Short photoperiods end autumn migration in a naïve diurnal migrant

Many migratory animals use daylength, or photoperiod, to signal when to migrate and transition be-tween annual phenological states. Whether animals use photoperiod as a temporal or spatial cue while migrating, however, requires additional empirical support. We used hatch-year dunnocks, Prunella modularis (a songbird), caught during their ﬁ rst migration in southern Sweden to elucidate whether migratory animals incorporate photoperiod as a spatiotemporal cue into their endogenous migratory program during migration. We exposed the migratory-naïve to light environments that simulated either the local photic conditions or a shorter daylength and larger transitions between photoperiods. All birds experienced local geomagnetic conditions. We hypothesized that migratory dunnocks used photoperiod to inform their ﬁ rst migration and predicted that the experimental treatment represented either a spatial displacement to the north or a temporal advancement towards winter at the capture site compared to the local control conditions. We found, though, that the short photoperiods terminated the expression of the migratory phenotype compared to controls by reducing body mass gain and ending migratory activity, indicating that the endogenous migratory program integrates photoperiod during migration. The incorporation of photoperiod into the endogenous program may complement geomagnetic cues to ensure ending migration at the correct time and location. The incorporation of photoperiod can also provide a mechanism that facilitates poleward shifts of overwintering distribution under climate change by allowing migrants to overwinter in newly suitable habitat at higher latitude (i.e. short stopping). © 2024 The Author

Animals, from insects to birds, use photoperiod as a cue for transitioning between nonmigratory and migratory phenotypes and scheduling their annual cycle.For example, migratory bluethroats, Luscinia svecica (a songbird), started depositing fat as a migratory fuel quicker when exposed to shorter photoperiods simulating the advancement of autumn compared to natural photic conditions (Lindstr€ om et al., 1994), while European stonechats, Saxicola torquata rubicola (a songbird), that experienced a slower progression in photoperiods during vernal migration delayed annual transitions in physiology and behaviour, such as moult and the following autumn migration (Helm & Gwinner, 2005;Helm & Liedvogel, 2024).Migratory monarch butterflies, Danaus plexippus, and milkweed bugs, Oncopeltus fasciatus, initiate migration and transition to a migratory phenotype when exposed to the short photoperiods of autumn (Saunders, 2009).Additionally, although mammals may schedule migrations based on foraging opportunities (Sawyer & Kauffman, 2011), the annual moulting and growth of their pelage and sexual reproduction can be organized in line with photoperiod (Bradshaw & Holzapfel, 2010;Webster & Barrell, 1985), and therefore photoperiod may initiate the pelage and reproductive status conducive to migration in mammals.Despite much support for the role of photoperiod in transitioning to a migratory phenotype and initiating migration, evidence remains elusive on whether migratory animals use photoperiod during migration and incorporate photoperiod in their endogenous migratory program during their first migration.
The endogenous migratory program in animals can control the duration, direction and necessary physiology in naïve migrants (Åkesson & Helm, 2020;Helm & Liedvogel, 2024).This endogenous program controls the expression of activity attributed to migration in diurnal and nocturnal migrants (Gwinner, 1986(Gwinner, , 1996;;Stey et al., 2017), while the change in physiology includes migratory fuelling.Migratory fuelling is the accumulation of fat which substantially increases body mass during anticipated annual phases of migration (Farner et al., 1961;Fransson et al., 2001;Hedenstr€ om & Alerstam, 1997;Lindstr€ om et al., 1994).The endogenous migratory program can also incorporate a circannual rhythm that controls whether the autumnal and vernal migrations occur at the correct phase of the annual cycle and proceed for the appropriate duration and in the appropriate compass direction (Åkesson & Helm, 2020;Gwinner, 1986;Gwinner & Wiltschko, 1978).Additionally, photoperiodic cues used for initiating migration and magnetic cues that guide birds towards the correct destination are both well documented components of the migratory program (Åkesson & Helm, 2020;Beck & Wiltschko, 1981;Coppack & Pulido, 2004).For songbirds during autumn, magnetic displacement south to their overwintering grounds can attenuate the migratory phenotype in naïve individuals despite a photoperiod either of 12 h or of the capture site (Bulte et al., 2017;Ilieva et al., 2018;Kullberg et al., 2007), while photoperiod can trigger body mass gain and migratory activity in migratory birds (Coppack & Pulido, 2004;Dawson et al., 2001;Farner et al., 1961;Gwinner, 1986;Lindstr€ om et al., 1994).
However, few investigations elucidate whether the endogenous migratory program incorporates photoperiod as either a spatial or a temporal cue during the first migration.Indeed, most investigations of the endogenous migratory program hold photoperiod constant to reveal the endogenous expression of the migratory phenotype (Gwinner, 1986(Gwinner, , 1996;;Gwinner & Wiltschko, 1978;Helm, 2020;Stey et al., 2017) or identify specific photoperiods for the initiation of this phenotype (Coppack & Pulido, 2004;Lindstr€ om et al., 1994).One of the rare examples investigating photoperiod during the first migration revealed that the daily average of locomotor activity of the long-tailed tit, Aegithalos caudatus (a songbird), was photosensitive (Bojarinova & Babushkina, 2015), yet these results did not identify whether the photosensitivity was associated with a change in migratory phenotype.
We used a controlled experiment, advanced computer vision and captive dunnocks, Prunella modularis (a songbird), that migrate during daylight to test whether photoperiod conveys spatiotemporal information as part of the endogenous migratory program.Diurnal migrants are suitable subjects because they exhibit migratory restlessness (Stey et al., 2017) and because typical migratory cues, such as the geomagnetic field, affect their daily activity during the migratory season, indicating that the activity should be migratory (Ilieva et al., 2018).We subjected a treatment group of hatch-year dunnocks to a photic environment representing a northward geographical displacement to above the polar circle.This photic environment included large transitions between increasingly shorter short photoperiods, which is characteristic at polar latitude during autumn.The birds could interpret this photic environment as a spatial displacement, because of the combination of the magnitude of transition between, and the short duration of, photoperiods.Alternatively, they could interpret this photic environment as a temporal advancement towards winter at the capture site, because of the combination of short photoperiods and the local geomagnetic field.We subjected another control group to a photic environment representing the location of capture.We predicted that the birds in the treatment group would increase migration-associated activity and body mass compared to the control group because the experimental photic environment conveys a position that is either temporally delayed at the capture site or spatially displaced poleward, both situations that should increase motivation to migrate to their overwintering grounds.Supporting this prediction, the long-tailed tits that experienced a photic advancement of autumn increased their daily locomotor activity (Bojarinova & Babushkina, 2015).In addition, naïve migratory songbirds captured later in autumn increased migratory fuelling because of a perceived time pressure to complete migration (Kullberg et al., 2003(Kullberg et al., , 2007)), while naïve northern wheatears, Oenanthe oenanthe (a songbird), magnetically displaced poleward increased fuelling during autumn migration compared to controls (Bostr€ om et al., 2012).We could alternatively predict that the migratory phenotype would end if short photoperiods, independent of the transition between photoperiods and the geomagnetic field, were the most important cue.This outcome would suggest that the short photoperiods experienced by the treatment group represented winter on the overwintering grounds.Some support for this alternative prediction is that the investigations on the longtailed tit and European stonechat demonstrated an effect of a manipulated photoperiod while not manipulating the geomagnetic field (Bojarinova & Babushkina, 2015;Helm & Gwinner, 2005).Consequently, our protocol should reveal whether dunnocks incorporate photoperiodic cues as spatiotemporal information during their first migration.

Study Species
We collected 23 hatch-year dunnocks that had never completed a migration in southern Sweden on 30 September 2022 near Stensoffa Ecological Field Station (55.68 N, 13.43 E).S.Å., an experienced ringer, identified the age of the dunnocks following Svensson (1992).Dunnocks' breeding distribution extends from eastern Europe and southern Sweden to above the polar circle into northern Norway and Russia, while they can overwinter across central and southern Europe (Hatchwell, 2020).We could not identify the origin of captured birds, but capture occurred during the known migratory season for the region; indeed, most dunnocks ringed in Sweden have left Sweden by October (Fransson & Hall-Karlsson, 2008).Dunnocks that pass through southern Sweden generally migrate southwest to southern France to overwinter (Fransson & Hall-Karlsson, 2008;Zink, 1975), and the southwest migratory direction is consistent for hatch-year birds during their first autumn migration (Fransson & Hall-Karlsson, 2008).They predominantly schedule migratory activity around sunrise and during the beginning of the day (Åkesson et al., 2021;Dorka, 1966;Ilieva et al., 2018;Michalik et al., 2020).They may partially migrate at night (Ulfstrand et al., 1974), but this is rare (Michalik et al., 2020) and does not occur in experiments recording migratory activity in captive dunnocks (Åkesson et al., 2021;Ilieva et al., 2018;see Results).

Experimental Set-up
We randomly assigned individuals to the treatment (N ¼ 12) and control (N ¼ 11) groups.We conducted the experiments in six nonmagnetic, wooden huts.Each hut contained four circular cages with opaque walls and a transparent mesh top (cage height ¼ 700 mm and diameter ¼ 550 mm), and one bird was in each cage.The cages contained a single circular perch attached directly to a balance underneath the cage.Birds had ad libitum food (mealworms, Tenebrio molitor) and water.We replaced food and water daily at 1300 local time using access ports on the floor of the cage; the birds were unable to see us when we were in the huts, including when changing food and water.We recorded daily each individual's mass to the nearest 0.01 g when changing their food and water.We also weighed the food into and out of the cage to calculate the food consumed to the nearest 0.001 g.A closed-circuit video-surveillance system connected to a local network continuously recorded the four cages in each hut.The huts had nontranslucent roofs, and we used square photoperiods and turned lights on and off using automated switches.We included nightlight in each hut during subjective nighttime (colour temperature ¼ 1500 K, 75 W Night Heat Lamp, Exo-Terra, https://exo-terra.com).Light intensity at the top of the cages was approximately 600 lx during subjective daytime and below the lower threshold of our lux meter (<0.1 lx) during subjective nighttime.
We used the latitude and longitude of the study site and a poleward displacement to 71.15 N to identify the photoperiod schedule used from the beginning (2 October 2022) until the end (14 October 2022) of the experiment in the control and treatment groups, respectively (see Supplementary Material).The latitude of 71.15 N is approximately the northernmost area in continental Norway, an area near the northern limit of where dunnocks can be recorded and their breeding distribution (Hatchwell, 2020).The range of photoperiods experienced was 10.50 to 11.63 h for the control group and 8.88 to 11.05 h for the treatment group.The photoperiod schedule followed the change in photoperiod at 55.58 N and 71.15 N for the control and treatment groups, respectively, using the longitude of the field station, and we obtained the schedule from www.timeanddate.com(last accessed: 17 April 2024).In two huts, the automatic switches malfunctioned (hut 1 ¼ 3.5 days, hut 3 ¼ 1 day), but the results did not change whether we removed the affected data or not.We therefore present results excluding the affected data (see Supplementary Material).An error caused relatively shorter photoperiods than planned for both groups on day 7 of the experiment (see Supplementary Material).All birds experienced 1 day with local photic conditions in the huts before beginning the experiment.

Ethical Note
The experiment adhered to relevant laws and regulations.Permissions were given by the Malm€ o/Lund Ethical Committee for Scientific Work on Animals (Dnr 5.8.18-12719/2017; 5.8.18-09591/ 2021; Sweden), the Swedish Board of Agriculture for housing facilities (Dnr 5.2.18-5398/16; 5.2.18-04121/2019) and work with animals (Dnr 5.2.18-10992/18), the Swedish Nature Protection Agency and the Swedish Ringing Centre (No. 440) for catching and ringing birds.We captured hatch-year dunnocks (N ¼ 23) of unknown sex from the wild using a mist nest and playback.Capture occurred within 500 m of the huts and field station, limiting transport.Before birds entered the huts, they spent 24 h in individual cages indoors and exposed to the natural photoperiod; they had ad libitum food and water during this time (Åkesson et al., 2021;Ilieva et al., 2018).Once transferred to the huts, one bird occupied each of the four cages in each hut (see Experimental setup for the description of cages).Stress could negatively affect the birds' natural behaviour that we were interested in investigating.We therefore strived to reduce all unnecessary sources of stress and disturbance.We collected activity data remotely reducing the need to enter the huts.When we did enter the huts each day to provide fresh food and water and measure body mass, the birds could not see us.We released all birds back into the wild after the experiment.All birds were healthy when released.

Data Collection Using Computer Vision
We automatically assessed the birds' activity through the video footage captured by the network camera in each hut, following Ilieva et al. (2018) and Åkesson et al. (2021).Each camera was positioned under the roof and was framing all four cages (Ilieva et al., 2018).The computer vision methods used for this analysis quantified the proportion of time the birds actively spent in flight, excluding nonflight motions, such as wing whirring, fluttering, feeding or walking, following Ilieva et al. (2018).We focused on flight only because flight is the primary movement used during songbird migration and, unlike most other investigations, we could separate flight from other activities using our computer vision approach.Individual birds were tracked using dynamic background modelling (Zivkovic & van der Heijden, 2006).A fixed threshold was applied to the speed distribution to distinguish between periods of flight and nonflight (Ilieva et al., 2018).The flight/nonflight ratio was then quantified in 72 equal periods per day.The resulting data, representing the proportion of flight time in 20 min intervals, served as an indicator of migratory activity (Åkesson et al., 2021;Ilieva et al., 2018).These results are presented as raw data in the Supplementary material and visually depicted in actograms (see Results and Supplementary Material).

Statistical Analyses
We used R v. 4.3.3 for all analyses (R Core Team, 2024), and the 'lme4' package v. 1.1-35.1 for all linear mixed-effects models (LMMs; Bates et al., 2015).We tested our hypotheses by comparing a model with the relevant predictors and random effects with an intercept only, or null, model that included the relevant random effects using a likelihood ratio test (Tredennick et al., 2021).We identified the significance of the predictors at a ¼ 0.05.Model coefficients (b) were significant if their 95% confidence interval (CI) did not overlap zero.We used base solutions and the R-package 'report' v. 0.5.8 for model summary statistics (Makowski et al., 2023).We visually checked assumptions of linear mixed-effects models using histograms of residuals and the R-package 'performance' v. 0.11.0 (Lüdecke & Ben-Shachar, 2021).We calculated estimated marginal means of model predictions and 95% CIs for plotting using the R-package 'emmeans' v. 1.10.0(Lenth, 2023) and plotted these using 'ggplot2' v. 3.5.0(Wickham, 2016).
We tested the prediction that the birds in the treatment group will increase migratory-associated body mass compared to the control group by modelling the relationship between body mass and an interaction between day (continuous) and group (categorical) and an additive interaction between food consumed (continuous) and group.We controlled for repeated measures and different initial mass of each individual across the experiment by including a random intercept for bird ID (categorical) and a random slope for day, respectively.
We tested the prediction that the birds in the treatment group will increase migratory activity compared to the control group by modelling the relationship between mean daily time spent flying and an interaction between day and group.We controlled for repeated measures attributed to an individual by including a random intercept for bird ID.
We additionally tested whether the birds in the treatment group changed their mean evening flight time 1 h before lights off compared to the control group by modelling the relationship between mean evening flight time and an interaction between day and group.We controlled for repeated measures attributed to an individual by including a random intercept for bird ID.

RESULTS
The experimental photic environment affected birds' migratory phenotype by reducing body mass gain and attenuating migratory activity across the experiment.Birds exposed to the experimental photic environment gained significantly less body mass over the experiment compared to birds in the control photic environment when controlling for individual differences in starting mass (LMM: c 2 5 ¼ 38.04, P < 0.0001, R 2 ¼ 0.92; Fig. 1).The change over time in body mass was significantly different by 0.23 g per day compared to the body mass of birds in the control group, which gained 0.33 g per day (Fig. 1, Table 1).This change in body mass occurred despite no significant influence of the amount of food consumed in each group on body mass (Table 1).
The experimental photic environment also caused birds to become less active over the experiment, while birds experiencing the control photic environment tended to be more active over the experiment (LMM: c 2 3 ¼ 17.94, P ¼ 0.0005, R 2 ¼ 0.56; Figs. 1 and 2).The change over time spent flying in the experimental group was significantly different by À0.11% per day (Table 2) compared to the flying time of birds in the control group (Fig. 1), and birds in the treatment group spent substantially less time flying by the last day of the experiment compared to the control group (treatment group ¼ 3.00%, 95% CI ¼ 2.12 to 3.87%, control group ¼ 2.07%, 95% CI ¼ 1.22 to 2.91%; Fig. 1).The birds in the treatment group also had significantly less evening activity 1 h before lights off compared to birds in the control group (LMM: c 2 3 ¼ 9.71, P ¼ 0.02, R 2 ¼ 0.46; Figs. 1 and 2).However, the significance of this model was caused by the difference between groups and not by the interaction between day and group (Table 3).

DISCUSSION
Short photoperiods caused migratory dunnocks to attenuate their migratory phenotype significantly and to end their migratory activity during their first migration; therefore, photoperiod could provide spatial or temporal cues to migrating birds as part of the endogenous migratory program.Contrary to our main prediction, however, the short photoperiods reduced expression of the migratory phenotype by decreasing migratory activity and body mass gain by the end of the experiment instead of increasing motivation to migrate compared to controls.This contrasts with long-tailed tits which increased daily activity during a simulated advancement of autumn (Bojarinova & Babushkina, 2015).The treatment group also appeared more incapable of predicting lights off during the larger transitions in photoperiod compared to the control group because the evening activity peak, which may or may not be migratory activity (Åkesson et al., 2021;Ilieva et al., 2018;Michalik et al., 2020;Ulfstrand et al., 1974), decreased substantially from day 1 of the experiment.The birds did not generally appear to interpret the polar photic environment as a spatial displacement north, suggesting they interpreted the short photoperiods as being later in the annual cycle at potential overwintering grounds.
We do not know whether these potential overwintering grounds were a temporal advancement at the capture site or at their typical overwintering area in southern Europe.However, dunnocks migrating in early October may never experience such short photoperiods as those experienced towards the end of the experiment until they reach their overwintering grounds.This is because the 8.88 h photoperiod on day 13 of the experiment occurs at approximately 6 November at the capture site, a date by which most dunnocks have left Sweden (Fransson & Hall-Karlsson, 2008;Ulfstrand et al., 1974), while similar photoperiods only occur in December at latitudes where dunnocks ringed in Sweden normally overwinter (approximately 46 N; Fransson & Hall-Karlsson, 2008;Ilieva et al., 2018;Zink, 1975).Additionally, dunnocks captured in southern Sweden during autumn and exposed to the local photoperiod reduced their migratory activity when displaced magnetically to their expected overwintering grounds near southern France (Ilieva et al., 2018).In combination with our results, the evidence suggests that the birds interpreted the short photoperiods as a spatiotemporal displacement to their expected overwintering grounds later in the annual cycle.
Despite numerous investigations on the role of photoperiod in initiating migratory behaviour and physiology (Åkesson & Helm, 2020;Coppack & Pulido, 2004;Dawson et al., 2001;Farner et al., 1961;Gwinner, 1986;Lindstr€ om et al., 1994), few, if any, have demonstrated experimentally whether a migratory animal can use photoperiod during their first migration as either a spatial or a temporal cue (cf.Bojarinova & Babushkina, 2015).The predictability of photoperiod by latitude and date may be a potent cue used by migratory animals not only to initiate migration (Åkesson & Helm, 2020;Farner et al., 1961;Gwinner, 1986), but also to organize ongoing migrations (Figs. 1 and 2).Hatch-year migratory dunnocks incorporated photoperiod as a cue to end their first Black lines and shaded areas are the model predictions and 95% confidence intervals, respectively, from models that explained significant variation in the response variables compared to a null model.For evening flying time, only groups were significantly different and not the interaction between day and group (Table 3).Models controlled for repeated measures of individuals using a random intercept of bird ID, and the body mass model also controlled for possible effects of different initial mass of each individual using a random slope for day.
N. P. Huffeldt et al. / Animal Behaviour 215 (2024) 23e29 migration (Figs. 1 and 2), supporting the integration of photoperiod into the endogenous migratory program during migration, and not just as a cue to initiate the migratory phenotype.
The endogenous migratory program is vital for determining where, when and how far animals migrate (Åkesson & Helm, 2020).When photoperiod is held constant, songbirds typically reduce the migratory phenotype when they reach the magnetic environment of their overwintering grounds (Bulte et al., 2017;Ilieva et al., 2018).Yet, photoperiod may be a more prominent cue for initiating and ending migration because, with short photoperiods, the dunnocks strongly attenuated the migratory phenotype despite experiencing northern geomagnetic conditions (Figs. 1 and 2), while photoperiod is well known to initiate the migratory phenotype (Åkesson & Helm, 2020;Coppack & Pulido, 2004;Dawson et al., 2001;Farner et al., 1961;Gwinner, 1986;Lindstr€ om et al., 1994).Incorporation of photoperiod into the endogenous migratory program during migration, in combination with geomagnetic cues, may ensure birds stop migration at the correct location and date.
The integration of photoperiod into the endogenous migratory program may also facilitate adaptation to the warming climate.The change in photoperiod by latitude and date is not altered by climate change, unlike temperature gradients which are shifting poleward as the climate warms (Bradshaw & Holzapfel, 2010;Coppack & Pulido, 2004;Huffeldt, 2020).Unihemispherical migrants that do not cross the equator should have less rigid annual programs than transequatorial migrants, which should allow them to track suitable habitat to higher latitudes more easily (Huffeldt, 2021).At higher latitudes, photoperiods are shorter earlier in the autumn.Thus, if migrants remain at higher latitudes because they find newly suitable habitat there, then the short photoperiods they experience earlier in the autumn at higher latitudes may facilitate exiting the migratory phenotype.The influence of short photoperiods earlier in autumn at higher latitudes may allow them to remain in the newly suitable habitat until the return to their normal breeding grounds (i.e.short stopping; Elmberg et al., 2014).
The interaction between latitudinal differences in photoperiod and the incorporation of photoperiod into the endogenous migratory program during autumn migration would provide a mechanism for shifting winter distribution in migratory birds, which would be adaptive under climate change (Helm & Liedvogel, 2024).However, an upper latitudinal limit to short stopping may occur because the dunnocks exposed to the polar photic environment with large transitions between photoperiods seemed to predict poorly the time of lights-off compared to the control group that experienced a more temperate transition in photoperiod (Fig. 2).This suggests a mismatch between endogenous timing mechanisms and the photic environment (Huffeldt, 2020).Whether this potentially poor prediction affects the ability to remain at very high latitudes during winter is unknown.Additionally, whether unihemispherical, equatorial and transequatorial migrants can all benefit from the above mechanism which could facilitate short stopping requires additional insight.

Conclusion
The result that the endogenous migratory program of naïve migrants incorporates photoperiod advances our understanding of the environmental cues used to end their first migration, and photic and geomagnetic cues may complement each other to ensure the birds end migration at the correct time and location.Additionally, the integration of photoperiod into the endogenous migratory program during migration may provide a mechanism for migrants to shift winter distributions polewards under climate change via short stopping, demonstrating the importance of understanding how traits and the photic environment may interact (Bradshaw & Holzapfel, 2010;Coppack & Pulido, 2004;Huffeldt, 2020).Whether the incorporation of photoperiod into the migratory program during migration is a ubiquitous trait awaits confirmation, and we encourage further research on how the photic environment and associated traits affect adaptation to anthropogenic change.

Figure 1 .
Figure1.Change in (a) body mass, (b) flying time (migratory activity) and (c) evening flying time of dunnocks during the experiment.Black lines and shaded areas are the model predictions and 95% confidence intervals, respectively, from models that explained significant variation in the response variables compared to a null model.For evening flying time, only groups were significantly different and not the interaction between day and group (Table3).Models controlled for repeated measures of individuals using a random intercept of bird ID, and the body mass model also controlled for possible effects of different initial mass of each individual using a random slope for day.

Figure 2 .
Figure 2. Actogram of migratory activity in hatch-year dunnocks experiencing (a) local (control) and (b) polar (treatment) photic environments during the autumn.Flying time is the mean percentage duration that all individuals in a group spent actively flying in a 20 min interval and does not include nonflight motions, such as wing whirring, fluttering, feeding or walking.

Table 1
Summary of model predicting body mass