Linking vertical movements of large pelagic predators with distribution patterns of biomass in the open ocean

We analyzed data from a total of 344 electronic tags that yielded 3D movement data spanning 46,659 tracking days from 12 species of bony and cartilaginous fish predators in the North Atlantic Ocean (SI Appendix, Table S1 and Fig. 1A), including sharks (white, Carcharodon carcharias; basking, Cetorhinus maximus; tiger, Galeocerdo cuvier; shortfin mako, Isurus oxyrinchus; porbeagle, Lamna nasus; whale, Rhincodon typus; blue, Prionace glauca); rays (Chilean devil, Mobula tarapacana); tunas (bigeye, Thunnus obesus; yellowfin, Thunnus albacares); and billfishes (blue marlin, Makaira nigricans; swordfish, Xiphias gladius). Individual tracks averaged 134 d (range of mean durations = 62 to 231 d from 11 to 86 tags per species). Considering only 2D movements, individuals cumulatively traversed 1.55 million km of the pelagic ocean averaging 4,600 km per individual (range of species-specific mean track distance = 2,074 to 10,318 km; SI Appendix, Table S1).

The tag data indicated that the deep pelagic ocean is an important habitat for all species tracked in this study. Maximum recorded depth exceeded 900 m for all species and 1,500 m for seven of the 12 species (SI Appendix, Table S2). The among-species mean of daily maximum depth indicated widespread use of the deep pelagic ocean throughout the North Atlantic (overall mean 914 m, s.d. 420 m; Fig. 1B). Maximum depths (SI Appendix, Fig. S1) and daily occupation of the deep ocean (Fig. 2 and SI Appendix, Fig. S2) revealed, however, that vertical habitat use was variable among species, ranging from $>$60% daily residency at depth (deeper than 200 m) for basking sharks to $<$5% for blue marlin, tiger sharks, whale sharks, yellowfin tuna, and Chilean devil rays. Despite the variability across taxa, all species exhibited examples of individuals spending at least 50% of the day in the mesopelagic, and six species had examples of days for which 100% of time was spent deeper than 200 m (SI Appendix, Table S2).

High-resolution time series data (1 to 600 s resolution) were available for a subset of 219 tagged individuals representing 13,330 tracking days ($\sim$30% of total days; SI Appendix, Table S2). The time series data captured 11,922 complete deep dive profiles in which an individual was recorded moving below the epipelagic ($>$200 m) before returning to the epipelagic after a period of time. The number of deep dives per day averaged 0.4 to 4.2 across the study species while individuals made as many as 6 to 24 of these dives in a single day (SI Appendix, Table S2 and Fig. S3).

A hierarchical generalized additive model found alignment of vertical habitat use with the predicted depth of the primary deep scattering layer (DSL, Fig. 3 and SI Appendix, Table S3) as modulated by local oceanographic conditions (represented by $\mathrm{\Delta }$T, change in temperature between 220 m and the surface) and species-specific physiology. The model also highlighted significant variability among individuals within and across species and an apparent disconnect between predators and the DSL for other species. To quantify the relationship of the observed vertical habitat use relative to the expected depth of the DSL, we developed the metric $Z^{\prime }$ (Eq. 1) where values of $Z^{\prime }$ significantly $>$ 0 (blues in Fig. 3B) or $<$ 0 (reds in Fig. 3B) indicate depth of tagged individuals significantly deeper or shallower than the DSL, respectively. For example, despite some use of deeper waters, blue marlin and tiger sharks primarily occupied the upper mesopelagic and did not regularly intersect with expected DSL depths (95 and 79% of days with $Z^{\prime }$$<$$-$200 m, respectively; red colors in Fig. 3B). In contrast, swordfish, shortfin mako sharks, and white sharks all appear to track DSL depth (51, 41, and 53% of days with $Z^{\prime }$$\pm$ 100 m, respectively), largely independent of temperature. For blue and basking sharks, depth preferences tracked DSL depth but were modulated by thermal constraints whereby tagged individuals moved deeper as $\mathrm{\Delta }$T decreased. Basking sharks were consistently observed well below expected DSL depths (52% of days with $Z^{\prime }$$>$ +200 m) when the DSL was located $\sim$400 m or deeper.

To further explore how scattering layers might influence predator behavior, we collocated the tag dataset to 361 research expeditions on vessels equipped with active acoustics systems operating at frequencies of 75 kHz or lower, allowing them to insonify the DSL. The vessel data represented 3,892 d at sea, spanning much of the North Atlantic (Appendix, Fig. S4), and resulted in $>$290,000 matches where a ship and a tagged predator were within 100 km and $\pm$30 d of each other. We selected four representative fish-ship matches (for different predator species) to highlight the opportunities and inherent challenges in the analysis of these data (Fig. 4). The blue shark example suggested that the tagged individual was aligned with the depth of the migrating mesopelagic community ($\sim$80 to 100 m) at night. The shark’s daytime behavior suggested occupation of a portion of the water column that contained a scattering layer that was acoustically weaker than the primary DSL below. Tag-derived detections of bioluminescence events indicated the presence of potential prey (Fig. 4A and Materials and Methods). In contrast to diurnal variability observed in the blue shark example, data from a tagged swordfish highlighted tight coupling between the predator’s vertical habitat use and migration of the primary DSL from $\sim$500 m during daytime to the surface mixed layer at night (Fig. 4B). The dive profile recovered from a tagged basking shark revealed considerable time spent below the effective range of the shipboard acoustic instrument during the day, but a nighttime portion of the example suggests: 1) nighttime overlap with the migrating DSL in the near-surface; 2) potential repeated interaction with a non-migrating deep scattering layer at night around 450 m; and 3) additional overlap with a patchy shallow scattering layer at $\sim$250 m (Fig. 4C). It is worth noting that the basking shark example was hindered by strong noise in the acoustics data (light colored region centered around 18:00) which can result from changes in weather conditions and interference from other onboard instrumentation, highlighting a primary challenge in making inference from shipboard acoustic instrumentation. Finally, the shortfin mako shark example identified movement throughout the upper mixed layer at night while overlapping with the migrating component of the DSL. During daytime, the shark was primarily located in the surface ocean but made multiple, relatively rapid dives that terminate at the top of the primary DSL in this region ($\sim$300 m and Fig. 4D).

The deep ocean provides a wide range of vital ecosystem services to the planet and society (5) but remains largely unexplored and enigmatic. Recent research initiatives have sought to improve our understanding of deep ocean ecosystems amid international calls to action (24, 27). However, these efforts have often overlooked links between midwater animals and a group of large pelagic predators that are targets of large commercial fisheries, bycatch, or species of conservation concern. While nearly all species studied here are considered primarily epipelagic, our results demonstrate that all predators occupied the deep pelagic ocean to some degree, yet some did so less frequently or spent less time at depth. The density of prey in some deep scattering layers (28) suggests the potential for highly efficient foraging that could skew the relationship between time spent at depth and the relative importance of mesopelagic foraging. Indeed, studies of catch-at-depth for many of the species evaluated here indicate high catches in deep waters where tag records suggest that minimal time is spent (29); however, additional work is needed to quantify the relative importance of deep versus shallow foraging modes.

Our results demonstrate that the deep ocean is a critical habitat for large pelagic predators despite their range of physiological capabilities and limitations to accessing the ocean twilight and midnight zones. The ocean’s twilight zone, in particular, is typified by a strong primary DSL that may contain 10 to 100x more biomass than annual global fishery landings (2) and includes a number of important prey species, such as lanternfishes and squids, that are consumed by most pelagic predators (30). Taken together, our results suggest the frequent occupation of the deep pelagic ocean by several predator taxa in this study is likely a function of the high biomass in this zone and is modulated by overlying oceanographic conditions for which some species have specialized adaptations (e.g., endothermic capacities, SI Appendix, Table S1). For example, swordfish are well known to closely track diel vertical migration of the DSL and spend nearly all daylight hours at depth (31–33). These vertical movements are facilitated by a number of adaptations that make them a formidable predator in the dark, cold environment of the deep pelagic ocean, e.g., thermal plasticity, (34); visual acuity, (35). Bigeye tuna and blue sharks also spent significant time at depth but exhibit multiple, shorter dives per day. While bigeye tuna can also alter whole-body heat transfer, their time at depth appears limited by abrupt termination of dives when muscle temperature decreases to $\sim$17 ${}^{°}$C (36). As an ectotherm, blue sharks face clear thermal constraints to the length of time that they can spend in the twilight zone (37, 38). The inability to maintain elevated internal temperature likely contributes to the observed gradient of increasing time at depth from north to south in our study area as water temperatures also increase at depth (decreasing $\mathrm{\Delta }$T) as one moves closer to the equator, potentially releasing them from thermal constraints to foraging (15). Similarly, low oxygen at depth in the eastern Atlantic may limit the vertical extent and/or time at depth of many predator species relative to the normoxic conditions elsewhere in the North Atlantic (39, 40). Together, the observed patterns of deep ocean habitat use are both consistent with previous findings (18, 41, 42) and suggest that the vertical movements of pelagic predators are responding to variability in the distributions of mesopelagic prey in the open ocean.

Our results provide strong, but indirect, evidence for foraging as a key driver of deep ocean habitat use. These data are consistent with observations of active foraging at depth by some deep-diving marine mammals (17, 43). However, previous work has also pointed to other potential functional roles of these vertical movements, including predator avoidance (44) and navigation (45), among others (reviewed in ref. 18). For example, our model results highlight an apparent disconnect between predicted location of the DSL and predator dive depths for devil rays and whale sharks and significant variability in this relationship for basking and porbeagle sharks that both exhibited prolonged periods of deep ocean residency. This mismatch may be partially explained by the diet of filter-feeding predators (e.g., basking and whale sharks) that target zooplankton as these prey are not apparent in the low-frequency acoustics used to study deep scattering layers. In addition, a number of large pelagic predators, including whale sharks and mobulid rays, are known to use passive glides to descend, interspersed with active ascents, to optimize movement efficiency (46–49), which may represent a distinct functional role for occupying the deep ocean. Theoretical and empirical evidence has shown that this passive gliding, particularly in negatively buoyant animals, can result in significant energetic advantages (50). Furthermore, previous studies have indicated thermoregulation both as a modulator of and motivation for occupying the deep ocean for certain species in specific oceanographic regimes. For example, porbeagles are known to overwinter at depth in warm Gulf Stream water (51) and Atlantic bluefin tuna may exploit deep pelagic habitats to reduce energetic cost of swimming against strong currents and to reduce excess heat during spawning migrations in the (sub)tropical Gulf of Mexico (52). The complex relationship between predator behavior and expected DSL depths for these species demonstrates the need for additional work to disentangle the relative influence of foraging and thermal preferences and adaptations, along with their spatiotemporal variability.

While most of the study species showed correlations between vertical movements and DSL depth, we found no evidence for a link between DSL location and dive behavior in blue marlin and tiger sharks. These species are likely limited in their thermal and/or visual capabilities at depth (33, 53, 54) and only occasionally occupy the upper mesopelagic or conduct relatively short “bounce” dives to deeper habitats (e.g., refs. 55 and 56). Some research has suggested that species facing significant physiological limitations and/or exhibiting frequent surface occupation during daytime may rely on the nighttime migration of scattering layer prey into the epipelagic (56–58). This interpretation is supported, in some cases, by mesopelagic prey items in the diet of truly epipelagic species, e.g., shortbill spearfish: (59, 60). However, recent work suggests that some predators may also target shallow scattering layers in the open ocean during the day (61, 62), which may be a more common strategy among surface-oriented, or otherwise physiologically limited, pelagic predators than previously recognized.

Scattering layers are often coherent patches of fish and invertebrate nekton in the pelagic ocean that occupy specific light “comfort zones” (7). Most research to date has focused on the primary DSL, which is largely thought to be a ubiquitous feature throughout the global ocean but with a number of complexities, including multiple modes in the “primary” DSL (63) and varying proportions of migrating and non-migratory individuals and migration timing (64). This complexity likely contributes to the strikingly divergent predictions of a very simple DSL depth metric in this study derived from three published models (SI Appendix, Figs. S5 and S6). Our mean representation of this variability in DSL depth captured a general characteristic of the phenomenon. However, our model remains naive to variations in biomass, spatial, and temporal patchiness and differing community composition within the DSL and among other scattering layers, which are all too poorly constrained to use for this work. Nevertheless, a comparison between the geometric means of the predicted DSL for the case studies (shown in Fig. 4) shows that they differ from local maxima in acoustic backscatter by only 1 to 86 m (SI Appendix, Fig. S6 and Table S4). This comparison, though, is limited to the depths that can be observed with 75-kHz acoustics. For example, in addition to shallow, transient scattering layers and the primary deep scattering layer, there is also growing evidence for additional DSLs at the meso-to-bathypelagic boundary around 1,000 m (3, 65), consistent with some of the vertical habitat use of basking sharks shown here. Furthermore, some top predators are known to make extreme deep dives well below the primary DSL to $>$2,000 m, e.g., elephant seals, (66); beaked whales, (67); Chilean devil rays, this study and (68). While we lack an estimate of global biomass in the bathypelagic, recent work found higher biomass than in mesopelagic waters above the Mid-Atlantic Ridge (3) and others found that biomass may be significant even into the abyssal pelagic to $\sim$4,000 m, (69). Similarly, the global distribution and biogeography of shallow ($\sim$150 to 350 m) scattering layers remain poorly understood, highlighting how little we know about the ocean’s midwater ecosystems and the limitations of current approaches.

Regardless of these complexities, our finding that nearly all species studied here frequented the deep pelagic ocean shows that this zone may provide underappreciated ecosystem services to a suite of commercially and ecologically important pelagic fish species. We were unable to directly quantify foraging at depth, and therefore, the absolute value of deep resources to these predators remains unknown. However, our results suggest that mesopelagic forage species contribute significantly to the energy budgets of pelagic predators and highlight the need for additional research to better characterize the ecological value of this connectivity, particularly in light of climate-induced changes to midwater ecosystems (70, 71) and planned harvest of mesopelagic resources (72, 73). Despite extensive diet data supporting the importance of mesopelagic prey for a suite of predator taxa [reviewed in ref. 30], the relative importance of deep versus shallow foraging modes within and among taxa remains largely unknown. Uncertainties in the relative importance of pelagic prey have led to the continued discounting of large marine predators as mediators of carbon export via the biological pump or, if considered, have been assumed to be primarily epipelagic foragers [reviewed in ref. 74]. Our results show that many predators engage in regular vertical migrations of hundreds to thousands of meters between the surface and the ocean twilight and midnight zones. Better accounting is needed to more accurately quantify the potentially significant role of predators and the impacts of associated fisheries for predators (75) and prey (72, 73) in the marine carbon cycle. Together, the overlap in ongoing fishing effort and pelagic predator distributions (76), expected climate-induced changes in pelagic ecosystems (70, 71, 77), and the potential extraction of mesopelagic biomass (72, 73) suggest that this critical link may be in jeopardy.

We thank J. Collins for discussions about the marine carbon cycle, A. Chase for help with optical oceanography, and L. Gallagher (Fishpics) who created the species art, and a number of individuals who contributed to fieldwork. This work was funded in part by The Coastal Research Fund in Support of Scientific Staff and the Investment in Science Fund at the Woods Hole Oceanographic Institution (WHOI) (to C.D.B.), the WHOI President’s Innovation Fund and Postdoctoral Scholar Program at WHOI with funding provided by the Dr. George D. Grice Postdoctoral Scholarship Fund (to M.C.A.), United Kingdom Natural Environment Research Council (to D.W.S.), the European Research Council (to D.W.S.), a Marine Biological Association Senior Research Fellowship (to D.W.S.) and the King Abdullah University of Science and Technology (baseline research funds to M.L.B.). B.C.L.M. was supported by the projects IslandShark (PTDC/BIA-BMA/32204/2017), AEROS-Az (ACORES-01-0145-FEDER-000131), MEESO (EU H2020-LC-BG-03-2018), and Mission Atlantic (H2020-LC-BG-08-2018-862428). This work was part of the Woods Hole Oceanographic Institution’s Ocean Twilight Zone Project, funded as part of the Audacious Project housed at TED. N.Q. was funded by Fundação para a Ciência e a Tecnologia (FCT) PTDC/BMA/3536/2021 and CEECIND/02857/2018.