Dive data obtained in this study were similar to the values reported in previous studies (Le Boeuf et al. 1988, 2000; Robinson et al. 2012) and all seals returned to the colony on a normal schedule with typical mass gain. This suggests that attachment of the KKLs to the lower jaws may not have a significant effect on overall foraging success.
Thresholds and JME Count
Validation using the raw acceleration data suggested that JME decreased exponentially as the threshold level increased. This indicated that the very low thresholds (0·1 and 0·2 g) may include signals from non-feeding behaviour and signals from capturing very small prey. While separation of the feeding signals from noise is difficult at present in this study, we classified the signals ranging between 0 and 0·2 g as miscellaneous signals, (Fig. S2, Supporting information). The difficulty in classifying small JME might have been caused by the signals from very small prey, such as crustaceans, for which a quick strike capture motion may not occur. Further refinement of the KKL threshold will require simultaneous video data in the mesopelagic environment.
The JME algorithm used an interval threshold of 2 s to separate feeding events, assuming a distinct feeding event would not occur within 2 s of a previous event. Therefore, we validated our 2-s threshold on the raw acceleration data by examining the intervals between all peaks (>0·3 g) from the full-resolution data set. Most peaks had disappeared after a 0·5–1 s interval, suggesting that this interval would largely exclude repetitive peaks. We also compared the number of JMEs detected using a variety of interval thresholds. This comparison showed that the JME count using the 2-s threshold was 56·4–62·2% of the value for the 0·5-s threshold and 75·8–79·5% of the value for the 1-s threshold (Fig. S3, Supporting information). Therefore, we suggest that the number of JMEs counted by the 2-s and 0·3 g thresholds was less effective for small prey, particularly prey in a dense swarm, and represents a conservative estimate of JMEs.
The areas where FNES were most successfully foraging were already identified by examining the foraging-related metrics such as changes in drift rate and transit speed from a large sample (297) of FNES tracked from 2004 to 2010 (Robinson et al. 2010b, 2012). The area where FNES had the greatest change in drift rates, that is, lipid gain, was the area between the subarctic and subtropical gyres of the north-east Pacific Ocean (Robinson et al. 2012). Importantly, this is the same region where we observed the greatest number feeding events (JMEs). Interestingly, the seal with the largest mass gain (73 kg) had a low JME rate (Table 1), suggesting that this individual was feeding on larger or more energy dense prey. While it is clearly understood by the JME observations that the north-east Pacific gyres provide FNES an adequate amount of prey to gain the body mass necessary for recovery from the loss during breeding at the colony, it is also likely that, at a coarse scale, the prey were distributed rather evenly in the broad area of mesopelagic zone of this region, as shown by steady daily foraging throughout much of the migration. However, prey distribution and prey type likely vary on a local scale because the JME rate, body mass gain and migratory paths differed across seals.
Prey Type, Diel Pattern
Although several dive profiles, and their associated functions, were discussed, that is, transit, foraging, and resting (Le Boeuf et al. 1988; Crocker, Le Boeuf & Costa 1997), we classified dives into only two dive types, JME dives and non-JME dives. It is clear that non-JME dives were not performed with foraging intention but with other purposes, that is, digestion, rest and sleep (Crocker, Le Boeuf & Costa 1997; Mitani et al. 2010) because the dive profiles distinguished this dive type (shallow dive depth with slow descent and ascent rate) from JME dive type. JME dives occurred at a very high rate (80–91%) throughout much of the migrations. This suggested that the deep diving behaviour of elephant seals is an adaptation to exploit prey resources in the mesopelagic zone. However, the dominant prey species have yet to be adequately defined.
To assess the prey of FNES, we used three lines of evidence: inferred feeding mode, number and rate of JME, and photographs from a seal head-mounted camera. The distinctive spiky patterns observed in the raw acceleration data were short (<0·5 s) as reported in a captive seal and toothed whales (Bloodworth & Marshall 2005; Marshall, Kovacs & Lydersen 2008) suggesting that FNES use suction-feeding to capture small prey. Suction feeding by northern elephant seals has been confirmed in captive feeding experiments (Y. Naito & H. Louis, unpublished data). Most of the JME observed in the FNES were composed of a single peak. In contrast, predators feeding on large prey items are likely to use biting and mastication movements, which appear as a repetitive peak-acceleration pattern (Liebsch et al. 2007; Suzuki et al. 2009). Dental morphology studies also suggested that elephant seal teeth do not function in procurement or mastication (Abbott & Verstraete 2005) and that prey is swallowed whole (Antonelis, Lowry & Fiscus 1987; Antonelis et al. 1994), lending further support that FNES likely uses suction to feed on small prey.
We calculated an estimated prey size using field metabolic rates of similarly sized FNES [energy acquisition, 92 kJ min−1 (Costa et al. 1986)] and the mean caloric value of diel migrating prey species and caloric value of the mesopelagic prey [2·837 kcal g−1 wet weight of shrimp, squid and myctophid fishes (Benoit-Bird 2004)]. Using these values and the JME rates in this study (529·3–1117 JME per day), we estimated the mean prey size to be very small (9·9–21·1 g) with a daily consumption rate of 2·9–3·7% of body mass. Considering that the KKL probably did not detect feeding motion for small prey, our estimates of feeding rates is likely conservative and, therefore, the prey size could be even smaller. Conversely, our estimate assumed that all JME were successful prey capture events and, if so, the actual prey size estimate would be larger. Nevertheless, taking into account both of these potential sources of error, we suggest that seals forage on small prey items. Stable isotope analysis on female southern elephant seals, Mirounga leonina, indicates that their offshore prey probably consists mainly of small mesopelagic myctophids (<12 cm) (Cherel et al. 2008). Likewise, myctophids are potential prey for FNES in this study. The images from the seal-head cameras further support the consumption of small prey (Fig. 4).
Female northern elephant seals are known to dive on a diel schedule (Kuhn et al. 2009; Robinson et al. 2012). Using data from a long-term jaw-motion recorder, we confirmed that their diel diving pattern was caused by their foraging activity. Our JME data showed that seals adopted different foraging strategies by day and night, primarily foraging in the ‘deep scattering layer’, which consists of small organisms at 400–800 m (Barham 1966; Robinson et al. 2012). The larger number of JME at (57·1–61%) may suggest that FNES are primarily nocturnal foragers. Many pelagic foragers adopt this nocturnal foraging strategy, for example, Antarctic fur seals (Arctocephalus gazella). However, the daytime irregular or bimodal pattern of JME suggests that FNES employ unique foraging strategy during daylight hours. Why they adopt this foraging behaviour is unknown. Given the dives of tagged FNES were deeper with less bottom time, it is likely that they are foraging less efficiently compared to at night. These observations are consistent with the bi-modal diving pattern reported for FNES (Le Boeuf et al. 1988). Notably, very deep JME dives exceeding 900 m (V-shaped dives, referred to here as ‘big dives’, Fig. 1b) occurred only in the daytime. The occurrence of these big dives (63 dives for four seals, mean dive depth: 998 ± 63 m, mean number of JME: 253, mean JME depth: 982 ± 39 m) suggests that seals may search for alternative prey during the daylight hours. Daytime shallow foraging on different prey types is also possible. However, how prey type selection occurs in relation to depth is uncertain. We currently lack an adequate understanding of potential prey items, and this area should be a target for future studies.
Comparison with Deep Diving Toothed Whales
How does the foraging pattern of elephant seals, a visual predator, compare to an echolocating toothed whale? The hallmark of these deep diving toothed whales is their ability to produce high-frequency clicks or biosonar to find prey in the dark as well as assess the prey field. We compared the JME rate of FNES (529·3–1117 JME per day) with the buzz rate of pelagic foraging toothed whales. The echolocation buzzes of toothed whales are considered to be feeding attempts (Johnson et al. 2004). The buzz rate (buzz per day) was calculated from the number of buzzes per dive and number of dives per day of toothed whales (312, 345, 320–360, 25·2–80·4 buzz per day) for Cuvier's beaked whale Ziphius cavirostris, Blainville's beaked whale Mesoplodon densirostris, sperm whales Physeter macrocephalus and short-finned pilot whales Globicephala macrorhynchus, respectively (Miller, Johnson & Tyack 2004; Tyack et al. 2006; Aguilar Soto et al. 2008). These data suggest that toothed whales make fewer feeding attempts than northern elephant seals. Considering the far larger body mass of toothed whales, the lower rate of feeding attempts suggest that toothed whales are capable of finding and routinely capturing much larger prey or prey with a higher energy density. These simple comparisons suggest that FNES are less selective and employ a less efficient feeding mode than toothed whales in terms of successful prey capture events per dive. This is not surprising when we consider that toothed whales employ ‘biosonar’ clicks and buzzes, that allows them to identify prey many meters away in total darkness and also provides them information on the size and composition of the potential prey (Arranz et al. 2011). In contrast, the range over which FNES may sense and identify potential prey is much smaller. They may thus be limited to taking whatever prey they encounter, and to forage in areas where prey are numerous but of small size. We hypothesize that the continuous diving mode by FNES could be attributed to their reliance on small prey and their less efficient ‘passive sensors’ for prey search, that is, their use of vision to detect luminescence (Levenson & Schusterman 1999) and whiskers to detect the vortices shed by swimming prey (Miersch et al. 2011).
Our findings partly explain the paradox of deep diving behaviour among some species of marine mammals and the diversity of this behaviour. However, we still require precise knowledge of their prey species and of the deep water community for a more complete understanding.