foundation for Innovation, Research Manitoba, the Natural Sciences and Engineering Research Council's Discovery Grant Program and the University of Manitoba.