Diving is indicative of foraging in cormorants (Phalacrocoracidae). We have investigated a range of parameters associated with diving in Great Cormorants Phalacrocorax carbo to provide insight into the bases of cormorant predatory strategies. We hypothesize that if vision is important in cormorant foraging behaviour, and if they are not constrained by the position of their prey in the water column, then diving behaviour will be modulated primarily in response to the diel variation in ambient light levels. Specifically, we propose that cormorants forage at shallower depths when light levels are low, and more deeply when light levels are high. We provide evidence that this is the case. We recorded the occurrence of cormorant diving behaviour using implanted data loggers and recorded ambient light levels and water temperature using leg-mounted loggers in a sample of free-living Great Cormorants in Greenland. Our results show that dives are shallower at the beginning and end of each day when light levels are lower. We suggest that these data support the hypothesis that cormorant foraging is visually-guided even though recent evidence has shown that their underwater visual acuity is poor.
Recent evidence shows that Great Cormorants Phalacrocorax carbo have relatively poor visual acuity underwater (White et al. 2007) and that their acuity declines further in turbid conditions (Strod et al. 2004). Furthermore, a model of prey detectability based upon these acuity data indicated that even high-contrast fish will be visible only at short range (< 1 m; White et al. 2007). This raises a number of questions concerning the role of vision in guiding the foraging behaviour of this species. Nevertheless, a range of anatomical, behavioural and dietary observations indicate that vision is important in cormorant foraging. Thus, their eyes have been described as being well suited for an amphibious lifestyle (Hess 1909, Hess 1913) and they are known to have a high capacity to alter the refractive power of their lens to compensate for the loss of corneal refraction that occurs upon immersion (Glasser & Howland 1996, Katzir & Howland 2003).
The great majority of cormorant foraging dives occur during the day, although they have been recorded foraging at night (King et al. 1998, Sapoznikow & Quintana 2002, Grémillet et al. 2005a). In Greenland, the diving of Great Cormorants is apparently constrained by day length, and extends into the night only during the shortest midwinter days at high latitudes when presumably daytime foraging opportunities are curtailed (Grémillet et al. 2005b). In addition, 70% of night absences from colonies and roosts (an indicator that birds were foraging) recorded for Rock Shags Phalacrocorax magellanicus and Imperial Shags P. atriceps coincide with a half-full or nearly full moon (Sapoznikow & Quintana 2002).
In turbid conditions, where visual acuity will be very low (Strod et al. 2004, White et al. 2007), cormorants may use mass fishing techniques to drive fish to relatively clear surface waters where they are more likely to be detected when seen from below in silhouette against the downwelling light (Van Eerden & Voslamber 1995). Indeed, Great Cormorants have been shown to feed exclusively on pelagic fish during social foraging (Grémillet et al. 1998). When Double-Crested Cormorants P. auritus dive at night they forage on shad Alosa spp. that are very close to the water surface (King et al. 1998).
The present study overcomes some of the limitations of this piecemeal evidence on the role of vision in cormorant foraging by using data loggers to determine both the diving behaviour of individual Great Cormorants and the light levels that they experience throughout the year at a study site in Greenland. From this, we have been able to determine that cormorants modulate their foraging behaviour in response to variation in ambient illumination, supporting the hypothesis that vision is the primary cue employed to guide the foraging of cormorants, even though their underwater acuity is relatively poor.
We employed two types of data loggers to determine the occurrence, depth and duration of dives in Great Cormorants during different parts of their annual cycle, and to determine ambient light levels and water temperatures associated with diving.
Occurrence, depth and duration of dives
Ten breeding male Great Cormorants raising young chicks on Disko (69°30′N, 54°05′W), West Greenland, were equipped with depth data loggers (DLs; dimensions 60 × 24 × 7 mm, mass 20 g, c. 0.6% of the body mass of the cormorants studied; Woakes et al. 1995) in June 2002. The DLs were programmed to record data every second day, so that the logger memory would last an entire year. On the days during which DLs were recording, pressure (depth) was stored every 2 s. All devices were calibrated before and after use (depth resolution 0.1 m). Loggers were surgically implanted under isoflurane anaesthesia following Stephenson et al. (1986); additional details and analyses of these data are provided elsewhere (Grémillet et al. 2005a, 2005b). Loggers were recovered from the birds using the same surgical procedure. All surgeries were conducted by trained veterinary personnel under permits of the ethics committee of the French Polar Institute, The Arctic Station Godhavn (Copenhagen University), the Danish Polar Center, the Danish veterinary administration, and the Greenland Home Rule Government (permits MP/14/18/04/03 and MP/Pj/12/20/04/04).
All of the implanted birds bred normally, producing 2.9 ± 0.9 (sd) chicks per nest compared with 3.1 ± 0.9 chicks for 50 control nests (Grémillet et al. 2005b). Nine of the implanted birds were re-trapped in June 2003, and the tenth was re-trapped a year later. Eight of the implanted birds were resighted while breeding in June 2004, compared with nine of 15 control birds marked with metal rings (Grémillet et al. 2005b). Only seven of the 10 loggers provided complete recordings. The six loggers that were recovered after 1 year recorded data for periods of 7–12 months. The seventh logger, which was recovered after 2 years, recorded data for 14 months.
Light levels and water temperature
Ten leg-ring mounted light and temperature loggers (GeoLT, Earth and Ocean Technologies, Germany; 14 mm diameter × 45 mm long, 8.2 g; c. 0.2% of body mass of the cormorants studied) were deployed in June 2004, but only four were recovered (all birds were males). Two of the recovered loggers recorded ambient light every 30 s for 324 days, and two recorded ambient light every 45 s for 391 days. Light resolution was 4.5% to < 0.025% of reading, depending on level. The logger used a silicon photodiode sensor with a filter providing a spectral range of 300–700 nm and maximum sensitivity at 550 nm. The filter was near-photopic, with a slight blue–green shift, and was therefore comparable with normal luxmeters. Minimum detectable light level was 0.23–0.25 lux.
Recorded temperatures of the light loggers were used to identify bouts of diving, from which water temperature and ambient aquatic light levels were estimated. Bouts of diving were identified on the basis of stable temperatures that persisted for several minutes (typically > 10 min). This stable value was assumed to represent water temperature. Because the GLS loggers did not record depth, it was not possible to be certain that the measured illumination at any point in time was representative of illumination at depth. For each dive bout (i.e. period of stable temperature), a single estimate of ambient illumination was taken for a randomly selected point in time within the bout.
Data treatment and analysis
Diel variation in diving behaviour (occurrence, depth and duration) was assessed using custom-written BASIC software. This software was used to determine the number of birds that were recorded diving during each 15-min period of each day, and to display this value on a 16-bin blue–green–red scale, where each colour bin corresponds to a 30-beats/min range.
Although the resolution of the depth data loggers was 0.1 m, dives shallower than 1 m could not be reliably detected as they were obscured by electronic and mechanical noise, and therefore only dives deeper than 1 m are considered. Exploratory analysis revealed that dives extended into the night during midwinter (Fig. 1 and Grémillet et al. 2005a). We selected dives occurring during the period 23 December to 8 January for detailed analysis because during this period birds dive over a wide range of ambient illumination conditions.
Diving activity frequently occurred in bouts, such that clusters of dives (diving bouts) were separated by longer periods without diving. In order to select comparable periods of diving activity for each of the birds, it was necessary to identify objectively diving bouts, so that comparable bouts (e.g. the first bout of the day, see below) could be examined for each bird. This was accomplished using standard bout-end criterion analysis (Gentry & Kooyman 1986) and sequential differences analysis (Mori et al. 2001). These techniques are used to examine the frequency distribution of surface intervals (bout-end criterion) or the frequency distribution of differences between successive dives (sequential differences analysis) to separate bouts. Both methods identified a bout-end time of approximately 200 s, so a bout was defined as a series of consecutive dives that are separated by surface intervals of no more than 200 s. Visual inspection of dive data suggested that these analyses were differentiating dive bouts adequately. Each of the dives during the first (dawn), second (daytime) and final bouts of each day during this period were extracted for further analysis. Dives occurring during other bouts were not considered in the analysis, and the final bout of a day was considered only if it occurred at night. The second bout of the day was selected as representative of daytime foraging, as this bout corresponded most closely with midday (95% of dives during the second bout occurred between 09:46 and 12:29 h).
For each dive, the duration and maximum depth were calculated, as was the duration of the surface period following each dive. The frequency distributions of dive duration and dive depth were then compared between bouts occurring at different times of day. Most data were approximately normal (Shapiro–Wilk P < 0.05) and so met the requirements for parametric tests (but see below). The effect of bout timing (dawn, daytime, night) on the relationship between dive depth and duration was then examined using a General Linear Model (GLM) with dive duration as the dependent variable, dive depth and bout timing as fixed factors, and bird ID as a random factor. The effect of bout timing on dive depth was also examined, considering only dives of 35–45 s in duration, using a GLM with dive depth as the dependent variable and bout timing and bird ID as fixed and random independent factors, respectively. There were some variables where the Shapiro–Wilk test indicated departure from a normal distribution. In these cases, the above analyses were repeated using non-parametric anova with rank-transformed data. The results of the non-parametric analyses were qualitatively similar to the parametric tests (i.e. significance did not change, although the P-values for the non-parametric tests were generally higher), suggesting that the conclusions were not influenced by the non-normality of the data. For simplicity, only the results of the parametric tests are presented.
Dive/pause ratio was calculated as dive duration divided by the duration of the subsequent surface interval. Differences in dive/pause ratio, log(ambient illumination) and water temperature between bouts was tested using GLMs with dive/pause ratio, log(ambient illumination) and water temperature as the dependent variable, and bout and bird ID as fixed and random independent factors, respectively. Tukey's HSD was used to identify significant differences between bouts. Alpha was set at 0.05 for all tests.
During the mid-winter period, diving frequency (assessed as the proportion of monitored birds that were recorded diving during each 15-min interval) rose at dawn (Fig. 1). This is at a time that corresponded well with the rise in ambient illumination to detectable levels (> 0.25 lux). During this period, but unlike other times of the year, dives extended into the night, albeit at low frequency (Fig. 1).
Dive duration and depth
A summary of dive parameters for Great Cormorants in Greenland is provided in Table 1. The relationship between dive duration and depth was significantly different between dive bouts (F2,1534 = 39.1, P < 0.0001; Fig. 2). The relationship was not significantly different between the dawn and night bouts (Tukey's HSD). The relationship for both the dawn and the night bout differed significantly from the relationship for the daytime bout (Tukey's HSD). For dives of 35–45 s duration, depth was significantly different between bouts (F2,527 = 155, P < 0.0001), with the dawn, daytime and night bouts each being significantly different from each of the other bouts (Tukey's HSD).
Table 1. Diving behaviour of Great Cormorants during winter in Greenland (data are shown as grand mean ± se of individual medians).
Dawn bout (n = 7 birds)
Daytime bout (n = 7 birds)
Night bout (n = 3 birds)
At this time of year, Great Cormorants in Greenland spend around 65 min/day in water (Grémillet et al. 2005b).
The frequency distribution of dive durations and depths also exhibited clear differences between day and night foraging bouts (Fig. 3). During the dawn and daytime bouts, the frequency distribution of dive depths was strongly bimodal, and dives during the daytime bout were deeper. Dives occurring during the dawn and daytime bouts reached maximum depths of around 30 m. However, during the night diving behaviour was different. Night-time dives were far shallower than those during the day, the frequency distribution of dive depths was unimodal and dives appeared to be strongly constrained to depths of around 10 m (Fig. 3). The bimodal distribution of dives during the day arose, in part, because individual birds dived to different depths, although some individuals showed a similar bimodal distribution of dive depth.
Dive/pause ratio was significantly different between bouts (F2,1524 = 9.3, P = 0.0001; Fig. 4). Dive/pause ratios for the dawn and daytime bouts were not significantly different; both were significantly different from the dive/pause ratio of the night bout (Tukey's HSD).
Water temperature and ambient illumination
Water temperature did not differ significantly between dive bouts (F2,78 = 1.3, P = 0.20; Fig. 5). Ambient illumination was significantly different between the dawn and daytime bouts of the day (F1,60 = 26.9, P < 0.0001; Fig. 5). It was not possible to determine if ambient illumination of the night bout was different from the dawn and daytime bouts, because ambient illumination during this period was consistently below the minimum level that could be detected by the logger.
It is clear that Great Cormorants typically restrict their foraging (diving) behaviour to discrete bouts, showing a strong circadian entrainment (Fig. 1). During winter, diving is most strongly entrained with the diel light cycle, but occurs occasionally during the night (Fig. 1). However, during this period it is also clear that diving behaviour (number of dives, dive depth) differed between bouts in a systematic way (Figs 2 & 4) and dive depth was correlated with ambient light levels (Fig. 6). We interpret these findings as providing clear support for the hypothesis that foraging behaviour is guided primarily by vision.
Diving behaviour and ambient light levels
Figure 1 shows that Great Cormorants in Greenland exhibit a strong circadian entrainment of diving behaviour during winter with most diving confined to daylight hours when ambient illumination is above c. 0.25 lux. The birds equipped with light loggers are known to winter throughout southwest Greenland between Julianehåbsbugten and the Arctic circle south of Sisimiut (our unpubl. data), but the position of the birds equipped with depth data loggers is unknown. It seems likely, however, that these birds also remained within southwest Greenland, due to the close temporal correspondence of variation in light levels and diving behaviour recorded in different years (Fig. 1).
The lower level of illumination encountered by diving Great Cormorants (< 0.25 lux) is similar to the lower limit encountered by diving Blue-eyed Shags Phalacrocorax atriceps (Wanless et al. 1999). Visual acuity of Great Cormorants declines precipitously below this level of illumination (White et al. 2007). This suggests that cormorants are visually guided in their foraging, but constrained in the range of ambient illumination over which foraging can occur.
This constraint of foraging behaviour by ambient light levels in Great Cormorants is supported by observations in other diving avian predators. Thus, Little Penguins Eudyptula minor forage with reduced success and engage in prey pursuits less frequently at low light levels (Cannell & Cullen 1998, Ropert-Coudert et al. 2006b). Ambient illumination has no effect on the prey capture rate of Double-crested Cormorants, but only over the higher range of light levels (1.8–120 lux; Enstipp et al. 2007), at which visual acuity changes by only a small amount (White et al. 2007). Emperor Aptenodytes forsteri, King A. patagonicus, African Spheniscus demersus, Adélie Pygoscelis adeliae, Gentoo P. papua and Chinstrap P. antarctica Penguins respond to diel variation in ambient illumination by varying dive depth such that their deepest dives occur when ambient illumination is highest (Wilson et al. 1993, Kooyman & Kooyman 1995). Similarly, the present study also shows that Great Cormorants modulate their diving depth during the day, and that this may correspond with reduced visual resolution that would result when diving at the lower ambient illumination levels of early morning and at night.
The relationship between dive depth and duration in Great Cormorants is similar in the dawn and night bouts of the day, when light levels are relatively low. However, these relationships are different from the second bout of the day, when light levels are relatively high. These depth–duration relationships differ such that dives of a given depth are relatively long when light levels are low, and dives of a given duration are relatively shallow when light levels are low (Fig. 2). For example, dives of 10 m depth have an average duration of 35 s in the dawn (lower light level) bout of diving, and 28 s in the daytime dive (higher light level) bout, while a 60-s dive reaches an average depth of 18 m if made around dawn and 23 m if made later in the day. Night-time dives to 10 m have a similar length to those around dawn, but dives longer than 60 s never occur at night. Similarly, if only dives of 35–45-s duration are considered, night-time dives are shallowest (least-squares mean depth ± se = 5.5 ± 1.2 m), dives during the dawn bout are intermediate (8.3 ± 1.2 m) and dives in the daytime bout are deepest (11.4 ± 1.2 m). Thus, for a given duration, birds forage deeply when ambient illumination is high (i.e. during the day), and at shallower depths when ambient illumination is low (i.e. at night; Fig. 6).
Other factors that may affect diving behaviour
Although we argue that the positive relationship between dive depth and ambient illumination provides strong general support for the hypothesis that vision plays a role in cormorant foraging, there could be other explanations for the diel variation in foraging behaviour. We discuss these possibilities, before briefly examining the role of vision in the predatory behaviour of cormorants.
Diel variation in diving effort
It seems likely that the difference between dive bouts does not arise due to differences in motivation or effort. The dive/pause ratio, which is an index of diving effort (Dewar 1924, Grémillet et al. 2003), does not differ between the dawn and daytime bouts (Fig. 4), despite the difference in the relationship between dive depth and duration (Fig. 2). It therefore seems most plausible that Great Cormorants forage at shallower depths when light levels are low.
Diel variation in prey behaviour
The differences in cormorant diving behaviour between dive bouts could be the result of changes in the behaviour of their prey. However, we argue that this is unlikely to underlie the pattern of Great Cormorant behaviour that we describe here. At our Greenland study site, Cormorants forage mainly on sculpin Myoxocephalus spp. throughout the year (Grémillet et al. 2001, 2004, 2006, Lilliendahl & Solmundsson 2006). Sculpin are a group of mainly benthic fish that are diurnally active predators of benthic prey (Nickell & Sayer 1998, Norderhaug et al. 2005). Sculpin represent 60–80% of Cormorants’ diet by mass, and 40–60% by number, even when other potential prey such as Capelin Mallotus villosus are abundant (Grémillet et al. 2004, 2006). Myoxocephalus scorpius are found throughout the depth range utilized by Cormorants (up to 33.2 m; Kato et al. 2006, Ropert-Coudert et al. 2006a), and also at much greater depths (0–451 m and commonly 20–80 m; Fedorov 1986, Cardinale 2000). Given their wide depth distribution and diurnal habit, it is possible that the movement of benthic Sculpin to shallow water at night can account for the diel variation in dive behaviour of Cormorants. However, we suggest that Cormorants are primarily constrained to feed on Sculpin at shallow depths (up to c. 10 m) at night by low levels of ambient illumination. During the day, when ambient illumination is higher, they are able to dive to the greater depths where Sculpin are more common.
Diel variation in thermoregulatory costs
Subtidal water temperatures are known to vary by up to 4–5 °C with diel and tidal cycles (e.g. Kaplan et al. 2003, Magill & Sayer 2004) and such changes in water temperature could influence diving behaviour through the effect of temperature on diving metabolic rate. However, although temperature has a clear effect on diving metabolic rate in cormorants, the difference between resting and diving is typically greater than the difference associated with changing temperatures (Grémillet et al. 2001, Enstipp et al. 2005, Enstipp et al. 2006). Furthermore, in Greenland, Great Cormorants dive in water with a temperature of around 5 °C in summer and below –1 °C in winter (Grémillet et al. 2001), but we did not find a significant diel variation in water temperature during our study (Fig. 5). It seems unlikely therefore that increased thermoregulatory costs at night associated with a diel variation in water temperature can account for the pronounced diel variation in diving behaviour.
The role of vision in cormorant predatory behaviour
The possibility that sensory cues other than vision (e.g. tactile cues and olfaction) could also have a role in cormorant foraging has been discussed, but not investigated (Voslamber et al. 1995, Grémillet et al. 2005a). Furthermore, Harbour Seals Phoca vitulina have low aquatic visual acuity similar to that of cormorants (Weiffen et al. 2006) and they are known to be able to use vibrissae to provide touch and movement cues to guide their behaviour (Dehnhardt & Kaminski 1995, Dehnhardt et al. 1998). Nevertheless, even if information from such cues is available to foraging cormorants, the use of these cues cannot explain the association between diving behaviour and the level of ambient illumination. We suggest therefore that vision has a primary role in cormorant foraging despite their poor visual resolution at the light levels at which they are known to forage. We have argued elsewhere (White et al. 2007, Martin et al. 2008) that this apparent paradox is resolved by the hypothesis that cormorants primarily forage using close-quarter prey detection and capture, possibly involving a flushing technique in which they disturb prey on the substratum or within a natural refuge, and that they lunge (by rapid extension of the neck) at escaping prey.
We are grateful to Jon Green for helpful discussions, and to John Quinn and four anonymous referees for their detailed suggestions that greatly improved this paper. This research was funded by the Natural Environment Research Council (Grant NER/A/2003/00542), the French Polar Institute Paul-Emile Victor (Grant 388), and the Centre National de la Recherche Scientifique.