Migratory naïve songbirds rely on an endogenous migration program guiding them from the sites where they are born to their wintering areas (Gwinner, 1975, 1996; Berthold, 1996). The genetic program encode time, distance, direction, and fueling in preparation of migration flights that enable birds to follow population-specific routes (e.g., Berthold and Querner, 1981; Helbig et al., 1989; Berthold, 1996; Helbig, 1996, 2003; Pulido, 2007; Liedvogel et al., 2011). Additional phenotypical adaptations are needed to enable successful long-distance migrations, including morphology, physiology, and behavior (Åkesson and Hedenström, 2007). Several of those adaptations are tightly connected to the endogenous migration program (Gwinner, 1975, 1996; Åkesson et al., 2017) and the ability to navigate (e.g., Emlen, 1975; Able, 1980; Åkesson et al., 2014), while others are related to disposition of time between fueling and flight (Hedenström and Alerstam, 1997).

Both timing and length of the autumn migratory restlessness period have been shown to be correlated with migratory distance in songbirds (Berthold, 1973) and contribute to the individual variation and flexibility in migratory traits observed in birds (Åkesson and Helm, 2020). The periods of controlled migratory restlessness in individual birds are regulated by two environmental cycles, extending over a day or a full year (Aschoff, 1960; Gwinner, 1975, 1996; Gwinner and Brandstätter, 2001). The annual migratory restlessness period is controlled by an endogenous oscillator with external diel input, which in absence of seasonal variations has been shown to free run for several years (Gwinner, 1986). The circadian oscillator is dependent on control mechanisms from the day and night cycle, including variations of light levels (e.g., Aschoff, 1960; Daan, 1976; Gwinner, 1996). The complex circadian pace-making system in birds include three main components (i.e., encephalic photoreceptors, pineal gland, and retina), shown to enable coordination of time over annual and circadian cycles (Gwinner and Brandstätter, 2001).

The migration process involves periods of fueling and flight, which has been predicted to approach a fueling: flight ratio of 7:1 (Hedenström and Alerstam, 1997). Endogenous control of fueling rate and mass increase have been shown to be timed and triggered by geomagnetic information to increase in front of predicted barriers (Fransson et al., 2001), and to be regulated by daylength (Kvist and Lindström, 2000). Migratory restlessness and compass orientation have extensively been investigated in caged songbirds (e.g., Gwinner, 1975, 1986; Åkesson, 1994; Bäckman et al., 1997; Muheim et al., 2006; Helm et al., 2009; Ilieva et al., 2018). However, the combination of recording fueling, and migratory restlessness in the same experimental setup, has so far been limited (Bulte et al., 2017; Ilieva et al., 2018) as well as studies of how daylength affects migration behavior including timing of departure, activity pattern, and fueling (Kvist and Lindström, 2000).

To investigate if diel schedule has any effect on the timing and level of migratory activity, and fueling in juvenile songbirds, we designed an experiment where we experimentally increased the daylength by 2 h in the morning (i.e., prolonged photoperiod shifted forward; Müller et al., 2018b) and kept the geomagnetic parameters at the local site. By this procedure, we could separate the anticipated responses relative to the diel cycle from the geomagnetic cues (i.e., field intensity, angle of inclination) providing latitudinal information (Skiles, 1985). This was important since geomagnetic information have been shown to affect both migratory activity (Ilieva et al., 2018) and fueling (Fransson et al., 2001; Boström et al., 2010) in songbirds. We compared the behavior of the experimental group with control birds kept in the local photoperiod of southern Sweden in autumn. We studied two species of temperate zone passerine migrants captured in autumn on passage to their wintering areas in southwestern Europe, the dunnock, Prunella modularis and the European robin, Erithacus rubecula (Zink, 1975) (hereafter robin). Dunnocks migrate primarily at daytime (Dorka, 1966; Ilieva et al., 2018; Michalik et al., 2020), while the European robin is a strictly nocturnal migrant (Cramp, 1988; Åkesson et al., 2015). The daylength manipulation was substantial (2 h), and therefore out of range from what the species’ normally meet on a daily basis during natural migrations, and in a direction that would be expected during spring rather than during autumn. However, since the period of the increase was abrupt and with a fixed duration (not gradual) it is unlikely that birds will interpret this change as springtime. The length of the light manipulation was selected because we wanted to investigate how the birds reacted to an earlier onset of the day and at the same time reliably measure the time needed for adjusting the clock to a new time. We were interested to investigate if the birds would shift the onset of activity and if the length of the period of migratory restlessness would stay unaffected or change in response to the manipulation. The increase in photoperiod by 2 h correspond to a 1 month shift back in time, or a geographical shift to the south of the expected wintering area in late autumn, where the experimentally increased daylength would occur (Supplementary Figure 1). We expected the two migrants may respond differently to the treatment, since they predominantly migrate at day (dunnock) and at night (robin). The advanced and increased photoperiod was used to investigate if the dunnock responded to the increased daylength by shifting the (fixed) period of migration restlessness in relation to the new sunrise time, or if the sunrise may result in an extended period of expressed migratory activity. For the nocturnally migrating robin we instead expected the earlier sunrise, and thus increased day-length, not to affect the start of migratory restlessness, but rather result in a shorter period of nocturnal restlessness (or no change in length) and an increased fueling rate in response to the increased period available for foraging (Kvist and Lindström, 2000).

We set up the following predictions:

We used two species of temperate zone passerine migrants, the dunnock and the European robin, breeding in large parts of Sweden (Fransson, and Hall-Karlsson, 2008), where different populations are exposed to largely different photoperiods in summer depending on breeding latitude. Dunnock is a diurnal migrant (Dorka, 1966; Michalik et al., 2020) passing through southern Sweden on autumn migration in September and October. Swedish dunnocks migrate to wintering areas in southern France (Zink, 1975; Fransson, and Hall-Karlsson, 2008). The European robin is a nocturnal migrant, captured in large numbers on autumn migration passage in southern Sweden, migrating to wintering areas from Germany and Denmark in the north to North Africa in the south (Zink, 1975; Fransson, and Hall-Karlsson, 2008).

First-year migratory European robins (n = 24) and dunnocks (n = 24) were captured with mist-nets at stopover sites near Stensoffa Ecological Field Station (55°41′N 13°26′E) in southern Sweden on 23–24th September and 6–10th October 2018, respectively. We kept the birds indoors in individual cages for a few days (maximum 5 days) until they were all simultaneously moved to the experimental facility (on 26th September for robins and 11th October for dunnocks) when the experiments started. In the individual cages, the birds were provided with food and water ad libitum.

The experiments were performed in the experimental facilities at the Lund University Stensoffa Ecological Field Station in southwest Sweden (Ilieva et al., 2016). The facility is composed of six wooden houses built in non-magnetic material placed a 5–10 m apart. Each house accommodates four birds in separate circular cages (n = 24 birds in total). For each species, we used three of the houses as control and three houses for altering the photoperiod. All houses have a semi-transparent roof that provided natural light conditions (transmission spectrum visible in Supplementary Figure 2). The daylength was altered adding a LED light source (Lumak Pro; 8,000 lm luminous flux) with daylight color temperature (emitting spectrum in Supplementary Figure 2) in each house that was automatically switched on and off by an electronic timer. The light source was mounted pointing toward the translucent plastic roof of the house, hence, providing indirect illumination to the cages with similar diffuse light as the incoming natural light during daytime. The LED lights were switched on at 07:00 (local sunrise time) in the control houses and 2 h earlier in the treatment houses (05:00) resulting in 2 h longer days for the experimental birds. Light irradiance provided by the lamps was measured on top of the cages with an electronic radiometer (IL 1400A, International Light Technologies, Inc., United States) to be around 0.3 mW/cm² when complete dark outside and comparable with the light intensity measured soon after sunrise in the houses. The lights were automatically switched off in all houses at 10:00 (local time), when the natural light irradiance was above 4 mW/cm² (measured under different weather conditions) and the contribution due to the LED lights to the internal luminosity of the houses was not measurable anymore. We introduced the birds into the cages in the early afternoon to familiarize with the new environment and kept them in the cages during 14 days in total.

The cages were designed to electronically measure the weight of the individual birds while sitting on the perch positioned in the middle of the cage (for more details on setup see, Ilieva et al., 2016), and to let an operator provide food and water from the bottom of the cage. With these procedures, the captive birds were not able to see the operator during the weighting and while providing water and food (once a day at 12:00 local time) reducing potential stress-related effects on the birds.

In each house, a network camera was filming all four cages from above and continuously recording videos on external storing devices. Videos were then used to extract the detailed birds’ activity using the computer vision procedure detailed in Ilieva et al. (2018). Briefly, the algorithm automatically detected and tracked the motion of each bird individually and measured the fraction of time that the birds were in flying mode (i.e., excluding any fluttering, jumping or walking) in 20-min intervals. Raw data from the current experiments and extracted by the above procedure are available as Supplementary Appendix 1, and summarized in the actogram in Figure 1. The daily activity recorded for the individual bird was measured as the average number of 20 min active intervals (i.e., intervals when the bird was flying for more than 1 min; Ilieva et al., 2018), and presented for the full experimental day (from 12:00 to 12:00 of the next day, local time), but also divided into first half of the day (from 00:00 to 12:00) and second half of the day (from 12:00 to 00:00) periods (Figure 2). In addition, we performed statistical tests in intervals before the sunrise time (either at natural or experimentally manipulated time) to investigate whether nocturnal activity in European robins was affected by the expected sunrise time. We finally tested both species for difference in “departure” and “arrival” times calculated relative to the overall diurnal activity for dunnocks (between midnight and feeding time: 00:00–12:00, local time) and nocturnal activity for robins (between sunset and sunrise: 19:00–07:00). The departure (i.e., start of migratory activity) was defined as the time at which each individual reached the 10th percentile of the cumulative activity in the considered migratory period that has been demonstrated to be an accurate predictor by Schmaljohann et al. (2015). Similarly, the arrival (i.e., end of migratory activity) was defined as the time at which the individual reached the 90th percentile of the cumulative activity, that is, when the bird is close to landing and possibly searching for shelter or a foraging ground.

All the plots and statistical tests were done in the software R version 3.6.1 (R Core Team, 2019). One-way and repeated-measures ANOVA were used for testing differences in activity, food intake, fuel load and arrival/departure time between the treatment and control groups of each species.

The actograms show clear differences in activity patterns between dunnocks and robins (Figure 1), with the highest activity in the first half of the day starting at sunrise for dunnocks, and a prolonged period with elevated level of activity during the entire night for robins. The actograms also show a strong effect of the treatment on the timing of activity in both species.

Dunnocks showed no activity at night and initiated their migratory activity at sunrise (Figure 1). Thereafter, the activity continued for several hours, and ended before noon. The start of activity coincided with local sunrise time for the control group exposed to the local diel cycle and was initiated 2 h before for the experimental group (a time corresponding to the apparent sunrise time in this treatment group; Figure 1). Dunnocks in the local photoperiod were active for the rest of the day and expressed a peak of increased activity just before sunset (Figure 1). The highest level of activity in dunnocks was expressed in the first half of the day, starting at sunrise and ending before noon (Figure 1). The experimental dunnocks expressed a similar temporal pattern of activity throughout the day, but with an extended morning activity period starting at the experimentally advanced sunrise and ending at the normal time before noon. The morning activity (higher) is primarily associated with migration, while afternoon activity (lower) is associated with foraging in dunnocks.

Robins expressed a short peak of activity just before sunset, which for this nocturnal migratory species is expected to be associated with preparations for the migratory departure. Immediately after this peak of activity at sunset, the robins showed a restricted period of lower activity corresponding to a “quiescence period,” before the elevated nocturnal restlessness period started. Robins were active for the entire night, but with a reduced activity level, and showed an increased activity ending with a peak at sunrise (Figure 1). The timing of the peak of morning activity for the robins, probably associated with the end of the daily migratory activity and search for shelter, was affected by the treatment. That is, for the control group the peak of activity was at local sunrise time (07:00), whereas, for the experimental group the peak of activity occurred almost 2 h earlier around 05:00 when lamps were turned on for the experimental group (Figure 1; see also below). Both robins held in the local and extended photoperiod continued to be active during the mornings, and showed relatively limited activity during the afternoons (Figure 1).

To investigate whether there were differences in restlessness between groups, we analyzed the duration of activity (i.e., number of active intervals), and we found that dunnocks exposed to experimentally increased photoperiod were more active than the control group (Figure 2). The difference was significant during mornings (repeated measures ANOVA, days 1–13: F_1,22 = 4.394, p = 0.048), but not during evenings (F_1,22 = 2.239, p = 0.149). The higher activity during the morning, is a consequence of the fact that dunnocks in the daylength treatment group started their migratory activity earlier (see above) but continued to be active until around 10:00 which is similar to the time when migratory activity ceased in the control group (Figure 1; see also below).

Robins, on the other hand, were not affected in their overall activity by the increased photoperiod with earlier sunrise and did not change activity pattern (p > 0.486 for all time periods; Figure 2). However, we investigated if the peak activity at sunrise for robins (see above) was only affected by an external factor (i.e., the earlier sunrise of the treatment) or by an internal factor (i.e., birds can predict time of sunrise with their circadian clock). We compared the activity of the control group with the experimental group during 5 h before the sunrise, either at natural sunrise timing (07:00) or at the artificial sunrise time (05:00). The activity of robins 5 h before the natural sunrise (02:00–07:00) was not affected by treatment (repeated measures ANOVA, days 1–13: F_1,22 = 1.515, p = 0.231). However, the activity 5 h before the artificial sunrise (00:00–05:00) showed that the experimental group was significantly more active than the control group (F_1,22 = 8.887, p = 0.007), with higher level of activity for the experimental group. The latter difference was measured before any light was yet present in the houses. We did not find this response in the dunnocks relative to the earlier sunrise, as they expressed no activity at night (Figure 1).

As predicted by the actogram in Figure 1, we found a strong effect on time for onset of migratory activity between the control and the experimental group in dunnocks (repeated measures ANOVA, days 1–13: F_1,22 = 106.4, p < 0.001) with the experimental group starting to express migratory restlessness earlier following the artificial sunrise (Figure 3). We did, however, not find any time difference in the end of restlessness for the dunnocks (F_1,22 = 2.04, p = 0.167). Furthermore, we did not find any difference in start of migration activity between the control and the experimental group in robins (F_1,22 = 0.005, p = 0.945; Figure 3), but we did find that the end of activity for the experimental group was consistently earlier than the control group (F_1,22 = 47.33, p < 0.001), and this difference lasted during the entire experimental period (Figure 3).

We also measured how body mass, foraging activity and fueling varied across the experimental period, as well as between the experimental groups for the two species. There were no initial body mass differences between the control and experimental groups for both dunnocks (control: mean ± SD: 21.8 ± 3.0 g; treatment: 21.4 ± 1.6 g; one-way ANOVA: F_1,22 = 0.139, p = 0.713) and robins (control: 16.9 ± 0.9 g; treatment: 17.0 ± 1.3 g; F_1,22 = 0.096, p = 0.759), nor was there a significant difference between groups in food intake throughout the experiment (repeated measures ANOVA, days 1–13: dunnocks: F_1,22 = 2.488, p = 0.129; robins: F_1,22 = 0.064, p = 0.802; Figure 4). Both species showed a tendency to increase body mass during the experimental period (Figure 4). However, the fuel load (i.e., fraction of body mass increase relative to individual’s lean body mass) for dunnocks, was significantly lower for the daylength treatment group as compared to the control group (repeated measures ANOVA, days 1–13: F_1,22 = 5.806, p = 0.025; Figure 4), but there was no difference in fuel load between the two groups for robins (F_1,22 = 0.029, p = 0.866; Figure 4).

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

The animal study was reviewed and approved by Malmö/Lund Ethical Committee for scientific work on animals (Dnr 5.8.18-12719/2017), the Swedish Board of Agriculture for housing facilities (Dnr 5.2.18-5398/16), and work with animals (Dnr 5.2.18-10992/18).

MI performed the fieldwork with support from GB and SÅ. GB analyzed the data. All authors conceived the study, discussed the outcome of the results, and wrote the first draft of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.