From the polar night to the midnight sun: light and polar cod DVM in ice-covered Franklin Bay
In ice-free waters, the DVM of Arctic zooplankton are strong during the succession of day and night in spring and fall, and weak or absent during the polar night and the midnight sun (Buchanan and Haney 1980; Fischer and Visbeck 1993; Falkenhaugh et al. 1997; Blachowiak-Samolyk et al. 2006; Berge et al. 2009). The weakening of ∆I/I, the zeitgeber that synchronizes the migration, explains the weakening of the DVM during the polar night (e.g., Berge et al. 2009). Continuous irradiance and the resulting lack of night-time refuge against visual predators likewise explain the cessation of DVM in ice-free waters during the midnight sun (Hays 1995; Fortier et al. 2001). In ice-covered waters, the snow and sea-ice cover can modify the seasonal pattern in DVM by attenuating underwater light and modifying the intensity of ∆I/I. For example, in Svalbard during the polar night, the dimming of light to almost imperceptible levels stopped the DVM of zooplankton in ice-covered Rijpfjorden, while the migrations continued in ice-free Kongsfjorden (Berge et al. 2009). As well, in Barrow Strait (74.5°N) in summer, sea-ice attenuated irradiance at night below the reported threshold of visual perception of fish predators, and copepods maintained their DVM under the midnight sun (Fortier et al. 2001).
In the present study, seasonal changes in the duration of polar cod DVM closely tracked the lengthening of the photoperiod from the continuous polar night in December until the midnight sun in May. Even at the pinnacle of the polar night on 22 December, the weak upward and downward movements of the scattering layer in the 0–100 m layer were precisely synchronized with the theoretical maximum rate of change in twilight intensity (∆I/I) at 70°N. A first interpretation of this synchrony is that polar cod perceived variations in whatever little downwelling light penetrated the snow and ice cover and reached to 100 m depth. At 70°N in late December, civil twilight at the surface (during which objects can be distinguished by the human eye) prevails for a period of 4 h centred on local noon. In Amundsen Gulf (70°N to 72°N) in 2007–2008, maximum incident PAR during civil twilight raised slowly from 0.5 μmole photon m−2 s−1 on 26 December to 1 μmole photon m−2 s−1 on 5 January and increased exponentially afterwards to reach 7 μmole photon m−2 s−1 on 21 January (T. Papakyriakou, University of Manitoba, unpublished data). The minimum thickness of the snow and ice covers were 0.07 and 0.8 m, respectively, around the overwintering station in December 2003 (Langlois et al. 2006). Using the minimum light attenuation coefficients of snow and sea-ice (4.3 and 1.1 m−1, respectively, e.g., Perovich 1996) and the K
d for water measured in April (0.11 m−1), an incident PAR of 0.5 μmole photon m−2 s−1 in late December would have been attenuated to 2.3 × 10−6 μmole photon m−2 s−1 at 100 m depth. The same calculation with the maximum coefficient of attenuation of snow (40 m−1) and ice (1.5 m−1) yields a value of 1.4 × 10−7 μmole photon m−2 s−1 at 100 m depth. These rough estimates of the maximum and minimum light reaching 100 m around noon in late December bracket the 1.9 × 10−7 μmole photon m−2 s−1 threshold to which juvenile walleye pollock (Theragra chalcogramma) respond in the laboratory (Ryer and Olla 1998). Hence, it is conceivable that dark-adapted polar cod possessed the visual acuity to detect the weak solar signal that reached at 100 m in December in Franklin Bay. An alternative explanation is that the precise timing of the DVM of polar cod on maximum ∆I/I in late December was a remanence of an earlier synchronization to the stronger light signal with similar period that prevailed before the setting of the polar night in December.
Based on measured light penetration in April and May, migrating polar cod avoided moving above the 10−6 μmole photon m−2 s−1 isolume (Fig. 12). This threshold is two orders of magnitude lower than the ~1.89 × 10−4 μmole photon m−2 s−1 level at which the feeding of polar cod became significantly impaired in the laboratory (Girsa 1961 cited in Blaxter 1970), which suggests that the feeding of polar cod moving into the cold intermediate layer at night may have been limited by light. However, after several months of adaptation to nearly absolute darkness at depth under the snow and ice cover, polar cod may achieve higher visual acuity than measured under laboratory conditions. Alternatively, when stimulated mechanically or by turbulence, bioluminescent organisms such as Metridia longa (e.g., Lapota et al. 1989) can provide substantial background illumination that can be used by predators (Widder et al. 1992). Within the extremely dense aggregation at depth in Franklin Bay (Benoit et al. 2008), turbulence from swimming fish and/or direct contact between fish and M. longa may have triggered the emission of the light needed by polar cod to detect its prey under the 10−6 μmole photon m−2 s−1 solar isolume. The light emitted by M. longa would make it particularly vulnerable to detection and capture, explaining the relatively high frequency of this copepod in the gut content of polar cod despite its small weight and lipid content and, hence, energy value. Most importantly, bioluminescence by M. longa may also have provided polar cod with the background illumination to detect the large and energy-rich but non-bioluminescent Calanus hyperboreus and C. glacialis.
While the synchronized migration of the scattering layer often stops during the midnight sun, individual migrants may nevertheless perform unsynchronized migrations. For example, in a Svalbard fjord (79°N) in June, the vertical velocity of individual zooplankton (most likely Calanus finmarchicus and C. glacialis) indicated a continuous net downward movement in the surface layer and an upward movement in the deep layer, while the overall scattering layer exhibited no net movement (Cottier et al. 2006). In the present study, the vertical striation of the echogram under the midnight sun in late May suggests as well the rapid vertical migration of individuals or schools with no net synchronized movement of the main aggregation except for the avoidance of seals (Fig. 12b).
Prey, predators and the causes of polar cod DVM under the ice
Among various theories on the causes and trophic advantages of diel vertical migrations in the ocean, the complementary predator-evasion and hunger/satiation hypotheses are most popular (Lampert 1989; Hays 2003; Pearre 2003). Both hypotheses assume that (1) food is limited at depth and available in some shallower layer; (2) in daytime, vulnerability to visual predators is higher in the shallow layer than at depth; and (3) the energy gain of feeding in the shallow layer exceeds the energy expenditure of the migration (Gliwicz and Pijanowska 1988; Lampert 1989). Under these conditions, the maximum in ΔI/I at dusk prompts an upward migration, while the downward migration is triggered by satiation at anytime during the night or by maximum ΔI/I at dawn (Ringelberg 1995; Fortier et al. 2001; Sourisseau et al. 2008). According to the predator-evasion hypothesis, the need to avoid visual predators dictates the downward migration in the morning and the occupation of the deep layer during daytime. Increased risk of predation due to a large size or high visibility limits the amplitude and duration of incursions in the shallow layer at night and increases the depth of daytime residence (Giske et al. 1990; Fortier et al. 2001). According to the hunger/satiation hypothesis, the drive for the night-time upward migration should increase with hunger and only those individuals sufficiently hungry will migrate upward at night (see Pearre 2003 for a review).
Under the ice of Franklin Bay in winter and spring, the greater density of copepod prey in the deep layer (140–225 m) than in the lower part of the cold intermediate layer into which some polar cod migrated at night (100–140 m) seemed to contradict the assumption of food limitation at depth and food availability above. However, the precise assumption is that food limitation prevails at depth for those fish that migrate, not necessarily for all fish. With their heavy livers and heavy gut content relative to body weight, polar cod >25 g showed no sign of food limitation and did not migrate into the intermediate layer (Fig. 6). By contrast, the polar cod that migrated into the intermediate layer were small fish (<25 g) with small livers and gut contents and, perhaps, lesser visual acuity than their larger congeners (Fernald 1990). We suspect that a limitation of food intake due to foraging interference with large polar cod in the extremely dense aggregation at depth drove the upward migration of small polar cod into the cold intermediate layer where the ratio of copepod prey to fish was more favourable. That polar cod feed more successfully outside than inside schools (Hop et al. 1997) supports this interpretation.
Consistent with the hunger/satiation hypothesis, only a small fraction of the polar cod population migrated upward on a given day. Digestion in polar cod is slow, and a total evacuation time of about 400 h (ca. 17 days) has been reported (Hop and Tonn 1998; Saether et al. 1999). Hop and Graham (1995) suggested that polar cod feed infrequently in winter, just enough to minimize weight losses. Hence, small polar cod could have remained at depth for several days between feeding forays into the intermediate cold layer, which would explain the overall low fraction of the population participating in the migration on any given night. If so, the observed migration behaviour of polar cod would represent an extreme case of DVM in which the vertical interchange of individuals actually occurs on a time scale of several days. Under similar conditions of sub-zero temperature and low light availability in the Southern Ocean in winter, Antarctic cod (Notothenia coriiceps) enters a state of dormancy interrupted by awakenings of a few hours every 4–12 days, during which metabolic activity builds back to summer levels (Campbell et al. 2008). The intriguing possibility that small polar cod adopt a similar hibernating strategy, remaining in a state of torpor interrupted by fortnightly feeding incursions into the intermediate layer, could be tested using acoustic tags as in Campbell et al. (2008).
Studies of DVM in the ocean generally assume that the deeper distribution of migrants in daytime is an escape response to visual predators prowling the photic layer. Yet, these visual predators and their predatory impact are seldom identified and quantified. The one and apparently unique predator of polar cod under the ice of Franklin Bay was the ringed seal. Under coastal landfast ice in winter and spring, ringed seals dive beyond 200 m and maximum diving depth often corresponds to the bottom (Lydersen and Hammill 1993; Kelly and Wartzok 1996). In the nearly absolute darkness prevailing at depth under the ice in winter, ringed seals use their highly sensitive vibrissae to detect and capture their prey (Hyvärinen 1989). Dive duration and depth are strongly correlated to body mass (Kelly and Wartzok 1996; Teilmann et al. 1999; Kunnasranta et al. 2002). In the North Water (northern Baffin Bay) in winter and spring, sub-adults equipped with acoustic tags dove only to shallow depth, while an adult performed deep dives under the ice (Born et al. 2004). Born et al. (2004) surmised that immature seals preyed upon small polar cod and the hyperiid amphipod Themisto libellula in the surface layer, whereas adult seals preyed on large polar cod at depth.
Both large mature and small immature ringed seals occupied the moon pool of the ship and, consistent with the scenario of Born et al. (2004), two modes existed in the distribution of dive depths in April: relatively shallow (20–130 m) dives presumably by small seals and deeper dives (130–230 m) by large seals (Fig. 9). In agreement with the observations of Kelly and Wartzok (1996), ringed seals dove by night and by day, and no clear relationship between dive frequency and light was observed. However, the average depth of the deep dives varied with the depth of the fish aggregation (Fig. 10), implying that the energy expended by mature seals to feed on large polar cod varied with the depth of their prey. Hence, on average, residing at depth must have provided polar cod with some protection against predation by ringed seals.
Differences in diving depth and in the prey targeted, respectively, by small and large ringed seals could explain the size-related differences in the migration behaviour of polar cod observed in the present study. In agreement with the prediction of reduced migrations and a deeper distribution of large conspicuous individuals (Giske et al. 1990; Fortier et al. 2001), large polar cod with disproportionately large livers remained at depth, avoiding moving above 180 m (Fig. 6). In the North Atlantic, grey seals (Halichoerus grypus) have been observed to bite off the belly of Atlantic cod (Gadus morhua) including the lipid-rich liver (Fu et al. 2001; Chouinard et al. 2005). In the present study, the livers of large polar cod brought back to the moon pool by large seals were often extruding from the torn belly of the fish (L. Fortier, personnel observation), suggesting that ringed seals similarly sought the lipid-rich liver. For large polar cod targeted by large deep-diving seals and detected by vibrissae rather than visually, moving above 180 m at night afforded no particular foraging advantage or refuge against predation. Accordingly, for those large fish, the best strategy would be to remain as deep as possible at all times so as to increase foraging costs to their predators.
By contrast, the fish migrating at night into the intermediate layer were small polar cod with relatively small livers and, presumably, reduced energy value for large competent seals that can reach the larger fish with larger livers at depth. These small polar cod performed DVM that were precisely synchronized with maxima in ΔI/I at dawn and dusk. Assuming that immature seals are less competent at detecting and capturing their prey in the dark than mature seals, the combination of improved feeding conditions due to reduced interference with large congeners and reduced risk of visual detection by small immature seals at night would drive the DVM of these small polar cod into the intermediate cold layer.
Reverse DVM (shallow in daytime, deep at night) has been documented in zooplankton (see Cohen and Forward 2009 for a review), but the causes and adaptive value of this behaviour are poorly understood (Tester et al. 2004). In the week before the final transition from the day/night cycle to the midnight sun in mid-May, polar cod exhibited a reverse migration (Fig. 5f). During this period of nearly continuous daylight and reduced snow cover, some light may have reached over most of the water column at all times except at depth during the night. Hence, a possible interpretation of the reverse DVM is that, with the disappearance of their refuge from visual predators at depth in daytime, polar cod moved upward to feed in the well-illuminated intermediate cold layer. During the nocturnal twilight, most fish returned below 120 m where dimmer light conditions still afforded some protection against predation by ring seals.
As expected, synchronized DVM stopped under the midnight sun in late May. The end of synchronized DVM coincided with a change in the structure of the fish aggregation, which gradually developed into a dense shoal of schools and ribbons. Along with DVM, schooling is a widespread anti-predation behaviour in fish (e.g., Pitcher et al. 1988; Pitcher and Parrish 1993). After 20 May, the lack of day/night difference in refuge from visual predators apparently triggered schooling as an alternative behaviour to reduce predation (Fig. 12b). Starting in mid-May, the shift from normal DVM to reverse DVM, and then to schooling in the cold intermediate layer at all time of day coincided with the resumption of feeding and the replenishment of liver reserves. Feeding efficiency is another advantage attributed to schooling, which increases detection volume compared to solitary foraging when searching for food concentrations, especially in a zooplankton-poor environment such as often encountered in the Arctic.
Finally, under the midnight sun in May, the vertical distribution and movement of polar cod clearly responded to seal forays in the deep layer (Fig. 12b). In summary, under the ice of Franklin Bay, the lack of DVM by large polar cod and the DVM of small polar cod in winter and spring, as well as reverse DVM and the schooling of polar cod under the midnight sun in May, were all consistent with a need to evade the different predation pressures by mature and immature ringed seals. Our observations suggest that, as in most trophic interactions in Arctic waters (e.g., Falk-Petersen et al. 2009), the quest for lipids dictates the foraging behaviour of ringed seals and, consequently, the observed size-related differences in the DVM of polar cod under the ice in winter and spring.