Do seabirds dream of artificial lights? Understanding light preferences of Procellariiformes Free

The increase in urbanisation and development of new technologies continue to intensify anthropogenic pressures on ecosystems. The levels of artificial light at night (ALAN) at the Earth's surface, or light pollution, are increasing at estimated rates from 2% to 9.6% a year (Kyba et al., 2017, 2023), seeping into protected and remote habitats (Garrett et al., 2019) and continuously eroding nightscapes in marine and terrestrial realms (Gaston and Sánchez de Miguel, 2022; Marangoni et al., 2022). The pervasive impact is worsened by the recent widespread transition to LED technology, which is leading to increased emission of short-wavelength (blue-rich) light (Sánchez de Miguel et al., 2021, 2022), which produces larger effects on wildlife than other lamp types and wavelengths (Longcore et al., 2018; Longcore, 2023).

Seabirds are among the most threatened avian groups (Dias et al., 2019) and burrow-nesting species (mainly Procellariiformes) are particularly vulnerable to light pollution (Rodríguez et al., 2019). At the end of the breeding season, thousands of burrow-nesting seabird fledglings fall to the ground in urban areas after encountering light-polluted areas during their initial flights from nest to sea (Rodríguez et al., 2017a). Such fallout events have been documented and mitigated for decades through rescue programmes, e.g. in Hawaii (Raine et al., 2017), Canary Islands (Rodríguez et al., 2023), Réunion (Chevillon et al., 2022) and Azores (Fontaine et al., 2011). In these programmes, citizens and specialists search for grounded birds, collect them and, after inspection of the birds for health by trained staff, release them at sea, with the intention of mitigating direct mortality. Nights with a new moon, foggy conditions and strong inland winds increase the magnitude of fallouts (Rodríguez et al., 2017a,b; Syposz et al., 2018; Telfer et al., 1987). Closeness of the source colony to the artificially lit area can lead to a higher amount of grounded birds (Crymble et al., 2020; Rodríguez et al., 2015; Wilhelm et al., 2021), although fledglings can also be redirected to light-polluted areas from substantial distances (Friswold et al., 2023). The amount of light pollution viewed by the fledglings during their flights to the ocean, or after they have already reached the ocean, also seems to influence the number of affected birds (Troy et al., 2013). However, proximate behavioural and biological causes of fallouts remain mostly unresolved, and it remains uncertain how other mechanisms, such as dispersal from the colony, may contribute to the pool of grounded birds (Brown et al., 2023; Rodríguez et al., 2017a).

Behavioural responses of seabirds to light sources are complex, with evidence of an interplay between attraction and avoidance. Fledglings are observed near light sources doing erratic flights, apparently unable to leave the overall lit environment via flight. Fledglings display increased tortuosity in their flights the closer they are to lit areas and the more intense the light pollution levels are (Rodríguez et al., 2022). Apart from fledgling fallouts, there are many documented events detailing adult seabirds being drawn to artificial lights or even crashing into them; for example, in the collision and entrapment records from lighthouses, ships and other lit structures at sea (Montevecchi, 2006; Gjerdrum et al., 2021; Ryan et al., 2021; Coleman et al., 2022). Similarly, adult little penguins, Eudyptula minor, a burrow-nesting sphenisciforme, showed preference for lit paths on land over darker ones and for paths lit with shorter wavelengths (blue light) over longer wavelengths of light (red light), upon returning to their colony (Rodríguez et al., 2018). In contrast, adult Manx shearwaters, Puffinus puffinus, when in flight above the colony, temporarily avoided the area whenever a light was turned on, especially during exposure to blue-rich light (Syposz et al., 2021). Likewise, the sporadic presence of light from ships facing a colony of Yelkouan shearwaters, Puffinus yelkouan, was associated with fewer adults coming to the nests during the lit periods (Austad et al., 2023). The disparity of behaviours recorded suggests that reactions to artificial light may vary because of intrinsic qualities of the bird, such as species, age, body condition, personality, experience or phenology, or the characteristics of artificial light, such as distance to the light source, its intensity, emission spectra, temporal and spatial extent, and contrast to the surrounding. So, to effectively manage the effects of light pollution on seabirds, it is essential to understand their behaviours towards light at different life stages and under different scenarios.

In this study, we used a two-choice experimental Y-maze set-up to investigate behavioural responses to light stimuli at two life stages (fledglings and adults) of a burrow-nesting seabird, Cory's shearwater (Calonectris borealis). We aimed to clarify behaviours, within a close range proximity to artificial light sources, simulating grounding: first, we tested choices of fledglings regarding different light sources (white versus no light, and red versus blue), and second, we tested whether adults’ behaviour could explain the age bias observed in fallouts.

It is generally considered that blue-rich light is more disruptive to organisms (Falchi et al., 2011; Longcore et al., 2018; Longcore, 2023), having led to a higher magnitude of behavioural responses in experiments with varied taxa (sea turtles: Witherington and Bjorndal, 1991; migratory passerines: Doppler et al., 2015; Zhao et al., 2020; and seabirds: Atchoi et al., 2023; Rodríguez et al., 2018; Syposz et al., 2021). Thus, we expected both fledglings and adults to mirror these tendencies and show a preference for corridors with no light or lit with light of longer wavelength (i.e. less disruptive light). Fallouts affect age groups disproportionately as fledglings are grounded by the thousands, during their first flights from nest to sea; adults, however, are seldom found grounded (Rodríguez et al., 2017a), even while flying over lit areas frequently when travelling to and from their colonies throughout the breeding season, and are even found to breed successfully in inland colonies (Chevillon et al., 2022). Thus, because adults have previous experience with light-polluted landscapes, we expected them to react faster to the experimental stimuli.

Cory's shearwaters Calonectris borealis (Cory 1881) breed in subtropical north Atlantic archipelagos, with their largest population located in the Azores archipelago (BirdLife International, 2024, Species factsheet: Cory's Shearwater Calonectris borealis; https://datazone.birdlife.org/species/factsheet/corys-shearwater-calonectris-borealis on 30/09/2024). The breeding period spans from mid-February to early-November, during which breeding adults nest in burrows and raise a single chick (Monteiro et al., 1996). Fledging occurs in late October early November, when an estimated 6% of fledglings are annually grounded in fallout events in the Azores (Fontaine et al., 2011; Rodrigues et al., 2012). To minimise mortality, rescue campaigns (‘SOS Cagarro’) have been conducted in the Azores each year since 1995. During these campaigns, grounded fledglings are collected and released at sea, reducing direct mortality caused by light pollution (Fontaine et al., 2011; Atchoi et al., 2021).

In our study, we sampled 131 fledglings and 84 adults. Fledglings were sampled from the pool of rescued individuals during the SOS Cagarro campaign on Faial Island, Azores. All rescued individuals were brought to the operational centre in Horta, the main town of the island. We selected a cut-off minimum mass of the fledglings to ensure each individual was sufficiently healthy to take part in the experiment and be released thereafter. The minimum for selection was 660 g, and the mean (±s.d.) for selected studied fledglings was 820±74 g (n=128). The mass of fledglings rescued during the 2020 and 2021 SOS campaign ranged from 360 to 1080 g (mean 766±96 g, n=1111), comparable to the long-term average of 750 g (Cuesta-García et al., 2022). We selected 660 g to account for the number of available birds but still ensuring individuals were healthy enough to survive release. We ringed each bird with a unique numbered metal band and took biometrics of the fledglings in the morning following capture, then placed the birds into rescue boxes (cardboard boxes of 20×20×40 cm with ventilation holes used to temporarily keep rescued birds during the SOS Cagarro campaign) and left them in a quiet and dark place until the experimental test, which occurred in the evening of the same day. We tested fledglings for five nights in 2020 and for four nights in 2021, during October and November. The experimental maze was placed on the edge of a 20 m high cliff (38.5199N, 28.6202W) on the outskirts of Horta. The two exits of the maze, which were parallel to each other, were set perpendicular to the cliff face, oriented at 100°E, facing the sea. Fledglings exiting the maze would have a small drop (<50 cm) to a grassy slope, from where they could fly towards the sea whenever ready.

Adult shearwaters were captured and tested at the breeding colony of Capelinhos (38.5895N, 28.8233W), a Special Protection Area (SPA) from the Natura 2000 Network, an area with low light pollution levels situated on Faial Island. Capture occurred after sunset, either at the nest or on the ground. Each adult was ringed with a unique numbered metal band and had its biometric measurements taken. After capture and ringing, adults were transferred into SOS Cagarro cardboard boxes and brought to the testing site, just at the foot of the colony (<20 m), where they were kept for a minimum of 15 min before being tested to give them a period of time to adjust to the darkness and adjust after the human manipulation. We tested adults for 11 nights between May and September 2022. The maze was placed at the foot of the colony, in a flat ground-levelled area facing the sea, <10 m from the ocean surf. Adults leaving the maze walked into the colony grounds and were gently redirected to abandon the area directly in front of the maze to avoid choice bias for the next bird tested.

Handling and data collection from Cory's shearwaters and the execution of the experiment were conducted under licences from the Azorean government (46/2020/DRA, 60/2021/DRAAC, 48/2022/DRAAC, 25/2021/DRCTD and 36/2022/DRCTD), and ethical licences from the University of the Azores (UAC/2020/4182, UAC/2021/5720). Biometric data from grounded Cory's shearwater fledglings (2020–2021) were collected by E.A. in the frame of the SOS Cagarro campaign, coordinated by the Regional Directorate for Maritime Policies (DRPM of the Azorean Regional Government), with the collaboration of a network of institutional partners, including the Faial Island Environment and Climate Change Service. The responsibility for all conclusions drawn from the data lies entirely with the authors.

The two-choice maze was a 100×50×40 cm plywood box, painted in matte black, with a separator in the middle to create two distinct exits (Fig. 1). Lights were placed on the top of the maze in such a way that each illuminated the corresponding chamber as well as the bird from above. A smaller wooden box (black box) with black painted inner walls was fitted inside the back of the maze. This box was able to move vertically, so that we were able to cover the bird inside it and pull the box upward to release the bird into the maze, revealing the choice arms. At its highest position, when the bird would be free to move around the maze, the walls of the box limited light spill from one side to the other. Individuals were placed one at a time inside the black box, at a centre position in the back area of the maze, and stayed under the cover of the black box for 1 min. After 1 min, the black box was pulled upward where it stayed for the duration of the test. Behaviours were directly observed via the observation window at the back of the maze and by visual confirmation at the choice arms whenever an individual exited the maze. To avoid confounding effects of odour, the maze was regularly cleaned with 70% ethanol (Bonadonna et al., 2006) after a couple of tests, or whenever an individual had defecated or regurgitated inside the maze. A black cardboard sheet was placed at the back of the maze where the individuals stood at the initial placement and replaced whenever the individual defecated or at the end of each experimental night.

For the white light stimulus, we used a commercially available LED strip (3000K warm white; AA-20-FLEX; by LEDsupply, Randolph, VT, USA) with a broad emission spectrum with one blue peak at 455 nm and a second, higher peak at 625.2 nm (Fig. S1A). This light was not adjusted to be isoluminant for the shearwaters’ visual system, unlike the red and blue stimuli, because it was not to be tested in a pair with either red or blue LEDs. The white light type was chosen to be a close simulation of the urban areas fledglings and adults encounter, i.e. new installations of LED lights in Azorean urban areas which range from 3000K to 4000K (personal communication from Electricidade dos Açores). After being installed in the maze, the white LEDs yielded an illuminance of 175 lx at the starting position of a bird. During control treatments, the illuminance at the bird’s starting position was <0.1 lx, which is the minimum detection range of the light meter used (Hagner Screen-Master light meter; B. Hagner, Solna, Sweden).

The red and blue LED lights used in this study, as well as the aim and methods of their design, were the same as in Atchoi et al. (2023), i.e. ‘they were designed to be as isoluminant as possible for the achromatic visual system, but to have strong contrast for the chromatic visual system’ (Atchoi et al., 2023). Each LED had only one peak, the blue around 450 nm, and the red around 620 nm (Fig. S2B,C). So, we were able to obtain (i) a strong chromatic contrast (60.7 just noticeable differences, JNDs; the discrimination threshold is 1 JND) between them; and (ii) a very low achromatic contrast, estimated at 4%. Usually, birds display very poor achromatic contrast sensitivity to stationary stimuli, being unable to detect contrasts below 10% (Lind et al., 2012; Olsson et al., 2018). The spectral sensitivities were modelled based on visual parameters of the wedge-tailed shearwater, Ardenna pacifica, the most closely related species to Cory's shearwater for which such data exist (Hart, 2004). We assumed similar capacities for Cory's shearwaters, and as such they would not be able to distinguish the red and blue light stimuli based solely on luminance. For the human visual system, the blue LEDs yielded an illuminance of 75 lx and the red LED of 120 lx at the starting position of a bird (measured with a Hagner Screen-Master light meter; B. Hagner). These illuminances correspond to light levels at civil twilight (Cronin, 2014) and are comparable to what a chick would be exposed to outside its burrow after sunset. Further details on the red and blue light system, including dimensions, measurements and sensitivity modelling, can be found in Atchoi et al. (2023), and in the Supplementary Materials and Methods.

We used a two-choice maze to test the behavioural preferences of fledglings and adults against light stimuli. Each bird was subjected to only one trial, and presented with one out of three treatments: (1) white–dark, where one arm had no light stimulus (dark), and the other was fitted with a white light; (2) blue–red, where one arm was illuminated by a red light and the other by a blue light of equal intensity (see ‘Light stimuli’, above); and (3) control, where both arms were without any experimental light, i.e. both choices were ‘dark’. To be able to identify any bias generated by the structure of the box itself, we registered whether the bird entered the ‘left’ or ‘right’ arm. To eliminate physical bias of the maze structure itself, the position of the lights was randomly interchanged between arms. We attributed to each treatment and order of lights a number (e.g. 1 – red in left arm blue in right arm, 2 – blue in left arm red in the right, and so forth), then we ran a randomiser in Excel with the function ‘RAND()’ on the list of numbers and followed it while applying the treatments. For each bird, we recorded (1) entry into the arm (when the bird moved from the original starting place and stepped into one of the maze's arms, even if it did not exit the maze) or when it remained still in the original starting place (a bird was considered still even when moving, if it did not step forward into an arm and stayed in the starting area; Fig. 1), (2) choice (white or dark, red or blue, left or right) and (3) latency to enter (in minutes). Whenever an individual was still, the test ended after 10 min, and the individual was removed from the box and released. Because the treatments were assigned in a random order, it resulted in an uneven number of trials per treatment (Fig. 2). Latency to enter was marked in minutes; thus, from the beginning 00:00 to 01:00 was attributed as 1 min response, 01:01 to 02:00 was attributed as 2 min, and so forth.

We ran a Fisher exact test to determine the differences between the two age groups concerning the number of entries and choices made. Choice was analysed with an exact two-tailed binomial test, for each treatment separated by age group. To estimate differences in latency to enter within treatments for each age group we used Welch two-sample t-tests. As a single measurement may not reflect body size in a reliable manner (Rising and Somers, 1989), we conducted a principal component analysis (PCA) on biometric data (wing, tarsus, culmen and gonys – mm; Fig. S4). We then used the estimates from the first component (which retained 51% of the variation) as a body size index (BSI), following Rodríguez et al. (2017a,b). Subsequently we calculated body condition indices (BCI) as the standardised residuals of a linear regression of body mass on BSI. Positive values of BSI and BCI, respectively, represent larger individuals and individuals heavier than average after controlling for body size, i.e. in a presumably better body condition (heavier fledglings have higher survival rates; Mougin et al., 2000). Body mass was highly correlated with both PC1 [Pearson correlation coefficient=0.553; 95% confidence interval (CI)=0.448 to 0.643] and the residuals from the chosen model (Pearson correlation coefficient=0.833; 95% CI=0.785 to 0.871).

Generalised linear models with binomial error distributions and logit link functions were conducted for each age group and each treatment to detect relationships between continuous explanatory variables (BSI and BCI) and the binomial response variables (a) entry (yes=1, no=0) and (b) choice (white=1, dark=0; blue=1, red=0; left=1, right=0). Linear models were conducted to determine relationship between the continuous explanatory variables (BSI and BCI) and a continuous variable (time in minutes) for the latency to enter. All statistical analyses and data manipulation were conducted using R version 4.1.2. (http://www.R-project.org/).

The proportion of fledglings entering an arm was significantly higher (1.44 times) than that of adults. Fledglings entered during 77% of trials (101/131) while adults did so in 54% of trials (45/84) (Fisher exact test: P<0.001, odds ratio=2.88; Table S1) with differences being detected among treatments (Fig. 2). During the control treatment, all fledglings entered an arm; in fact, most exited the maze, while adults only entered an arm in 70% of cases. In the white–dark treatment, the two groups showed similar entry rates (Fisher exact test: P=0.367, odds ratio=1.55; Table S1). Fledglings entered the maze's arms significantly more often than adults in the blue–red treatment (Fisher exact: P=0.015, odds ratio=3.4).

In the control treatment, there was no difference between left and right arm choices for either age group (Fig. 2; binomial exact tests, fledglings: P=0.627; adults: P=0.115). During the white–dark treatment, fledglings consistently chose the dark arm of the maze over the lit one, whereas adults were less clear, despite also choosing the dark arm more often (Fig. 2; binomial exact test fledglings P=0.009; adults P=0.3). Both age groups showed a clear choice for red light during the blue–red treatment (Fig. 2; binomial exact test, fledglings P=0.003; adults P=0.00052). Also, we found that fledglings were as likely as adults to choose the red and dark arms of the maze, without significant differences in the choices of the two age groups when compared within each treatment (Fisher exact test, red light: P=0.25; dark: P=0.5).

Fledglings were slower to enter any arm than adults (Fig. 3; Welch t-test P<0.01, 95% CI 0.531 to 1.766). Both age groups were quicker in the control treatment than in either of the other two treatments (Fig. 3). Fledglings’ latency to enter did not differ whether they chose the left or the right arm of the maze in the control treatment (Welch two-sample t-test: t₂₃=0.65, P=0.521), and regardless of choosing the light or dark arm during the white–dark treatment. Fledglings appeared to be quicker when choosing the lit arm yet the differences were not statistically significant (Welch two-sample t-test: t₇=−0.989, P=0.353; Fig. 3). Latency to enter did not differ between the blue or red arm in the blue–red treatment (Welch two-sample t-test: t₁₅=−1.04, P=0.314).

Adults’ latency to enter did not differ between the left and right arm during the control treatment (Welch two-sample t-test: t₅=1, P=0.363), or differ between the lit and the dark arm during white–dark treatment (Welch two-sample t-test: t₁₂=−0.923, P=0.373). It was not possible to determine differences for adults’ latency to enter an arm between blue and red as only one adult chose the blue arm.

Only two weak relationships were identified in the models. First, larger fledglings showed a tendency to enter an arm more often during the blue–red treatment (glm estimate: 0.581, 95% CI 0.007 to 1.253), and second, larger adults showed a tendency to enter an arm more often during the control treatment (glm estimate: 1.161, 95% CI 0.120, 3.000). No other significant relationship between BSI or BCI and choice (Fig. 4) or latency to entry were observed (Table S2, Fig. S3).

Our study verified behavioural choices of Cory's shearwaters for light stimuli and showed that these choices did not differ significantly between fledglings and adults. These findings are consistent with those of previous studies, which showed that (1) Cory's shearwater chicks display an affinity for dark environments when exposed to artificial light at close range (Atchoi et al., 2023), and (2) adults of closely related species, such as Manx and Yelkouan shearwaters, avoid their colony when the latter is sporadically exposed to artificial light (Syposz et al., 2021; Austad et al., 2023). In addition, both age groups displayed stronger avoidance of the arm with shorter wavelength artificial lights. We did uncover age-specific differences in the frequency and time taken to enter an arm of the maze.

In our study, all fledglings entered an arm, and made a choice, whenever light stimuli were absent from the experimental maze, i.e. in the control treatment. However, when light stimuli were present (white–dark and blue–red treatments), some fledglings were deterred from making a choice (Fig. 2). Furthermore, the fledglings that made a choice were quicker to do so in the control treatment, but slower in treatments with light stimuli (Fig. 3). Our data indicate that the presence of a close-range artificial light can be disruptive to Cory's shearwater fledglings, perhaps because of disorientation or fright, influencing their behaviour. More specifically, artificial light appears to impair the fledglings’ ability to exit the unknown lit environment, thus increasing their vulnerability to dangers present in urban areas for those that become grounded. Our results agree with the observations of irregular and chaotic flights displayed by fledglings during fallout events (E.A., T. Pipa, A.R., B.M. and J.B., personal observation). A proximate cause leading fledglings to converge on urban lit areas remains undetermined; it may be due to contrasting environments, or other factors. Regardless, when fledglings come into close range of artificial lights, they apparently become disoriented by the stimuli rather than being behaviourally attracted to them (see tracked flights in Rodríguez et al., 2022).

Disorientation could be a result of artificial light interfering with the bird's visual system. Studies have described an underdeveloped and untrained visual system in growing Procellariiformes nestlings: older Leach's storm petrel chicks Hydrobates leucorhous reacted more times to light stimuli than did younger ones, i.e. underdeveloped chicks (Mitkus et al., 2018), and growing Cory's shearwater chicks reacted more to light stimuli the more they had been previously exposed to light, i.e. trained chicks (Atchoi et al., 2023). In the current study, fledglings were slower to react than adult birds in general (Fig. 3), suggesting that the light-induced effect is stronger for the younger birds. Additionally, the slower reactions of fledglings, even in control treatments, indicate their general behavioural inexperience. Our results suggest that the underdeveloped and untrained visual system of the chicks of burrow-nesting seabirds (Atchoi et al., 2020) will increase their vulnerability to the adverse effects caused by light pollution upon fledgling, and that this vulnerability, observed by the apparent disorientation displayed by the first flights of fledglings, is reduced in adult birds with both visual and navigational experience (Imber, 1975).

The delayed development of the visual system of underground-nesting Procellariiformes could, in turn, be explained by the conditions faced during the chick development phase. Indeed, the chicks of these species spend the first weeks of their lives in dark burrows which lack sufficient light stimuli needed for the full development of a functional visual system (Mitkus et al., 2018; Atchoi et al., 2023). The interference of light with the visual abilities of fledglings as they begin to navigate these new lit environments in flight might be a factor explaining the age bias observed in fallout events. Our data showed a similar behavioural preference between age groups, so behavioural differences cannot account for the age bias observed in fallouts. Adults displayed similar behaviours to fledglings, with overall fewer choices made during treatments with light stimuli (Fig. 2) and were quicker to exit the maze during the control treatment. As with fledglings, the presence of artificial light also contributed to a disruption of the adults’ behaviour. Other studies observed adult Procellariiformes avoiding their colonies when a temporary light source was turned on (shearwaters: Austad et al., 2023; Syposz et al., 2021).

Fledglings and adults made similar choices for the dark arm of the maze versus the white light and for the red versus the blue LED arm (Fig. 2). The behavioural choices displayed by adults in the white–dark treatment were similar to those made by fledglings (Fig. 2; fledglings: 23% chose white, 77% chose dark; adults: 33% chose white, 66% chose dark), even if adult choices did not differ significantly between arms. The lower clarity in adult behaviour could be attributed to a smaller sample size (Fig. 1), and the tendency should be confirmed if more adults were sampled. We executed a power analysis which indicated the need for between 176 and 305 birds to confirm the tendencies (sample sizes difficult to obtain with free-living Cory's shearwater adults). Yet, the similarly low sample size used in the blue–red treatment was sufficient to result in clear preferences for both age groups (Fig. 2).

Blue-rich light has been found to be more disruptive to burrow-nesting seabirds (Rodríguez et al., 2017a,b,2018; Syposz et al., 2021). Blue light has also been recognised as the most disturbing wavelength for most taxa, and most recommendations to reduce the overall effect and the amount of light pollution highlight a shift of urban light spectral emissions towards longer wavelengths (Gaston and Sánchez de Miguel, 2022; Rodríguez, 2023). Blue-rich light has greater effects on diverse taxa, both diurnal and nocturnal (Longcore et al., 2018). Accordingly, in our study, we interpreted the preference for the red light and dark arms of the maze as evidence for stronger effects induced by the blue light, which is also present in the broad spectrum white light (it has a peak emission in the short wavelengths) (Figs S1 and S2). Indeed, Procellariiformes forage over, and dive into, clear waters typically rich in short-wavelength light (400–500 nm); accordingly, their visual systems seem to be attuned to such environments (Hart, 2004). Shearwaters forage during the day, so they need visual systems able to navigate flat landscapes with different light conditions, and they need to orientate themselves in the colonies at night; however, they may use the olfactory system and vocalise during the night, as they may be more effective than shearwater vision (Bonadonna and Bretagnolle, 2002).

Notably, the white–dark treatment had a less pronounced effect on adult choices (Fig. 2). The white–dark treatment revealed an increased reaction speed when fledglings chose white and a higher latency when choosing dark. Although we did not directly measure and compare emissions between the white and the blue or red LEDs for the shearwater colour visual model (see Materials and Methods) (Figs S1 and S2), photometric measurements within the maze (in lux) suggest that the intensity of the white light should have been similar to that of the blue and red LEDs, as they all shared a similar magnitude of intensity (white 175 lx, blue 75 lx, red 120 lx). Despite this similarity, during the white–dark treatment, fewer fledglings and adults entered an arm than in trials with the blue–red treatment, and even fewer than during the control treatment (Fig. 2). Furthermore, fledglings selecting the white-light arm exhibited much faster reactions than those choosing the dark arm (Fig. 3). Given the comparable intensity of the three light treatments, we do not attribute the lower amount of choices made during white–dark treatments to differences in light intensity. Rather, the white–dark treatment presented a more contrasting environment, with white light against a dark area [∼175 lx in the white area and <0.1 lx in the dark arm – see illuminance (lux) and luminance (cd m⁻²) values in Fig. S2], compared with the blue versus red light (∼45 lx difference between the two arms). This high contrast of the white–dark treatment may weaken the birds’ decision-making capabilities and worsen disorientation caused by light pollution after grounding. Our results agree with observations that a stronger contrast between well-defined lit and dark zones disorients shearwaters in flight (Guilford et al., 2018). The broad emission spectra of the white-light treatment may also contribute to a greater inability to make a choice, i.e. to navigate the lit environment. Indeed, Manx shearwater adults showed a greater aversion to flying into the colony when exposed to white light compared with blue light (Syposz et al., 2021). Additionally, broad white light emissions led to more grounded fledglings compared with more monochromatic lights (Rodríguez et al., 2017a,b).

All fledglings tested in our study had already been grounded by artificial lighting in urban areas. These individuals could have been biased to avoid light after having already associated lit areas with stress; however, as a group they were subjected to similar stress conditions. Testing individuals with limited or no exposure to artificial light during growth might lead to distinct results. In other words, unexposed shearwater fledglings might innately display a preference for light sources versus dark environments. Indeed, there is evidence of behavioural shifts in other cavity-nesting bird taxa; for example, in common starlings Sturnus vulgaris, which display negative phototaxis as young chicks, but switch to positive phototaxis as they approach fledging (Minot, 1988). However, this is unlikely. Previous work revealed avoidance of light stimuli in Cory's shearwater chicks aged from 3 weeks old to nearly fledging (Atchoi et al., 2023), suggesting our results were not skewed as a result of the previous exposure to ALAN, nor do Cory's shearwaters go through a behavioural shift similar to that of starlings. Testing light-unexposed fledglings should further clarify the behavioural preferences of this species, yet the practical constraints associated with fieldwork make it challenging to obtain an adequate sample size; in addition, the difficulty in selecting fledgling birds means we risk taking unprepared birds if we collect these young shearwaters from their nest. It is also unlikely that different age groups would display different behaviours, as our study showed that adults behave similarly to fledglings (Fig. 2).

In our study, not all birds made a choice, particularly adults (only half of them completed the trials) (Fig. 2). While it is difficult to determine why individuals make or do not make choices in such settings, it has been proposed that a choice may be derived from specific needs at the time of testing (Brooke, 1998) or personality differences (Bonadonna and Sanz-Aguilar, 2012). In our control treatment, 70% of adults made a choice, whereas all fledglings made a choice. This agrees with the assumption that fledglings are driven by hunger and an instinct to reach the ocean and forage (fledglings lose ca. 40 g per day until the first successful foraging event; Mougin et al., 2000). In contrast, especially in adults, the behavioural preferences of individuals should be activity dependent; adults are experienced, have encountered artificial lights before and have sufficient navigational and flight experience to have survived to breeding age. When comparing fledglings and adults, one needs to note that the two groups were captured and went through different situations before the test, i.e. fledglings were kept in a ventilated box until night for one day, then released into the maze, whereas adults were captured at the colonies when returning during the night, so have not been affected by light pollution and have not spent a day in captivity. We tried to match conditions as much as possible and kept adults for 10–30 min in a ventilated box before submitting them to the maze. The different stress conditions of each group could confound the choices they made. It also could partly explain the less clear choices displayed by adults. Adults come to the colony with different needs or aims (e.g. nest defence, foraging, incubating or chick rearing), and different personalities, all of which will affect their behaviour. However, despite our two age groups being under different pre-capture baselines, we believe their choices and our results would be similar even if we had used rare grounded adults or free fledglings at the colony. Working with free animals in situ is challenging and there is a limitation on variable control that researchers can apply.

The link between personality and body condition is poorly understood, with inconclusive results (Réale et al., 2007). While body condition could affect behavioural responses to stress (Seltmann et al., 2012), our results did not yield consistent patterns regarding body size, body condition and changes in behaviour (Fig. 4). Our models detected only two weak effects (see estimates and confidence intervals in Table S2).

We sampled fledglings based on their apparent health and heavier body mass (see Materials and Methods), limiting the representativeness of the sample. Random sampling of grounded fledglings without body mass thresholds could unveil relationships between body condition and behaviour towards light stimuli undetected in this study. The consequences of exhibiting bolder or shyer behaviour in the context of fallout events should be further investigated.

Finally, other factors, such as differential exposure to light during growth, via the amount of natural or artificial light that enters a burrow, might lead to differences in the development of the visual system (Atchoi et al., 2023) or even in personality (Ruiz-Raya et al., 2022). These factors might also partially explain the rate of responses and vulnerability to light pollution at the individual level. Further research is needed to identify the most vulnerable subgroups, and to determine how mitigative measures can be extended in both time and space to reduce the effects of light pollution on seabird populations.

The disruption of sensory systems by artificial light appears to lead fledging seabirds to present erratic flights, an inability to orientate towards the sea, and ultimately, groundings (Rodríguez et al., 2017a,b, 2022). Exposure to artificial light leads adult shearwaters to display short-term and local disorientation (Guilford et al., 2018), temporary geographical avoidance (Syposz et al., 2021; Austad et al., 2023) and avoidance of lit areas during flights from sea to the nest (Rodríguez et al., 2016). The increase in light pollution (Kyba et al., 2023) and its shift to a more intense and blue-rich spectral composition (Sánchez de Miguel et al., 2021) may worsen the behavioural consequences for seabirds. Our study adds to the growing body of knowledge regarding the effects of this anthropogenic pressure and supports recommendations to shift urban lighting to more monochrome and longer-wavelength lighting systems, while reducing the implementation of artificial night lighting in both natural and urban areas. Further research is needed to identify the most vulnerable subgroups, and how mitigative measures can be extended in both time and space to reduce the effects of light pollution on seabird populations.

The authors are very thankful to Institute of Marine Sciences - OKEANOS, University of the Azores, our core center for this study, which provided logistical, material, counsel and management support. We are grateful to Joseph Lewin for constructing the maze and Lorenzo Kunze for providing expert logistical support with the light stimuli. We thank Pranciškus Vitta, Olle Lind and Peter Olsson for the help and advice on the preparation of light stimuli. We thank OMA (Observatório do Mar dos Açores) for providing staff and support during the SOS Cagarro campaign and field work. We thank João Lagoa, Maria São Bento and Hannah Gretham for contributing to data collection, and Ellie Ga for constructive discussions on the topic which greatly improved the study. Thanks to Pedro Marques and Alberto Lopez for collecting data during the SOS Cagarro campaign and improving the discussion of the topic, and to Scott Cameron at LEDsupply for technical guidance and general notes on seabird welfare. We are grateful to Faial Island Environment and Climate Change Services and its rangers for the logistical support. Finally, the authors extend their gratitude to the Regional Directorate for Maritime Policies (DRPM of the Azorean Regional Government) for providing access to the rescued fledglings and to all volunteers and trained staff involved in the collection, rescue and processing of grounded Cory's shearwaters during the SOS Cagarro campaign on Faial Island. Rescued fledglings’ biometrics were collected as part of the SOS Cagarro campaign, coordinated by the Regional Directorate for Maritime Policies with the collaboration of a network of institutional partners, including the Island Environment and Climate Change Services. The responsibility for all conclusions drawn from the data lies entirely with the authors.