Locomotion and behavior of Humboldt squid, Dosidicus gigas, in relation to natural hypoxia in the Gulf of California, Mexico

The Humboldt squid, Dosidicus gigas, is a large muscular squid with a rapid growth rate, a short life span and extremely high fecundity (Nigmatullin et al., 2001; Markaida et al., 2004; Argüelles et al., 2001). It is associated with highly productive, upwelling-driven ecosystems in the California and Peru Current Systems of the eastern Pacific Ocean (Roper et al., 1984). Throughout its large range (Canada to Chile) D. gigas plays an important ecological role as both predator and prey. In the Gulf of California, prey consists primarily of mesopelagic micronekton, including myctophid fishes, squids, pteropods and crustaceans (Markaida and Sosa-Nishizaki, 2003; Markaida et al., 2008), but this menu expands to include a variety of neritic and pelagic fishes in continental shelf and slope environments along the Pacific coast in the northern hemisphere (Field et al., 2007; Bazzino et al., 2010). In turn, D. gigas is a major prey item for sperm whales and other marine mammals, large teleost fishes and sharks, and is thus an important transitional organism in trophic webs, transferring energy from micronekton to the largest apex predators.

An important feature associated with most of the eastern Pacific Ocean is a well-defined oxygen minimum zone (OMZ), a midwater region where the concentration of dissolved oxygen is naturally lower than 20 μmol kg⁻¹ (~0.5 ml l⁻¹ or 0.7 mg l⁻¹) (Levin, 2003; Paulmier and Ruiz-Pino, 2009). This value is close to the concentration where microbial respiration switches from O₂ to NO₃⁻² [~10 μmol kg⁻¹ (Hofmann et al., 2011)] and is considered severely hypoxic for many nektonic organisms, including fishes, squids and crustaceans. The depth of the OMZ in the eastern Pacific varies with latitude (Helly and Levin, 2004) but typically extends from ~250 m to 1200 m in the southern half of the Gulf of California (Alvarez-Borrego and Lara-Lara, 1991). Control of oxygen concentration in the OMZ involves both physical and biological factors, including delivery of source water and microbial-based decomposition of sinking organic material in a deep, stable water column (Azam, 1998; Fiedler and Talley, 2006; Pennington et al., 2006; Karstensen et al., 2008; Paulmier and Ruiz-Pino, 2009). A variety of mesopelagic organisms, including myctophid fishes and krill, which constitute the main prey items of D. gigas, have adapted to conditions in the upper OMZ (Childress and Seibel, 1998; Drazen and Seibel, 2007), where they can avoid aerobic, visual predators during the day. This community of organisms comprises a major acoustic deep scattering layer (DSL) and generally undertakes a diel migration to near-surface waters to feed on plankton at night (Forward, 1988; Ringelberg, 1995; Benoit-Bird et al., 2009).

Dosidicus gigas also migrates vertically in the water column, in a generally parallel fashion to that of the DSL, spending nighttime at shallow depths and daytime near the upper OMZ in the Gulf of California (Gilly et al., 2006) as well as in the Pacific Ocean off Baja California (Bazzino et al., 2010). Thus, D. gigas is potentially exposed to prey on a constant basis and may therefore be able to forage during both the day and the night. Although D. gigas can greatly reduce its resting metabolic rate in response to a cold, hypoxic challenge (Gilly et al., 2006; Rosa and Seibel, 2008; Rosa and Seibel, 2010), a previous tagging study did not detect any reduction in vertical velocities during daytime hours that were spent at very low oxygen (Bazzino et al., 2010). Effects of hypoxia on the metabolic rate of D. gigas during active swimming are not known, but presumably highly aerobic activity such as high-velocity jetting would be significantly impaired in the OMZ.

In this paper we address this puzzle by deploying electronic tags of improved resolution and an animal-borne video package (AVP) in different locations and seasons in the Guaymas Basin of the Gulf of California that are characterized by large differences in the depth of the upper OMZ. This allows a thorough analysis of the vertical distribution of D. gigas in relation to this environmental feature and provides new insight into the patterns and speed of vertical movements. Nearly all locomotion by D. gigas appears to involve fins-first active ascents and passive descents in a pattern of classic ‘climb-and-glide’ swimming (Weihs, 1973). Temporal summation of small climb-and-glide events underlies complex and rapid swimming behaviors in all parts of the water column, including the OMZ, but vertical velocities across a wide range are clearly reduced in the upper OMZ. We also find that when the upper OMZ is deeper than ~300 m, Humboldt squid decouple from this feature and spend daytime at shallower depths, which are not as severely hypoxic. In these cases, light and presumably the presence of the DSL probably act to determine the daytime habitat of this species.

Humboldt squid [Dosidicus gigas (D'Orbigny 1835)] were captured from research vessels in the Guaymas Basin (including its northern extension, known as the San Pedro Martir Basin) of the Gulf of California, Mexico, between 2004 and 2008 using conventional rod-and-reel or hand-line methods with weighted, luminescent jigs. Squid were brought on deck or held next to the vessel and assessed for suitable size for tagging [>70 cm dorsal mantle length (DML)] or AVP deployment (>80 cm DML) before proceeding as described below. Care was taken to ensure vitality of a tagged squid by holding it gently beneath the surface until it showed a burst of powerful jetting. Squid for tagging were caught after dusk; squid for AVP deployments were captured in mid-afternoon.

Pop-up archival transmitting (PAT) tags (Wildlife Computers, Redmond, WA, USA) were attached to the medial-ventral surface of a fin as described elsewhere (Gilly et al., 2006). Animals were tagged on deck with constant irrigation of the gills with ambient seawater. MK10 tags with 1 Hz sampling were used in most cases, and these devices measure depth from 0 to 2000 m with a resolution of 0.5 m and temperature from 0 to +40°C with a resolution of 0.05°C. Two PAT4 tags were also used (10/04 and 07/06B in Table 1). Depth resolution of these tags ranges from 0.5 m (0–100 m range) to 8 m (for depths >400 m). Only data from MK10 tags with 1 Hz sampling were used for analysis of vertical velocity at different depths.

Eleven of 13 tags deployed in the course of this study reported summary data to the Argos satellite system. Nine tags were physically recovered, and the complete archival data sets were analyzed (Fig. 1, Table 1). These tags are referred to by their month and year of deployment (mm/yy). Four tags were deployed in the vicinity of Santa Rosalia, Baja California Sur (10/04, 05/08, 11/08A,B), two off Guaymas, Sonora (03/07A,B), and three near Isla San Pedro Martir (07/06A,B,C). To compare swimming behavior and vertical distribution of individual squid with independently assessed hydrographic properties (oxygen and temperature), we analyzed vertical distribution and velocities for the first 3 days, which was assumed to be a reasonable time for the squid to remain in an area with similar conditions to those in the area of the tag deployment where the hydrographic surveys were made.

Tags 10/04 and 07/06C were recovered opportunistically and returned for a reward. All other tags were recovered through active searches at sea using a radio direction-finding system based on a mast-mounted antenna (DDF6095) and display processor (DDF6001F, Doppler Systems, Carefree, AZ, USA) and receiver (PRO-97, Radio Shack, Fort Worth, TX, USA) that allowed localization of the 401.650 mHz transmissions by the floating tag to Argos satellites. Updated tag positions were accessed through the internet using a portable satellite phone. We relied on a hand-held Argos receiver (TSUR-400, Telonics, Mesa, AZ, USA) to provide the final, close-range detection (generally several hundred meters) of the tag before commencing a dedicated visual search. Floats of the tags were painted International Orange to facilitate visual detection. Before each search, the radio direction-finding system was calibrated in place on the vessel using a SPOT5 tag (Wildlife Computers) to provide an Argos signal, typically from a distance of 100–200 m.

Tag data were analyzed with MATLAB (MathWorks, Natick, MA, USA), IGOR Pro (WaveMetrics, Eugene, OR, USA) and a custom graphical user interface (L. Giovangrandi, Electrical Engineering, Stanford University, CA, USA) that allows visualization and specific transformations of data. Vertical velocity was computed for those PAT tags with 1 Hz sampling as the simple, two-point time derivative of depth after smoothing the depth record with a Savitsky–Golay filter (order 6, frame size 33).

National Geographic's Crittercam (Marshall et al., 2007) was deployed on three squid in September 2009 at Punta Prieta north of Santa Rosalia (27.54°N, 112.3°W; see Fig. 1). This AVP (0.7 kg in air, 25.4×5.7 cm) is capable of storing digital black-and-white video (17 Hz) and logging sensor data at 1 Hz for temperature, depth, light and acceleration along three mutually orthogonal axes. The instrument package was fitted onto an animal being gently held at the surface behind the vessel's dive platform. A sleeve of stretchable nylon fabric (57.5 cm circumference), attached to a thin fiberglass sheet for docking the AVP, was slid over the mantle from the posterior (fins) end, and slack was taken up by cinching the sleeve with plastic cable ties in a way so as not to restrict normal mantle movements. The assembly was positioned so the docking sheet was located on the dorsal midline, and the sheet was then secured to the anterior edge of mantle on each side of the gladius by a cable tie looped through the muscle ~1 cm from the mantle edge and 4 cm from the midline. Finally, a syntactic foam cradle holding the Crittercam was attached to the docking plate by a programmable release mechanism. Software for programming the AVP and uploading data were provided by National Geographic's Remote Imaging Unit.

Two AVP deployments (nos 2 and 3) were made during daytime with no artificial lighting, and this allowed video observations under natural lighting conditions. Vertical movements were comparable to those recorded with PAT tags, indicating that the squid were not significantly impaired by the AVP. This is also suggested by a variety of social interactions between the squid carrying AVP 3 and a group of conspecifics that was encountered (see Results). Deployment of AVP 1 was carried out at night using red LEDs as a light source, but this squid was quickly attacked by a group of conspecifics. The use of artificial lighting was therefore abandoned, and data from AVP 1 are not considered in this paper.

Following its timed release and appearance at the surface, the AVP was located using a directional Yagi-type antenna to detect a VHF beacon. Video playback and editing was managed with the VLC media player (v1.1.2, VideoLan Project, Paris, France) and Adobe Premiere Pro CS4 (v4.2.1, Adobe Systems, San Jose, CA, USA). Image analysis was conducted using ImageJ (v1.42q, National Institutes of Health, Bethesda, MD, USA) and IGOR Pro software. Sensor data were analyzed with IGOR Pro and original programs in both MATLAB and R (v2.12.2, The R Foundation for Statistical Computing, Vienna, Austria).

Vertical velocity was computed as the time derivative of depth as described above. Pitch was defined as the angle between the squid's longitudinal axis (in the arms-to-fins direction) and a plane parallel to the sea surface, with positive angles being upward. When the squid is actively jetting in a fins-first direction, pitch is equivalent to the angle of inclination that describes the overall trajectory of the moving animal. This is not necessarily the case at all times (see Results).

Pitch was calculated from raw tri-axial acceleration data (A_x, A_y and A_z) as arctan[A_x/(A_z² + A_y²)^−1/2]. Absolute pitch angles from this analysis must be regarded as approximate, because data from tri-axial accelerometers alone cannot distinguish between static movements (i.e. pitch, roll) and small accelerations in a particular direction that would be generated by active jetting. Pitch measurements were deemed valid if values for (A_x²+A_y²++A_z)^1/2 deviated from 1.0 by less than 10% (Marshall et al., 2007). Most deviations outside this range lasted only for a few seconds. Video data were also used to confirm orientation of the squid relative to the surface based on background light intensity.

Hydrographic measurements at the sites of tag deployment were made on the same day to a maximum depth of 600 m with an SBE19plus conductivity-temperature-depth (CTD) profiler equipped with an SBE43 oxygen sensor that was calibrated annually (Seabird Electronics, Bellevue, WA, USA). Additional profiles were obtained over the course of several days preceding and following tag deployment. In several cases profiles were also made at the time of tag recovery. These data were analyzed with IGOR Pro and were imported into Ocean Data View 4 (http://odv.awi.de) for further analysis and plotting.

Sea surface temperatures throughout the year were measured in the Santa Rosalia area by monthly CTD casts during 2007–2008 (except August and September) using the equipment described in the previous paragraph.

In agreement with a previous study in this general location (Gilly et al., 2006), all of the squid carrying PAT tags generally spent nighttime at shallower depths than in the daytime and showed reliable movements towards the surface at dusk and back to depth at dawn. These overall features are evident in typical time-series data (Fig. 2). Although the diel migration pattern is clear, there is much high-frequency vertical movement, both at depth during the day and close to the surface at night. Physical features of the midwater environment will be established in the second part of this paper and related to squid distribution and behavior at that time.

Swimming by loliginid squid, coastal inhabitants that are much smaller than D. gigas, involves an active and complex combination of jet propulsion and fin beating at most swimming speeds (Bartol et al., 2001a; Bartol et al., 2001b; Anderson and DeMont, 2005; Bartol et al., 2008). In general, lower speeds involve more fin undulation and beating, whereas the highest speeds do not involve the fins at all (Hoar et al., 1994). In all cases, a cycle of jet propulsion involves inhalation of water into the mantle cavity, sometimes assisted by contraction of radial muscle fibers, followed by exhalation through the siphon driven entirely by contraction of circular muscle. The siphon is muscular and can aim in any direction, so jet-propelled movement can be fins-first or arms-first, with similar maximum velocities in both directions (Packard, 1969). Positive acceleration due to jetting occurs only during exhalation. Gliding, often accompanied by fin beating, occurs during refilling, making movement inherently pulsatile. The larger the squid, the more useful gliding will be as a component of locomotion, and a tagged D. gigas is a large squid with a mass of 20 kg or more.

Although the extensive archival records from PAT tags provide the richest variety of vertical movements over a large depth range, the basic features of locomotion are more readily explained in conjunction with squid carrying the AVP. Fig. 3A represents the complete depth (black, upper) and velocity (blue, lower) records obtained with AVP 2. After descending to ~100 m immediately after its release (arrow), the squid remained near this depth for ~1 h before beginning a gradual ascent to ~50 m. Vertical velocity transients of up to 1.0 m s⁻¹ peak amplitude are evident throughout this period. One such event (segment B) is expanded in Fig. 3B. A smooth climb of 3 m and associated velocity spike of 0.5 m s⁻¹ peak amplitude occurs in the middle of a slow descent with a mean (±s.d.) velocity of −0.056±0.039 m s⁻¹ (excluding 20 points associated with the velocity transient). Analysis of the squid's pitch (red traces and diagrams in Fig. 3B,C) indicates that the squid climbed in a fins-first direction at an inclination angle of 68 deg, tilted downward briefly (negative pitch angle) at the apex of the climb and then quickly assumed a pitch of near 0 deg during the subsequent slow descent.

Based on examination of many such records from both AVP and PAT depth sensors, we regard the event shown in Fig. 3B as representing the fundamental pattern of vertical movement – an active fins-first ascent, characterized by a brief (~10 s duration) velocity transient of ≥0.3 m s⁻¹ followed by much longer-lasting, slow descent at ~0.05 m s⁻¹. We assume that the slow sinking phase is essentially passive with respect to jetting for these slightly negatively buoyant animals. Fin beating probably occurs in combination with both jetting and gliding, but we cannot address its contribution to locomotion based on available data.

Although we discuss these events in terms of a single functional ‘jet’ during the ascent, it is likely that several cycles of contraction and partial relaxation are smoothly integrated to a degree that prohibits discrimination of the individual pulses. An alternative mechanism would be repetitive opening and closing of the siphon during a single cycle of mantle contraction, and this would likewise deliver pulses of thrust. These mechanisms could also be combined. Given that the inhalation rate under laboratory conditions at 7–10°C is ~0.33 Hz [unpublished data collected in conjunction with a previous study (Gilly et al., 2006)] and that siphon diameter is actively controlled in free-swimming animals (see Depth-maintenance behavior below), a combination of these mechanisms seems most likely. At present, we cannot specify the biomechanical details of the ascent phase of these individual ‘jetting events’.

This pattern of active ascent and passive descent defines classic ‘climb-and-glide’ locomotion seen in many flying and swimming organisms (Weihs, 1973). Repetitive respiratory movements of seawater into and out of the mantle cavity must occur during passive sinking occurring over such long periods, but evidently these events are too weak (or the siphon is oriented horizontally) to generate detectable vertical velocity transients. Based on observations in large tanks (W.F.G., personal observation) and from remotely operated vehicles (ROVs) (L.D.Z., personal observation), fin posture and activity can control the rate of descent, but this factor has not been quantitatively studied. We have no direct information regarding horizontal velocities with either AVP or PAT tag sensors, but some estimates can be made, as described in the next section.

Discrete climb-and-glide motions of this sort are the ‘elemental’ events out of which more complex behaviors are built. An ascent and descent of 10 m are illustrated in Fig. 3C. It is clear from inspection that the climb is composed of at least three elemental-type jetting events, as evidenced by the discrete peaks in the associated velocity transient. Larger steady climbs can thus be generated through temporal summation of small elemental events in a way that results in a relatively smooth velocity transient that is maintained throughout the ascent. Refilling of the mantle presumably occurs during the brief deceleration phase between individual jets, but it is unlikely that one or even several jets of this modest amplitude would deplete much of the seawater in the mantle cavity.

Varying the timing between jetting events, as indicated by the peaks of velocity transients, can control the rate of ascent, as demonstrated in Fig. 3D. Faster motion is generated by decreasing the time between mantle contractions to the point that little or no deceleration phase is visible (arrow), and the subsequent increase in time between jetting events during the second half of the record results in a slower rate of climb with little or no gliding. All of the above features of elemental events and their temporal summation are also seen in conjunction with active descents (Fig. 3E). A rapid downward jetting event is evident (arrow), and the smaller negative velocity transients during the subsequent descent (dashed line) are associated with a mean descent rate of 0.16±0.07 m s⁻¹, a value threefold faster than that associated with passive sinking.

It appears from records such as these that D. gigas extensively utilizes temporal summation of active jetting events and passive glides to carry out virtually all of its vertical movements. Elemental jetting events, as defined by the overall time course of the associated velocity transient (~10 s), can also vary somewhat in peak velocity (0.3 to 1 m s⁻¹; see initial event between vertical dashed lines in Fig. 3D), and this represents another mechanism for controlling overall velocity. This ability simply requires adjustment of the velocity of mantle contraction over a fixed time corresponding to an elemental event (~10 s).

An estimate of horizontal motion during a single climb-and-glide event can be derived from the changes in depth and pitch during individual events during the AVP deployments. A schematic event is illustrated in Fig. 4. Measurements of 27 such events from the AVP deployments with a vertical ascent of ≥1 m and similar descent showed a vertical (upward) movement (d_Vup) of 2.2±0.2 m (mean ± s.e.m.) traversed in 9.1±0.9 s (t_up), corresponding to a vertical velocity (V_Vup) of 0.27±0.02 m s⁻¹. Measured pitch angle for these climbs was 51.0±2.5 deg, and this value would equal the angle of inclination (θ). Pitch ranged from 30.5 to 77.8 deg, with higher angles more likely to be associated with longer ascents (data not shown). Descents occurred with the squid maintaining a posture of nearly zero pitch over a similar vertical distance (1.8±0.1 m) but much longer time (37.4±3.3 s) (t_down), defining an angle of glide (γ). Vertical velocity during the descent (V_Vdown) was −0.054±0.004 m s⁻¹, a value characteristic of passive sinking. In the following discussion, vertical ascent and descent distances are assumed to be equal for simplicity.

This approach is admittedly simplistic, neglecting drag and other factors during the descent, including weak, horizontally directed respiratory jets and fin activity, but it provides a rough lower estimate of horizontal motion that can be considered in conjunction with data from tags. Overall horizontal velocity during the climb-and-glide event would be given by (d_Hup+d_Hdown)/(t_up+t_down)=11.8/46.5=0.25 m s⁻¹ for α=0.75. For values of α ranging from 0.5 to 1.0, horizontal velocities would vary from 0.23 to 0.32 m s⁻¹, respectively. This range is consistent with that observed with more direct measurement methods (see Discussion).

Tagged squid (both AVP and PAT tags) often maintained their position in the water column for a considerable period of time over a narrow depth range. This behavior is evident throughout much of the depth records for both AVP deployments (Fig. 3A, Fig. 5A) and generally involves repetitive episodes of upward jetting and passive sinking, i.e. climb-and-glide. We operationally defined depth-maintenance behavior as maintaining vertical position within a range of ±15 m for at least 15 min. Typically, depth maintenance is carried out through simple climb-and-glide behavior (Fig. 5B) interspersed with other types of motions, such as the brief arms-first ascent at negative pitch (up arrow and diagram) and the quick downward jet toward the end of the record (down arrow).

Based on the above definition, squid carrying MK10 PAT tags spent 9.7±1.6% (mean ± s.e.m., N=7) of the total deployment time engaged in depth-maintenance behavior, with values ranging from 3.4% (tag 07/06B) to 18.5% (tag 03/07A). Bouts of depth maintenance occurred during both day and night, at all depths, and often occurred in clusters (data not shown). Durations of ascents were typically 10–40 s, and descents ranged from 20 to 150 s, with no consistent nature to the frequency distribution of durations. Vertical velocities during ascents and descents during depth maintenance were each exponentially distributed with means of ~0.5±0.1 and −0.15 m s⁻¹, respectively. Velocity distributions will be discussed in relation to oxygen concentrations at different depths in a following section (see Effects of OMZ conditions on vertical velocity).

One function for depth-maintenance behavior was demonstrated during the final 20 min of AVP 3 deployment (Fig. 5A,C) when the squid encountered a sizeable group of conspecifics at a depth of 50 m. During this period of extremely tight depth maintenance, there was little sign of canonical climb-and-glide swimming. Instead, the squid displayed a variety of rapid up-and-down movements of small amplitude, including some in which a negative velocity transient was accompanied by high positive pitch, indicating that the squid was descending arms-first (down arrows and diagrams), as opposed to the usual fins-first direction. All of the sudden vertical movements shown in Fig. 5C were probably associated with a close encounter with a conspecific, and a sampling of video frames captured is shown in Fig. 5D–H. Interactions included displays involving arm posture and apparent mating attempts, as evidenced by the appearance of spermatophores between the arms of the AVP-bearing squid (presumably a female based on its large size). Close interactions of this sort were always accompanied by chromatophore displays, and these behaviors will be described in detail in a subsequent publication.

Two features evident in the video frames are relevant to swimming mechanics. First, the arms are generally held tightly together and pointed in the anterior direction (Fig. 5D), with the prominent lateral keels on the third arm from the dorsal midline held in a position that resembles the posterior fins. Presumably this posture of the arms, like the fins, provides a stabilizing influence, but there was no indication of movement of the arms in a way that would generate propulsive thrust. Second, the siphon (arrow in Fig. 5Hi) is capable of large and rapid diameter changes (58.8 ms between sequential frames in Fig. 5Hi,Hii). This ability would provide an additional and highly effective control element for jet velocity.

Examples for tags deployed in autumn, spring and summer are shown in Fig. 6, and summary data for all PAT tags over the initial 3 days of deployment are provided in Table 2. Regardless of the time of year, all tagged squid spent daytime hours primarily at an average depth of 200–250 m where temperature was ~13°C. Oxygen concentration at these depths ranged from 12 to 70 μmol kg⁻¹ (Table 2). Tags deployed off Santa Rosalia in the autumn (10/04 and 11/08A, B) showed median daytime depths where oxygen concentration was characteristic of the upper OMZ (≤20 μmol kg⁻¹). Springtime deployments revealed a more variable picture, with oxygen at daytime depths ranging from 12 μmol kg⁻¹ (Guaymas, 03/07A) to ~60 μmol kg⁻¹ (Guaymas, 03/07B and Santa Rosalia, 05/08). Corresponding summertime values from San Pedro Martir were also ~60 μmol kg⁻¹ (07/06A, B). Although this upper range of oxygen concentration is higher than that found in the OMZ proper, it is low enough to be considered hypoxic for many pelagic fish (Diaz and Rosenberg, 1995; Hofmann et al., 2011).

In three cases, hydrographic properties were surveyed over a sizeable region to reveal the variation of oxygen at depth around the sites of each tag deployment (Fig. 7A). Oxygen (Fig. 7B) and temperature profiles (not shown) in the Santa Rosalia region in November 2008 revealed spatial variation in the thickness of the surface mixed layer (upper 75 m), but depths of the contours for both 20 and 60 μmol kg⁻¹ oxygen were quite constant over the area surveyed (Fig. 7C). More variation in oxygen at midwater depths occurred off Guaymas in March 2007, but the 20 μmol kg⁻¹ oxygen contour was again fairly constant (Fig. 7D). These results are consistent with the idea that some squid tagged in these areas in spring and autumn spent considerable time in the upper OMZ (Fig. 2A,B, Table 2) as well as in the less hypoxic, shallower region (20–60 μmol kg⁻¹ oxygen) that we will call the oxygen limited zone (OLZ).

Summer deployments in the San Pedro Martir Basin present a fundamentally different case. In this region the upper boundary of the OMZ is considerably deeper than in the areas discussed above (Fig. 7E). Depth of the OMZ generally increases with latitude in the Gulf of California and never falls to OMZ levels north of the shallow sill separating the Guaymas Basin from the Midriff Islands [yellow herringbone line in Fig. 1; see also Fig. 2 in Alvarez-Borrego and Lara-Lara (Alvarez-Borrego and Lara-Lara, 1991)]. These features are consistent with the high daytime oxygen concentrations observed for squid in this area (tags 07/06A,B; Table 2) at a ‘normal’ daytime median depth of ~250 m. Time-series data in Fig. 2C suggest that this squid entered the OMZ only once (deep dive on 08/02) during the entire 21-day deployment, because the pop-up location of the tag was well north of the sill.

Comparison of the median daytime depth of tagged squid with depth of the upper boundary for the OMZ (Fig. 7F) shows that these parameters are very similar in some cases, but not in others. Points for all autumn tags (Santa Rosalia) and one spring tag from Guaymas (03/07A) lie on or slightly above the dotted line of unity slope, whereas the other spring (03/07B and 05/08) and both summer tags (07/06B,C) do not. Daytime depths for these squid are all shallower than the local OMZ depth, and in the case of the San Pedro Martir tags the difference is extreme. Nonetheless, median daytime depth of all these tags would be classified as generally hypoxic by the 60 μmol kg⁻¹ criterion that we use to define the OLZ.

Overall, tagged squid spent nighttime at a median depth of ~60 m and a temperature of 19°C, a region with high oxygen concentration (~145 μmol kg⁻¹) (Table 2). Sizeable differences between individual squid in median nighttime depths and temperature reflect the strong seasonal variation in surface temperature in the Gulf of California. Lowest median nighttime temperatures and shallowest depths occupied by squid occur in spring (03/07A,B and 05/08) when surface temperatures are cool (Table 2, Fig. 8A), and highest temperatures and greatest depths occur during summer when surface temperatures reach a maximum.

From early summer through late autumn, surface temperatures in the Guaymas Basin exceed 25°C, but the squids' median nighttime temperature never exceeded 21°C (Fig. 8A), suggesting a potential thermal limit near this value. This idea is consistent with the amount of time individual squid spent exploring the upper 10 m of the water column in different seasons (Fig. 8B). Mean temperature in this depth zone was cool in spring, and the average duration of an excursion was ~3 h, with some lasting over 6 h. In summer, when surface waters were nearly 30°C, the longest excursion was <10 s. Occupancy of the upper 10 m also tended to be low in autumn, when temperature generally exceeded the thermal limit suggested above. Only one excursion lasting longer than 37 min was observed at temperatures greater than 21°C (7.7 h at 22.1°C).

PAT tags used in our study have a sensitive light sensor that permits a relative measure of illumination on a logarithmic scale. Fig. 9 (left panels) shows complete depth (upper, blue) and light (lower, gray) records for the tags discussed in conjunction with Figs 2 and 6. Arrows in each panel indicate the day on which the tag detached from the squid and began floating at the surface, with maxima and minima after this time representing day and night, respectively. Panels on the right show the distributions of time spent at different light levels for the tagged squid and floating tag. In general, light levels recorded by the tagged squid during both day and night are similar, and both are substantially lower than the light level at the surface at night. Variability in nighttime light levels for floating tags reflects the lunar phase; a transition from full moon on 9 August 2006 to new moon on 23 August 2006 is evident in Fig. 9Ci. All tags analyzed in this paper showed similar patterns for day and night light levels.

Although the tagged squid spent the vast majority of their time under conditions of extremely dim light, vertical excursions to levels of illumination characteristic of daylight are evident for each data set in Fig. 9. These excursions tend to be brief (Fig. 9Ai) and often occur over several sequential days (Fig. 9Bi,Ci). During these times, high-frequency vertical movements can obscure the regular diel pattern; this is most obvious in the data shown in Fig. 9Ci. This behavioral pattern is seen during both cold (March; Fig. 9Bi) and hot (July, Fig. 9Ci) seasons, and it occurs during daytime as well as nighttime (compare Fig. 9Bi,Ci with Fig. 2B,C).

Four tagged squid (03/07A,B and 11/08A,B) spent time in the upper OMZ (Fig. 2, Table 2) in an area where we made extensive oxygen surveys before and after tagging (Fig. 7). Vertical velocity distributions from these tags were analyzed for three specific depth zones, each 100 m thick, that are relevant to the effects of hypoxia on swimming performance: (1) the well-oxygenated ‘surface’ layer (0–100 m depth); (2) the seriously hypoxic ‘OMZ’ (100 m layer immediately below the 20 μmol kg⁻¹ oxygen contour); and a 100 m layer just above the OMZ that roughly corresponds to the OLZ (20–60 μmol kg⁻¹ oxygen; see Fig. 7C,D). Overall velocity distributions for these three zones are plotted in Fig. 10A. Ratios calculated for OMZ and OLZ values relative to the surface zone are plotted in Fig. 10B.

Vertical velocity can be divided into three categories (see Fig. 10A) based on comparisons between the velocity distributions for different depth zones: low (−0.1 to +0.3 m s⁻¹), intermediate (up to ±1 m s⁻¹) and high (>±1 m s⁻¹). Fractions of time spent in the low velocity range are similar across the three depth zones, and in each case this range accounts for >90% of the total distribution. These low velocities are associated with the most gentle vertical movements, including passive sinking, which alone accounts for 80% of the total time (−0.15 to +0.05 m s⁻¹).

Fractions for positive intermediate velocities are comparable in the surface zone and the OLZ, but OMZ fractions are approximately half of these (Fig. 10B). Negative velocities in the intermediate range tend to be greatest in the OLZ, although there is some variability (Fig. 10B). OMZ fractions are again reduced in this region relative to both of the other layers. Thus, vertical movements in this intermediate velocity range are substantially slowed in the OMZ. This velocity range accounts for 7.2% of the total time in the surface zone, 9.3% in OLZ and 4.1% in OMZ. Together, the low and intermediate velocity ranges account for >99% of all time spent in each zone.

Maximum velocities in the surface zone were −2 and +2.7 m s⁻¹, and the total range of velocities was progressively reduced in the other zones: −1.8 to +2.0 m s⁻¹ in the OLZ and −1.2 to +1.7 m s⁻¹ in the OMZ (Fig. 10A). Direct comparison of fractions of time spent at high velocities across depth zones is not straightforward, because these events are extremely rare and account for <0.02% of the total time. In order to assess higher velocities in another way, we counted the number of times each tagged squid exceeded a certain velocity threshold, regardless of how long the squid remained above that threshold. This provides a number of discrete ‘events’ when velocity exceeded the threshold value. Fewer than 10 events in each depth zone are responsible for all velocities exceeding −1.0 or +1.5 m s⁻¹ (Fig. 10C), and the relative number of high-velocity events is reduced in the OMZ, particularly for negative velocities. This approach is also informative about intermediate velocities. It is evident from inspection that the number of events (Fig. 10D) associated with the intermediate velocity range in the upper OMZ is also generally approximately half of the corresponding values in the OLZ or surface zones. This pattern is similar to that discussed in conjunction with Fig. 10B.

Vertical velocities during depth-maintenance behavior in all depth zones analyzed were confined to the low and intermediate ranges (Fig. 10E). Total time spent in depth-maintenance behavior in the surface zone was 16%, and values in the OLZ and OMZ were 5 and 20%, respectively. This range is similar to that computed for all depth maintenance during an entire deployment (3.4–18.5%). Velocities in the intermediate range associated with this behavior were greatly diminished in the OMZ and less so in the OLZ (Fig. 10F). The greater reduction of intermediate velocities in the OMZ versus OLZ is thus qualitatively similar to the pattern described above for overall jetting performance.

Previous studies have indicated that D. gigas spends considerable time at cold, hypoxic depths, both in the Gulf of California (Gilly et al., 2006) and in the Pacific Ocean (Bazzino et al., 2010; Zeidberg and Robison, 2007). This paper provides an expanded and refined assessment of this feature by exploring squid swimming behavior in the Guaymas Basin of the Gulf of California in locations that are characterized by large seasonal differences in surface temperature and depth of the upper boundary of the OMZ. This paper also presents the first data using an AVP with squid and provides insight into the basic patterns of jet-propelled locomotion in vivo from an oceanic squid, a group that has not been previously examined from this point of view. These insights into swimming behavior provide a foundation for evaluation of the effects of hypoxia on performance.

Data from both AVP and PAT-tag deployments are consistent with the idea that nearly all locomotor behavior of D. gigas is carried out through climb-and-glide swimming. Temporal summation of elemental climb-and-glide events underlies a rich assortment of vertical movements, and the strength of an elemental jetting event (as defined by its time course) can also vary. Neural mechanisms for control of jetting strength in D. gigas are presumably similar to those in a loliginid species, where jets of similar time course can vary over a large amplitude range depending on the motor pathway recruited and the intensity of neural activity (Otis and Gilly, 1990; Gilly et al., 1996). Another mechanism for varying the strength of individual jets in D. gigas would be active control of siphon diameter (Fig. 7H), but we cannot quantify siphon dynamics in free-swimming squid at present.

Gliding appears to account for the vast majority of a squid's time. Over 80% of the total time with PAT tags was spent at vertical velocities of −0.15 to +0.05 m s⁻¹, a range that includes passive sinking (Fig. 10A). Clearly, the vast majority of a squid's time is spent ascending slowly, presumably with weak jetting and fin beating, and descending even more slowly by gliding.

Based on the simple model for climb-and-glide swimming presented, Eqn 4 predicts a net horizontal velocity during passive gliding of ~0.25 m s⁻¹, and the total amount of time spent in the above velocity range would correspond to a total horizontal motion of ~22 km day⁻¹. This value is comparable to a figure of 30 km day⁻¹ estimated for D. gigas based on PAT tag deployments (Gilly et al., 2006) and is consistent with estimates of maintained horizontal speed during long-distance migrations of other oceanic squids (Dawe et al., 1981; Nakamura, 1993; O'Dor et al., 1994; Yatsu et al., 1999; Yano et al., 2000; Stark et al., 2005). Thus, it appears that D. gigas may generate a substantial fraction of total horizontal movement through passive gliding. Obviously weak horizontally directed jetting or fin activity during gliding could greatly extend total horizontal movement.

Virtually all of the additional 20% of a squid's time is spent actively ascending and descending at intermediate velocities of <1 m s⁻¹ magnitude. These motions can be quite complex, but are essentially composed of temporal summation of individual jetting events of varying strength and frequency. Because climb-and-glide swimming is also involved in these more active motions, it is likely to be an important component of most behaviors.

Fast vertical movements of >1 m s⁻¹ are extremely rare (Fig. 10C), accounting for only 0.2% of time in the upper 100 m and even less in the other depth zones analyzed (Fig. 10A,C). Maximum vertical velocities can exceed 3 m s⁻¹ (data not shown, but see Fig. 10A), but the functional importance of these rare and brief events is not known. Recent measurements using active acoustics (5–10 Hz sampling) of free-swimming D. gigas engaged in coordinated, nighttime foraging behavior have revealed short bursts (<1 s) of activity with vertical velocities approaching 4 m s⁻¹ and horizontal velocities of nearly 30 m s⁻¹ (Benoit-Bird and Gilly, 2012). Such high burst speeds are comparable to the fastest teleost fishes, but it is highly unlikely that they could be maintained by a squid for very long.

Metabolic rates of swimming squid, mostly loliginids, have been estimated in flumes (O'Dor and Webber, 1986; O'Dor et al., 1994; Bartol et al., 2001a; Bartol et al., 2001b), and such studies have confirmed that swimming by jet propulsion is much less efficient than tail-beating in fish (O'Dor, 1982; O'Dor, 1988; O'Dor, 2002; Webber et al., 2000). A large fraction (30–90%) of a squid's total energy expenditure in a flume is spent in maintaining horizontal position, and if climb-and-glide swimming were possible, a correspondingly large energy savings might be realized (O'Dor, 1988). Because passive sinking (gliding) accounts for ~80% of total time of a tagged squid in the natural oceanic setting, it is clear that climb-and-glide behavior by free-swimming D. gigas must indeed save a great deal of energy compared with horizontal swimming against a current in a physically constraining flume.

Flow-tank studies of active swimming by an oceanic squid such as D. gigas are therefore likely to overestimate metabolic demands to achieve a given horizontal velocity in an ecologically relevant context. If climb-and-glide locomotion were possible in such studies, the overall efficiency of jet-propelled swimming of a large Humboldt squid would probably be much closer to that of comparably sized teleosts than previously thought. Use of the fins by D. gigas in swimming would also tend to increase efficiency. But negatively buoyant, oceanic fish such as yellowfin tuna also employ climb-and-glide swimming (Schaefer et al., 2007), so locomotion in vivo for such teleosts would probably still be more efficient.

Depth maintenance (defined here as holding depth within a range of ±15 m for a period of ≥15 min) accounted for up to 20% of the total time at liberty for squid carrying PAT tags. This behavior generally involved simple patterns of climb-and-glide swimming and was always restricted to the low and intermediate vertical velocities that account for over 99% of the total time spent in locomotion.

It seems likely that depth maintenance is a generic pattern of activity that is expressed at a particular time and place when the context warrants careful attention of a specific region of the water column. One context is engagement in a variety of social interactions with conspecifics (Fig. 5), including mating. Another likely function would be foraging on patches of densely compacted micronekton, which can occur anywhere in the water column. Once such a rich resource was located, it is likely that a particular depth suitable for exploitation would be focused on through repetitive, small-amplitude climb-and-glide events.

Low oxygen concentrations found in the upper OMZ (<20 μmol kg⁻¹) have been proposed to limit the daytime depth occupied by D. gigas (Gilly et al., 2006; Rosa and Seibel, 2010), and data presented in this paper are to some extent consistent with this position. Comparison of the vertical distribution of squid determined with PAT tags with extensive hydrographic surveys in the areas of tag deployment indicate that daytime habitat (defined by median daytime depth) can correspond closely to the upper boundary of the OMZ in Santa Rosalia in autumn and Guaymas in spring (Fig. 7F). But in other locations and seasons this relationship breaks down, indicating that oxygen concentration is not the only relevant factor that determines preferred daytime habitat.

Decoupling of daytime habitat from the upper OMZ occurs in at least two ways. First, the OMZ can be substantially deeper than 250 m, either due to location (San Pedro Martir, tags 07/06B,C) or season (spring, tag 05/08), resulting in a daytime depth for squid at shallower depths and higher oxygen concentrations. Second, an individual squid may spend an unusually large fraction of time at shallow depths during the day, again yielding a median daytime depth that is shallower than the OMZ. This situation was observed in Guaymas (tag 03/07B) during spring when the upper water column was much cooler and had a higher oxygen concentration than in the autumn (Fig. 6A,B). These observations indicate that an association with the upper OMZ is not obligatory for large, adult squid, at least on a time scale of days to weeks.

Light is also a contributing factor to the daytime depth of squid. All tagged squid spent most of their time, both day and night, at depths that are darker than a moonless night at the sea surface. Diel vertical migration (Figs 2, 9) is clearly a relevant behavioral mechanism, but it is doubtful that light directly sets the vertical distribution of D. gigas. Numerous transient excursions to much brighter depths occur during both day and night, indicating that squid explore a variety of light levels. But light is likely to be a critical factor in determining the vertical distribution of the mesopelagic micronekton (fish, molluscs and crustaceans) of the DSL (Boden and Campa, 1967) on which D. gigas preys (Markaida et al., 2008). We hypothesize that D. gigas can forage on these organisms at depth during daytime, and that the vertical distribution of these prey organisms is the primary driver of the distribution of D. gigas.

In autumn and spring in the Guaymas Basin, the depths of the DSL and the upper OMZ boundary often both occur at ~250 m (K. Benoit-Bird, personal communication), and this depth would then set the daytime depth of D. gigas. When the OMZ is much deeper than 250 m, as discussed above, we hypothesize that DSL organisms on which D. gigas preys also decouple from the OMZ and remain at their preferred light level at 250–300 m, along with the squid that can utilize the DSL as a daytime foraging ground.

We have not seen a case in the Guaymas Basin where the OMZ is significantly shallower than the median daytime depth of squid as documented in this study (200–250 m; Table 2). In the southern Gulf of California (below ~25.5°S) the upper boundary of the OMZ can be much shallower (Alvarez-Borrego and Lara-Lara, 1991) (W.F.G., unpublished data), but tag data from this area are not available. If a shallow OMZ does present a limit to daytime depth at certain places and times, it is not entirely clear whether the directly impacted species would be D. gigas or some members of the DSL community on which D. gigas forages. Acoustic studies to simultaneously quantify vertical distributions of both the DSL and squid in conjunction with spatial or seasonal variation in OMZ depth, along with biochemical and physiological studies on both squid and their prey species would clarify this issue.

Aversion of adult D. gigas to temperatures of >22°C was previously proposed based on an analysis of diving dynamics (Davis et al., 2007). Tagging data from the present study support this idea and suggest an upper thermal limit of 21–22°C for median nighttime depth of D. gigas (Fig. 8A). When surface temperatures are cooler than this limit, considerable time is spent at shallow depths, with excursions into the upper 10 m of water column lasting up to 12 h (Fig. 8B). But very little time is spent there when the surface temperature exceeds this limit, and excursions are extremely brief. These observations are consistent with opportunistic foraging on small epipelagic fishes such as anchovies when surface temperatures are permissive in spring (Markaida et al., 2008), and brief forays into warm surface waters are also probably related to foraging. Physiological limitations are likely to underlie these observations, because the temperature dependence of resting metabolic rate in this thermal range is extremely high (Rosa and Seibel, 2008; Rosa and Seibel, 2010).

Tagged squid can spend a substantial amount of time, primarily during daytime, moving in and out of the upper OMZ (Fig. 2A,B). We hypothesize that foraging on myctophids and other DSL organisms can occur during these times. Maintained activity under hypoxic conditions raises some questions, because laboratory studies of D. gigas have revealed a large suppression (>80%) of resting aerobic metabolic activity under temperature and oxygen conditions that mimic those of the upper OMZ (Gilly et al., 2006; Rosa and Seibel, 2008). Dosidicus gigas has not been studied swimming under OMZ conditions in the laboratory, but fast swimming for any substantial period of time would almost certainly be greatly impaired (Rosa and Seibel, 2010).

Vertical velocity in the field provides an indicator of performance that can be used to assess the effects of cold temperatures and hypoxia as found within the OMZ. Results in this study indicate that low-velocity movements (−0.1 m s⁻¹ to +0.3 m s⁻¹) are not impacted by OMZ conditions, but ascending and descending velocities in the intermediate range (Fig. 10A,B) are reduced by ~50% in the upper OMZ relative to both of the shallower zones analyzed (OLZ and oxygenated surface). Maximum observed velocities and the number of high-velocity events (>1 m s⁻¹ magnitude) are reduced even more severely in the upper OMZ (Fig. 10C,D). Oxygen is likely to be the limiting factor underlying the diminished vertical velocities in the OMZ relative to the OLZ, because the temperature differential between these layers is rather small.

Intermediate vertical velocities during depth-maintenance behavior are reduced to an even greater extent in the upper OMZ relative to the surface zone (Fig. 10E,F). A similar but less marked effect is evident for the OLZ. Whether depth maintenance serves the same functional roles in the different depth zones is unknown, but it is clear that velocities of >0.5 m s⁻¹ associated with this behavior are slowed down by at least 80% in the upper OMZ. This figure is comparable to the reduction in resting metabolic rate measured in the laboratory under these conditions.

Behavior of marine-mammal predators also suggests that D. gigas may be slowed down under OMZ conditions. Sperm whales in the Guaymas Basin in fall preferentially hunt near the upper OMZ during both day and night, possibly because escape responses and fast jetting of the squid are impaired, and the increase in probability of capture might outweigh the decreased probability of encounter (Davis et al., 2007). Data in the present study confirm that squid in the upper OMZ move much less rapidly than at more oxygenated depths, consistent with the mechanism proposed for this pattern of selective foraging. It remains unclear, though, to what extent escape responses are impaired under OMZ conditions. Escape-response behavior has not been studied in D. gigas, but the species possesses a specialized giant-axon system (Villegas, 1969) similar to that in loliginid squids where activation of giant axons is associated extremely powerful jets that are seldom, if ever, employed in conjunction with ‘normal’ locomotion like that studied here (Otis and Gilly, 1990).

Data in the present study show that D. gigas spends a great deal of time at midwater depths that are generally considered hypoxic (<60 μmol kg⁻¹ oxygen), and in certain places and times occupancy extends to the upper boundary of the OMZ itself where oxygen concentrations are <20 μmol kg⁻¹. Below this value, metabolic suppression reduces resting metabolism in D. gigas by ~80% at low temperatures (Gilly et al., 2006; Rosa and Seibel, 2008), and vertical velocities during rapid movements and depth-maintenance behavior in vivo are comparably reduced (Fig. 10). These results need to be evaluated in light of the fact that D. gigas must often be in the presence of a high concentration of micronektonic prey during the daytime at hypoxic depths.

Depth maintenance probably underlies much foraging behavior, and because of the relatively low vertical velocities involved, such activity might be metabolically feasible even in the upper boundary of the OMZ where metabolism is suppressed. The observed reduction of intermediate vertical velocities associated with depth maintenance in the upper OMZ – but not in the frequency of the behavior – is consistent with this idea. Depth maintenance thus appears to be an important activity that continues to be displayed under OMZ conditions but at greatly reduced vertical velocities. Such a strategy may be viable, because prey items encountered in this region are also likely to be lethargic. The lack of horizontal velocity information for D. gigas in nature represents a major limitation in our current knowledge.

It seems unlikely that the observed reductions in vertical velocity are severe enough to preclude consumption of lethargic prey in and just above the upper OMZ. ROV observations show that squid are capable of feeding under OMZ conditions (Zeidberg and Robison, 2007) (L.D.Z., unpublished data), but it remains unclear how often they do so, or whether they do so in the absence of artificial light. Low-light camera observations, particularly with an AVP approach, would provide an excellent way of documenting foraging at depth in areas of the Guaymas Basin (or further south in the Gulf) where the upper OMZ is close to (or even less than) the lower limit of the photic zone.

Daytime foraging at DSL/OMZ depths as suggested for D. gigas is hardly unique to this species. Many other top predators show repetitive diving behavior to depths occupied by the DSL during daytime (Dewar et al., 2011), including sperm whales (Davis et al., 2007), tunas (Schaefer and Fuller, 2010; Schaefer et al., 2011) and swordfish (Carey and Robison, 1981; Sepulveda et al., 2010; Dewar et al., 2011), all species that prey on or compete with D. gigas. Although utilization of this rich daytime foraging ground may be a common feature of many top pelagic predators, only D. gigas appears to be able to tolerate the relevant cold, hypoxic conditions for several hours or more. Thus, the DSL/OMZ environment would appear to greatly favor D. gigas, and it is clearly an environment that this species inhabits and utilizes.

We thank Barbara Block and John O'Sullivan (Monterey Bay Aquarium) for continued support and access to Argos and data-archiving servers, Laurent Giovangrandi for MATLAB programs, Jorge Ramos-Castillejos for CTD casts in Santa Rosalia, Patrick Daniel for help with analyzing velocity distributions, and Sal Jorgenson, Gareth Lawson, Kelly Benoit-Bird and Brad Seibel for valuable discussions. We also acknowledge the invaluable assistance of captains and crews of all research vessels used: RV New Horizon (Scripps Institute of Oceanography), RV BIPXII (CIBNOR, Guaymas, Mexico), RV Pacific Storm (Bruce Mate, Oregon State University Marine Mammal Institute), the Marylee (John Barnes, La Paz, Mexico) and the Sand Man (Ron Steele, La Paz). We are grateful to Unai Markaida and Cesar Salinas for all manners of assistance in Mexico.

 
• AVP

  animal-borne video package

 
• DML

  dorsal mantle length

 
• DSL

  deep scattering layer

 
• OLZ

  oxygen limited zone

 
• OMZ

  oxygen minimum zone

 
• PAT

  pop-up archival transmitting

 
• ROV

  remotely operated vehicle