Journal list menu

Volume 65, Issue 4 p. 707-716
Free Access

Response of small sharks to nonlinear internal waves

Jesús Pineda

Corresponding Author

Jesús Pineda

Biology Department, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts

Correspondence: [email protected]Search for more papers by this author
Sally Rouse

Sally Rouse

Scottish Association for Marine Science, Oban, Argyll, UK

Search for more papers by this author
Victoria Starczak

Victoria Starczak

Biology Department, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts

Search for more papers by this author
Karl Helfrich

Karl Helfrich

Physical Oceanography Department, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts

Search for more papers by this author
David Wiley

David Wiley

National Oceanic and Atmospheric Administration/Office of National Marine Sanctuaries, Stellwagen Bank National Marine Sanctuary, Scituate, Massachusetts

Search for more papers by this author
First published: 22 October 2019
Citations: 4
Associate editor: Julia Mullarney


Plankton and nekton may respond passively or actively to large-amplitude, nonlinear internal waves (NLIW), with periods and wavelengths on the order of minutes and hundreds of meters, and the NLIW can cause direct or indirect changes in distribution. NLIW are ubiquitous in the coastal ocean, but understanding the influence of NLIW on organism response and distribution is challenging, because of NLIW unpredictability, short temporal and spatial scales, and the difficulty in resolving the biological response. Measurements of currents, temperature, and shark acoustic traces in Massachusetts Bay were used to evaluate the short-term response of individuals as well as the mean effects on the distribution of an aggregation of small sharks, Squalus acanthias. In two NLIW events, we detected 527 and 3240 shark traces. Individuals moved up and down in response to the currents associated with the sinking and rising of the thermocline. However, mean distribution deepened during one of the events, suggesting that organisms did not merely move in concert with the thermocline oscillation, but that sharks instead might have responded actively. Measurements of vertical currents and shark's depth change during one of the NLIW events indicate that with downward currents (sinking of the thermocline), the sharks tend to react passively. However, in response to the fastest upward currents, sharks appeared to swim down, supporting an active response in the rising phase of the wave. NLIW and other high-frequency processes can have a profound influence on the distribution of pelagic organisms, yet their ecological consequences remain largely unaccounted for.

Understanding the distribution of animals in space and time is a fundamental goal of ecology. For species that are commercially exploited or vulnerable to disturbance from anthropogenic activities, such information is vital for the development of conservation strategies and management plans. In marine systems, the sharpest environmental gradients, including light, temperature, and stratification, align with the depth axis. Thus, the depth (vertical) distribution of individuals within the water column is an important component of a spatiotemporal distribution. Shark distribution is the result of active movement, but currents may also influence their distribution. Recent advances in sensor autonomy and miniaturization have revolutionized the understanding of shark and other megafauna movements in the horizontal dimension (Hussey et al. 2015).

The depth distribution of shark species is not well known (Cortés et al. 2010). However, studies using tags have addressed seasonal, lunar, and diel variability in vertical movement (Vianna et al. 2013). Positioning in the water column of sharks and other pelagic elasmobranchs is commonly interpreted as a response to thermal regulation, metabolic rate, or as a result of foraging behavior (Sims et al. 2005). Little consideration has been given to the response of highly mobile pelagic predators to high-frequency variability (e.g., periods of seconds to a few hours), despite such features influencing the energetic costs associated with swimming or even the prey field. This gap is significant, because high-frequency vertical variability caused by internal waves and internal tides is pervasive in all oceans, and these phenomena can be a major structuring factor in coastal ecosystems, with effects ranging from nutrient input (Leichter et al. 2003) to larval transport (Pineda 1999) and prey aggregation (Greer et al. 2014).

Two trains of internal waves generate semidiurnally at low tide at Race Point channel, near the tip of Cape Cod, Massachusetts (Fig. 1), propagating, on average, in two directions, 250°T and 275°T (da Silva and Helfrich 2008). Waves propagating 275° may approach the southwestern flank of Stellwagen Bank, which is at ~ 10 km distance. Synthetic aperture radar (SAR) and in situ observations at the southwestern flank revealed nonlinear internal waves (NLIW) consistent with Race Point origin in ~ 62–68% of cases, and wave trains had long crests (19–28 km) and a propagation speeds of 30–50 cm s−1 (da Silva and Helfrich 2008; Pineda et al. 2015). Because of the predictability of the NLIW at the southwestern flank, this site is ideal for investigating their ecological consequences. The occurrence of top predators at Stellwagen Bank southern flank is suggestive of the relevance of the NLIW (chapter 4, U.S. Department of Commerce 2010).

Details are in the caption following the image
Detailed and wide-perspective maps of the study area. The arrow in the detailed map points in the estimated direction of propagation in the 01 July nonlinear internal waves event, and arrow origin indicates the site of the observations at the southwest flank of Stellwagen Bank. The suspected generation site is near Race Point channel (da Silva and Helfrich 2008), about 10 km distant from the study site. Contour lines every 15 m. Digital bathymetry from Butman et al. (2007). Upper right corner in wide-perspective map is eastern United States and Canada. Stellwagen Bank is designated by the box near the horizontal arrow. Digital coastline from NOAA's National Geophysical Data Center.

Squalus acanthias (spiny dogfish) is a coastal, circumboreal species, and the most abundant shark in the western North Atlantic (Bigelow and Schroeder 2002). Globally, the species is considered vulnerable by the International Union for Conservation of Nature (IUCN), and its population is declining (Fordham et al. 2006). It is slow growing; females reach maturity at 12 yr (Nammack et al. 1985) and can live to at least 35 yr. Spiny dogfish are opportunistic predators on ctenophores, squid, bivalves, herring, hake, sand lance, and other benthic and pelagic invertebrates and fish (Link et al. 2002). S. acanthias are found in a range of water temperatures, with preferred temperatures in adults ranging from 6 to 13°C (Shepherd et al. 2002; reviewed in Stehlik 2007). Tagged S. acanthias display day and night activity over a range of depths, including dives to over 600 m depths (Sulikowski et al. 2010). Offshore–onshore movement is related to water temperature and segregation of genders (Bigelow and Schroeder 2002), and aggregations migrate northward from North Carolina in the spring to Massachusetts and the Gulf of Maine (Jensen 1965).

We observed NLIW approaching Stellwagen Bank southwestern flank with measurements of currents, acoustic backscatter, and temperature. Coincidentally, we detected aggregations of small sharks S. acanthias during two NLIW events. Analyses of these measurements were used to resolve shark individual and mean (aggregate) depth displacements, vertical currents, and the NLIW. Do sharks change their depth distribution in response to the NLIW? If sharks respond to the NLIW, do they move up and down passively with the wave vertical currents, or do they respond actively?


Trains of NLIW approaching the Southwest flank of Stellwagen Bank were measured on 23 July 2010 and 01 July 2011 from the anchored RV Auk, in ~ 55 m deep water (Fig. 1). Concurrently, large numbers of sharks were observed with acoustic backscatter for a total of 527 acoustic traces on 01 July, and 3240 on 23 July. Boat position changed during the day, pivoting around the anchor with the tidal currents; the semidiurnal tide in the field site has a maximum range of ~ 3.1 m. These observations were taken while testing the hypothesis that humpback whales aggregate in response to zooplankton and fish aggregation by predicable NLIW (Pineda et al. 2015).

The 23 July event featured small amplitude waves propagating over a perturbed field, and some waves appear superimposed. The 01 July event included clearly defined, large-amplitude waves propagating over a mostly unperturbed thermocline. Our analysis focused on the 01 July event, when we estimated wave phase speed and direction of propagation from Doppler profiler data, and derived vertical currents associated with individual shark traces. For the 23 July event, we present only shark distribution data.

Physical measurements

Current velocity, temperature, acoustic backscatter, and density

East, north, and vertical current velocities were measured with a Doppler current meter. The 600 kHz RDI ADCP (four transducers and beam angle 20°) sampled from the vessel looking-down every ping (1.4 s), and bin height was 1.5 m. Propagation direction φ and phase speed c of the NLIW on 01 July were estimated by solving the “antenna” problem (e.g., Ufford 1947; Lee 1961). The timing of the four leading depressions observed with ADCP's acoustic backscatter at a selected depth (~ 27 m) was noted for each beam, and time lags among beam pairs were calculated for each wave depression. Lags among beam pairs in the four wave depressions were then averaged for the appropriate beam pair. Derivations of φ and c considered the mean heading of the instrument during the event and magnetic declination. Phase speed, c, was used to derive along-wave distance from time. Acoustic backscatter was also measured with a Biosonics DTX echosounder (200 and 120 kHz transducers). The horizontal currents in the direction of internal wave propagation were obtained by rotating the east–north currents according to the propagation direction φ (e.g., Scotti et al. 2005). Beam spread in the ADCP results in sampling heterogeneous currents fields, and may lead to inaccurate horizontal velocity estimates. Furthermore, averaging of the current field by the four beams likely smooths horizontal current velocities. The ADCP was programmed to produce horizontal earth coordinates (north–south and east–west). Thus, the data cannot be corrected for beam spread using techniques that require beam coordinate data. Horizontal currents were only used to provide a visualization of the flow field in the 01 July event. The ADCP vertical currents were evaluated against independent estimates of vertical velocity derived from the Biosonics echosounder acoustic backscatter. Acoustic backscatter patterns from the echosounder were measured in the vicinity of the shark traces (within 10 m above and below), and vertical displacement of backscatter patterns with time was estimated (see Fig. S1 legend for details). Temperature loggers (RBR 1060), attached on a line that was suspended from a large float, measured temperature every 10 s at depths ranging from 0.6 to 51.7 m. The float was tethered to the boat, and the line was kept taut by suspended weights. Logger spacing was 2.25 m on 01 July, with shorter separation near the surface and bottom; spacing was 4.5 m on 23 July. On 01 July, three temperature loggers did not record data. Onboard the RV Auk, the acoustic backscatter transducer and the Doppler current meter were ~ 5.5 m apart, whereas the temperature line was separated up to 30 m from the boat. On 01 July, temperature was advanced to match the NLIW in the acoustic backscatter and current meter data, and on 23 July, temperature time was slowed to match backscatter. Conductivity, temperature, and depth profiles were taken with an RBR XR-620 CTD from downcast measurements, and density was derived from conductivity and temperature for the 01 July event.

Shark observations

S. acanthias is the most abundant coastal shark in the region (Bigelow and Schroeder 2002). Acoustic backscatter measurements were used to estimate shark depth and depth change with time. The 200 and 120 kHz transducers were held ~1 m below the sea surface and sampled at a rate of five pings per second. Only the 200 kHz transducer was used on 01 July, and most measurements on 23 July used the 200 kHz frequency, but the 120 kHz data were used to resolve ambiguous data. The real-time acoustic data were used in conjunction with a live-feed cabled video camera (SplashCam Deep Blue Pro) to identify the backscatter traces characteristic of the sharks, with lighting provided by two small scuba LED flashlights. Traces of S. acanthias had a characteristic “rectangular” appearance, and were between −50 dB and 0 dB, and easily distinguishable in the echograms (Fig. 2B). Traces that were less than 4.5 s in duration were not considered in the analysis. In cases where two traces occurred consecutively, each trace was assumed to represent a shark unless there were faint or broken traces between the two. Faint traces that were perfectly parallel to a shallower trace were assumed to be an echo reflection and disregarded. Shark traces that were within 1–1.5 m of the camera were also disregarded, in case the lights influenced a shark's vertical position. We assumed individual traces were unique individual observations, but we likely measured some individuals more than once.

Details are in the caption following the image
Sharks and first wave of depression on 01 July. (a) Screen-capture photographs of Squalus acanthias from video camera. (b) Echogram with shark traces and an internal wave on 01 July (first wave of depression). Shark traces are bounded by white or yellow markers. The broken feature at ~ 1850 and 2400 pings and 26 m is the camera (red markers). The sharp feature at ca. 54 m is the bottom.
For each shark trace, depth Z, taken as the vertical middle of the trace, and ping number, an index of time T, were measured at the beginning (Z1, T1), and end of trace (Z2, T2). These measurements were used to estimate depth-change rate for individual traces, DS:

Depth of horizontal midpoint of the trace Zmd was also recorded. Vertical currents corresponding to each acoustic trace measurement (wi) were obtained, and an average of the vertical currents urn:x-wiley:00243590:media:lno11341:lno11341-math-0003 was computed for the duration of each shark trace. Individual traces were difficult to differentiate at very high densities of S. acanthias, which often occurred on 23 July. At very high densities, measurements of S. acanthias traces were taken from sections of full-length traces at any point where the trace was sufficiently distinct. Ping number and depth of a trace section was measured at the earliest point where the trace could be distinguished and before it merged into other traces. Finally, to describe the depth distribution of the shark aggregation, we calculated a 10-point running mean of shark depths, and we refer to this quantity as the mean depth of the aggregation.

To resolve whether DS related to the vertical currents on the 01 July NLIW, we parsed the shark observations among those that occur in the steepest sections of the NLIW train, and all others. The steep sections were associated with the five leading depressions, and occurred between 15:23 and 16:01 h (see below), but some records within this interval were determined to occur in the nonsteep section (e.g., around the waves crests and troughs, at times 15:29, 15:37, 15:37, 15:45, and 15:54 h). The steepest section of the NLIW was determined from wave patterns in ADCP acoustic backscatter (average of four beams for each wave depression), and the record was divided in sections with steep and nonsteep patterns. The steepest sections corresponded to those in which changes in backscatter isopleth-depth with time dz/dt were >|1.2|cm s−1 at about 20 to 32 m water depth. (For comparison, dt/dz values in the steepest sections at 20–32 m depth were up to dz/dt =|6–8|cm s−1, and higher absolute velocities at shallower depths.)


On 01 July, density stratification preceding the NLIW extended to surface waters (Fig. 3). There were five clearly defined wave troughs with periods of ~ 8 min and amplitudes of up to 20 m, comprising about 2/5 of the water column (Fig. 4). These waves are highly nonlinear with the nonlinearity parameter, the ratio of wave amplitude to upper layer depth, greater than one (Stanton and Ostrovsky 1998; Helfrich and Melville 2006). The well-defined waves had a similar amplitude and period, and were followed by a set of irregular, smaller-amplitude NLIW. The estimated phase speed was c = 0.30 m s−1, yielding a wavelength on the order of 150 m.

Details are in the caption following the image
Density (solid gray line) and temperature (dotted line) profiles at 13:42 h, before the 01 July nonlinear internal waves. The CTD sampled the entire water column.
Details are in the caption following the image
Internal wave train from temperature, 01 July. Along-wave distance was derived from time by using estimated phase speed c. Open symbols at approximately −160 m represent temperature loggers.

Some horizontal velocities in the direction of wave propagation u at 10–20 m were of similar magnitude to c (Fig. 5), suggestive of trapped cores (Helfrich and White 2010; Luzzatto-Fegiz and Helfrich 2014) and finite transport in the direction of wave propagation. These horizontal currents should be interpreted cautiously, however, because of biases in velocity estimates associated with beam spreading as mentioned earlier. The correlation between the ADCP vertical velocities and acoustic backscatter estimates of vertical velocity was 0.83 (n = 185; Fig. S1), suggesting that these two independent estimates of vertical velocity are consistent. The surface manifestation of the waves included bands of small breaking waves that rocked the boat as they passed by (Fig. S2). Acoustic backscatter measurements during the event yielded 527 shark traces, identified as S. acanthias with the help of live video (Fig. 2A), with 185 traces in the steepest sections of the NLIW, and 342 in nonsteep regions.

Details are in the caption following the image
First two depressions of the nonlinear internal waves train measured with acoustic backscatter (Biosonics), 14°C isotherm, and currents in the direction of propagation on 01 July (negative arrows are in the direction of propagation). Propagation speed was subtracted from the currents, and along-wave distance was derived from time by using estimated phase speed c. Therefore, currents are relative to estimated phase speed.

Shark distribution on 01 July centered around 20 m water depth at the beginning of the observations, about 14:00 h UTC (10:00 h local time, EST), with no sharks observed deeper than about 30 m (Fig. 6). Mean depth switched at about 14:45 h, when sharks were at 30 to 40 m depths, with only a few individuals observed around 20 m. Sharks moved deeper in response to the NLIW just past 15:20 h, with no sharks in shallow water. The mean depth of the aggregation appeared to move up and down with the isotherms during the first five depressions, and the deepest distributions on 01 July were associated with the second and fourth depressions (Fig. 6). Few sharks were observed from 16:30 h to about 17:00 h, and the distribution observed thereafter was broad, with sharks observed at ~ 15 to 50 m depth.

Details are in the caption following the image
Shark depth Zmd and 18°C, 14°C, and 10°C isotherms on 01 July. [Color figure can be viewed at]

Wave amplitude was smaller and sharks were more abundant on the 23 July event (number of shark traces = 3240). During the first two and largest depressions, shark mean depth moved up and down with the waves. The mean depth, estimated as the 10-point running mean, fluctuated around 35 m (Fig. 7). Sharks were abundant between 15:00 and 15:30 h, and between 15:15 and 15:20 h, a large number of them were in shallow 10–20 m waters. The depth range of the distribution tended to be large, although a relatively small number of sharks were observed in waters shallower than 20 m from approximately the beginning of the largest waves, just past 13:30 to ~ 15:01 h. Inspection of the echograms indicate that on 23 July the distribution was sometimes bimodal, with a group of individuals near the bottom, and another group at 35 or 25 m water depth.

Details are in the caption following the image
Shark depth Zmd and 20°C, 18°C, 14°C, and 10°C isotherms on 23 July. [Color figure can be viewed at]

DS in the steep portions of the NLIW train tended to correlate positively with the mean vertical currents, indicating that individual shark traces move up and down with the ascending and descending currents (Fig. 8A). However, no pattern was observed for the sharks in the nonsteep portion of the NLIW (Fig. 8B). The patterns in the steep portion of the NLIW may relate to a passive response, where sharks move up and down in concert with the vertical currents, or an active one, where individuals may swim up or down in response to the NLIW. To resolve whether DS may have changed with the sign and magnitude of the vertical currents, average vertical currents urn:x-wiley:00243590:media:lno11341:lno11341-math-0004 were subtracted from DS for each observation and plotted as a function of urn:x-wiley:00243590:media:lno11341:lno11341-math-0005. The trend in urn:x-wiley:00243590:media:lno11341:lno11341-math-0006 as a function of urn:x-wiley:00243590:media:lno11341:lno11341-math-0007 was characterized with a cubic polynomial, and a 10-point running mean. If sharks behave as passive particles, such a plot should show no trend. For the negative, downward currents, patterns suggest no response, that is, sharks appear to move down with similar velocities than those associated with the sinking thermocline (Fig. 9). For the positive, upward currents, however, the trend indicates a slight increase at the velocity 0.06 m s−1, and a more pronounced decrease at velocities larger than 0.08 m s−1, suggesting that sharks swam down at the highest positive currents (Fig. 9).

Details are in the caption following the image
Depth-change rate (DS) as a function of mean vertical currents, urn:x-wiley:00243590:media:lno11341:lno11341-math-0008, on 01 July. Individuals in (a) the steep regions, and (b) nonsteep regions of the nonlinear internal waves.
Details are in the caption following the image
DS with mean vertical currents urn:x-wiley:00243590:media:lno11341:lno11341-math-0009 removed as a function of urn:x-wiley:00243590:media:lno11341:lno11341-math-0010 for traces in the steep section of the nonlinear internal waves, 10-point running mean (dotted line), and cubic polynomial fit (solid line).

We tested whether the patterns described above were due to chance with a randomization analysis (Manly 1997). First, we calculated the sum of the squared residuals from fitting the polynomial cubic model to the urn:x-wiley:00243590:media:lno11341:lno11341-math-0011 vs. (urn:x-wiley:00243590:media:lno11341:lno11341-math-0012) data (e.g., Fig. 9). Then, we randomly reassigned (urn:x-wiley:00243590:media:lno11341:lno11341-math-0013) values to the urn:x-wiley:00243590:media:lno11341:lno11341-math-0014 data, fit a cubic polynomial, calculated the sum of squared residuals, and repeated 100,000 times (Fig. 10). The approximate p is the proportion of sum of squared residuals from the randomizations that are smaller than the sum of squared residuals derived from the nonrandomized data set (1.1685, p = 0.00005). These results suggest the trends described by the cubic polynomial fit are not random.

Details are in the caption following the image
Frequency distribution of the sum squared residuals (SSR) from cubic polynomial fit of 100,000 randomized DS − urn:x-wiley:00243590:media:lno11341:lno11341-math-0015 values. The SSR from the observed data and approximate p are also given.


Our study adds to the diversity of NLIW observed in Massachusetts Bay, including rank-ordered undular bores (Haury et al. 1979; Butman et al. 2006) and waves of elevation (Scotti and Pineda 2004). In particular, the 01 July event is remarkable as the set of clearly defined waves had a comparable period and amplitude, vis-à-vis a rank-order wave train, had a vivid surface expression, and O(1) nonlinearity. The current velocity and temperature measurements, with sharp horizontal temperature gradients near the surface suggest mass transport in the direction of wave propagation. Density stratification extended to the surface, and mass transport in trapped cores can occur in these conditions (Helfrich and White 2010). Studies in this region have identified two local generation sites: Stellwagen Bank (e.g., Haury et al. 1979) and Race Point channel (da Silva and Helfrich 2008). The NLIW on 1 and 23 July likely originated off Race Point, as do the majority of NLIW observed in the southwest flank of Stellwagen Bank (Pineda et al. 2015). Moreover, the observation of superimposed NLIW trains is consistent with generation of two NLIW sets at Race Point (da Silva and Helfrich 2008).

Nonlinear internal waves are ubiquitous in the world ocean. Yet, their unpredictability and small temporal scales, coupled with the difficulty in resolving the biological response, have hindered understanding of how NLIW influence the behavior and distribution of pelagic organisms, from microbes to whales. The response of organisms to NLIW and other processes where vertical currents are significant can be passive or active. Furthermore, their effects on distribution, or organism's physiological and ecological condition, can be direct, where currents and turbulence, or the interaction of these with individual behavior, produce changes in distribution (or individual condition) or indirect, where a variable forced by the waves results in a change in distribution or condition (Table 1). Temporal and spatial changes in horizontal distribution that parallel pycnocline deformation caused by the NLIW suggest a passive response (e.g., Haury et al. 1983). An active behavioral response with a direct influence on distribution may result from plankton swimming up against downwelling currents, resulting in accumulation and transport in internal tidal bore warm fronts (Pineda 1999). NLIW might also change indirectly the depth distribution of pelagic organisms, as when thinning and thickening of the surface turbid layer by NLIW changes the light environment at depth, resulting in short-term vertical migrations of fish (Kaartvedt et al. 2012).

Table 1. Hypothetical examples of organism's responses (passive/active) to processes featuring vertical currents and mode of influence in organism's distribution or physiological and ecological condition (indirect/direct).
Passive Active
Indirect Sessile benthic organisms buried by sediments resuspended by NLIW. Intensification of trophic interactions due to prey aggregation.
Direct Changes in organism's depth distribution parallels pycnocline depth. Organisms swim up in response to vertical displacement. Distribution can differ from that of a passive tracer.

Observing plankton and nekton active response to vertical currents in the field is challenging. Thus, observational studies have used indirect approaches, whereby behavior is inferred from pelagic distribution patterns and assumed or measured vertical currents (e.g., Pineda 1999; Lennert-Cody and Franks 2002). Laboratory studies can more easily resolve both circulation and behavior (e.g., DiBacco et al. 2011). Our unique data set allowed us to measure the response of sharks to the 01 July NLIW event. Results indicate a range of responses, from passive to active, and that the changes in organism distribution were likely in direct response to the NLIW. There was no trend in urn:x-wiley:00243590:media:lno11341:lno11341-math-0016 for downward currents, suggesting that sharks responded passively to the NLIW when the thermocline was sinking. At the highest velocities of the upward phase of the wave, however, sharks appeared to respond actively by swimming down. These large-amplitude waves clearly influenced the short-term depth distribution of the shark S. acanthias.

What are the consequences, and what might be the ultimate causes, explaining sharks swimming down during the fastest upward currents? The individual responses of the sharks might have influenced aggregation mean depth. During the 01 July event, the center of mass of the distribution deepened during the NLIW event and ~ 90% of the shark traces were associated with waters < 10.2°C. During the 23 July event, with smaller amplitude waves, deepening in the distribution occurred during the first two depressions, although this depth distribution appeared similar to other times during that day. On 23 July, about 90% of sharks were in waters < 11.8°C. On 01 July, the deeper distribution resulting from NLIW exposed sharks to temperatures as cold as 6.2°C, with metabolic consequences for the sharks (e.g., Brett and Blackburn 1978). The deeper distribution might have also exposed the sharks to a different prey field, with potential for bentho-pelagic coupling and community structure consequences. Strong NLIW-induced currents can extend to the bottom at this location (e.g., Pineda et al. 2015) and cause sediment resuspension (Butman et al. 2006), and this might intensify trophic interactions by, for example, resuspending chemical olfactory cues that may elicit a predatory response. The proximate and ultimate factors that caused sharks to respond to the ascending thermocline are unknown, but by swimming down, they might have escaped the horizontal shear associated with the upper water column. Thus, we speculate that sharks swam down in response to the upward fast currents to avoid areas of maximum shear.

This study characterized the response of S. acanthias to NLIW using high-frequency acoustic backscatter profiles. This enabled us to quantify shark distribution and behavior from hundreds of observations at temporal scales commensurate to NLIW variability. While shark (and other large nekton) spatial distribution and behavior can be measured by acoustic telemetry, and tagging studies have been used to measure top predator vertical movements in stratified waters (Eckert and Stewart 2001; Aspillaga et al. 2017), tagging approaches have a number of limitations, including physical interference with the animals and the number of individuals that can be tagged. In order to generate data comparable to that obtained in this study, a very large number of individuals would need to be tagged to ensure a reasonable chance that those individuals were present at the time and location of NLIW. Furthermore, acoustic telemetry studies typically capture data at a lower sampling frequency than in our study, but for a longer time period (Hussey et al. 2015). Although backscatter data have limitations, including potential difficulties in differentiating traces and the need to identify the animals producing traces (e.g., through live-feed video camera), acoustic backscatter data to study fish behavior allow simultaneous measurement of a large number of individuals and is a noninvasive technique (Fielding et al. 2004; Colbo et al. 2014). Future studies that combine approaches (e.g., telemetry and acoustic backscatter data) could be used to understand how fine-scale behavioral responses translate into longer-term, large-scale distribution patterns.

Understanding the full range of behavior and individual movements of S. acanthias has been identified as an urgent priority to aid the conservation and management of this globally declining species (Thorburn et al. 2018). Our results contribute to the growing body of research documenting how hydrodynamic processes influence marine vertebrate distributions (e.g., Yen et al. 2004; Hooker et al. 2011; Scales et al. 2014), and in particular show that high-frequency processes are useful in understanding the vertical component of large nekton distribution. While large aggregations of sharks were observed at Stellwagen Bank during the two NLIW events, it is not known whether sharks regularly aggregate at this site in response to NLIW. If sharks do persistently aggregate at sites with NLIW or other high-frequency processes, such areas may deserve special consideration for conservation activities. Additionally, understanding the magnitude and predictability of associations between sharks and NLIW will aid efforts to characterize biological–physical coupling in marine ecosystems, and ultimately help predict habitat use for large apex predators (Yen et al. 2004).


We thank the personnel that helped us in the cruise and logistics, including Brad Cabe, Michael Thompson, and the crew of the RV Auk. Andy Solow suggested to randomize the sum of square residuals, and the reviewers and editor provided constructive comments. This work was supported by Woods Hole Sea Grant, and the Woods Hole Oceanographic Institution.

    Conflict of Interest

    None declared.