Seasonal patterns in deep acoustic backscatter layers near vent plumes in the northeastern Pacific Ocean

Hydrothermal plumes at Endeavour Ridge originate with five major vent fields and multiple sources of diffuse venting within the axial valley of the ridge (e.g., Thomson et al. 2003). The five primary high temperature vent fields (Salty Dawg, High rise, Main Endeavour, Mothra, and Sasquatch) are located at around 2200 m depth near 47°57′N, 129°06′W (Fig. 1; Table 1). All sites are within the axial valley that splits the ridge and all are within the Canadian Endeavour Hydrothermal Vents Marine Protected Area. The high temperature vents and numerous other smaller low temperature vent features in the valley are thoroughly described by Kelley et al. (2012). Temperature and transmissometer profiles at the ridge, along with numerical simulations using nested-grid ocean circulation models (Thomson et al. 2005, 2009) reveal that the rising plumes in the axial valley, augmented by diffuse venting, spread laterally at their level of neutral buoyancy at around 1900–2100 m depth about 200 m above bottom (mab). Plumes drift both along and across the ridge axis with the prevailing currents (Thomson et al. 2003, 2005, 2009; Xu and Lavelle 2017). Plume remnants are also evident outside the confines of the axial valley and ridge complex, where water depths are much greater (Thomson et al. 1995). Other noteworthy features include a small volcano (Summit Volcano) located near the northern end of the west ridge of the axial valley and a small active vent site on the opposite side of the valley from the volcano (Kelley et al. 2012). Water depths are <2000 m over the top of the volcano. Both the topography of the volcano and the hydrothermal fluids from the vent site are expected to affect currents at the northern end of the valley.

The towed instrument package used in the historical 1991–1996 biological surveys—consisting of a 153 kHz Teledyne-RDI ADCP, seven opening/closing Tucker trawl nets, a Guildline conductivity, temperature, and depth measurement instrument (CTD), and a transmissometer—as well as data processing, interpretation methods and zooplankton communities throughout the water column are described in Thomson et al. (1991, 1992) and Burd and Thomson (1993, 1994, 1995, 2012, 2015). The ADCP had a downward-looking orientation and the processed data were mainly confined to downcasts to minimize disturbance to flow past the ADCP and CTD. Tows were conducted in July or August at various times of the day and night. Results from this earlier work provide context and descriptions of the deep scattering layer composition near the vent plumes in summer.

Time-series ADCP records from two separate sets of moorings from different years anchored at depths of 2110 to 2270 m were utilized to examine backscatter patterns over the lower water column (1375–2200 m depth) in the Endeavour Ridge axial valley. The mooring sets are from the Seabreeze project (ER03-ER05) (Institute of Ocean Sciences, Fisheries and Oceans Canada, Sidney, British Columbia; University of Washington, School of Oceanography, Seattle, Washington State) and Neptune Canada (RCM) (now a component of Ocean Networks Canada, University of Victoria, Victoria, British Columbia). Deployment dates, water depths and depth ranges of the ADCPs are shown in Table 1. All moorings included an upward-looking 75 kHz RDI-Teledyne Workhorse ADCP positioned at the top of the mooring. The four beams of the ADCP were at 30° angle to the vertical, and at 90° angles to each other. The ADCPs used during the Seabreeze project recorded a total of 64 bins with bin lengths of 8 m, whereas those used in the Ocean Networks Canada program recorded 128 bins with vertical bin lengths of 4 m. The lateral ensonification spread for the ADCPs used in both projects (calculated as bin length × bin number) reached a maximum of 512 m at a distance of 512 m from the transducer head. Distal bins with low signal/noise backscatter ratios were excluded from analyses. The records from Bin 1 adjacent to the transducer head were also excluded as the data were contaminated by high frequency transducer noise (data from Bin 1 were invariably inconsistent with data from the rest of the profile). There was also an acoustical “blanking range” of 1 m between Bin 1 and the transducer.

Because the ADCPs were subject to changes in orientation caused by current-induced torque on the moorings, time series from certain beams occasionally showed interference patterns that are likely associated with delayed secondary reflections of transmitted acoustic waves from the nearby valley walls. Although wall interference may have affected current velocity measurements (derived using Doppler frequency shifts of the acoustic signal from each beam; cf. Thomson and Emery 2014), the interference appears to have had a minimal effect on the acoustic backscatter intensity. To ensure that the signal from a disrupted beam is not overly emphasized, we averaged the backscatter data from all four beams. As illustrated by Fig. 2, there was little overall difference in the backscatter patterns among the separate beams.

The moorings can be grouped into three geographic clusters (Fig. 1):

ONC moorings are now operating in the southern and central Axial Valley but the time series are presently too short for seasonal analyses.

Based on results from combined net/ADCP profile data collected during the 1990s (Burd and Thomson 1994, 1995), acoustic backscatter profiles obtained from the upward-looking ADCPs mounted at 250 mab in the Neptune moorings may have partially missed some of the plume-related deep scattering layers. Conversely, acoustic profiles from the Seabreeze moorings beginning at 6 mab provide details of the scattering layers above and below the horizontally spreading plume that typically lies between 1700 and 2200 m depth, but will have missed migratory patterns farther up the water column. By combining results from the two mooring studies and the detailed net zooplankton data available from the earlier shipboard surveys, it is possible to generate a more complete annual cycle of the scattering layers from roughly 1375 m depth to near-bottom, with the underlying assumption that the deep scattering layer zooplankton patterns remained similar in character (if not detail) over the decade between the moored Seabreeze and Neptune ADCP measurements.

Because the short-term passive transport of zooplankton by the relatively strong (1–10 cm/s) tidal currents and other high frequency motions tend to blur active longer-term migratory shifts in backscatter features, a 30-h low-pass Kaiser-Bessel filter (cf. Thomson and Emery 2014) was applied to all backscatter time series to remove high-frequency variability. The smoothed time series were then reduced to 24-h intervals. The resulting daily time series of backscatter intensity for each bin was then averaged and a mean vertical (“background”) profile generated for each time series. To construct backscatter anomaly profiles, the mean backscatter value for each bin was subtracted from the daily mean value for that bin. This removal of the vertical background structure is meant to compensate for the well-known, but here undetermined, attenuation in acoustic signal strength due to spherical spreading and absorption of the acoustical energy with distance, R, from the ADCP (e.g., Thomson et al. 1991; Deines 1999; Gostiaux and van Haren 2010). Although this process can cause the acoustic backscatter from persistent, near-stationary backscatter layers to be blended into the derived background structure, the effect is mitigated by the marked temporal variability in the layer depths and distributions. As a consequence, acoustic backscatter anomalies associated with zooplankton aggregations ensonified by the moored ADCP are reduced in amplitude but not fully “lost” to the background profile estimates. A more sophisticated method for extracting background acoustic profiles from moored ADCP backscatter time series with steady transducer transmit power is presented in Gostiaux and van Haren (2010). As we are interested in relative temporal changes rather than absolute changes, the methodology has not been applied.

The Seabreeze program (www2.ocean.washington.edu/seabreeze/sb.pdf) at the Endeavour Segment of Juan de Fuca Ridge was a component of the RIDGE 2000 Integrated Study Site (ISS). Metadata and information on the mooring locations and history can be found at www2.ocean.washington.edu/seabreeze/sb03/locations and in Table 1.

The Seabreeze instruments were self-contained and powered by internal batteries. Each deployment period lasted for about 1 year (Table 1) and consisted of an upward-looking 75 kHz ADCP positioned 6 m above the seafloor in water depths ranging between 2110 and 2270 m. ADCP data were recorded between July 2003 and March 2006. For all deployments, there was a slight loss in ADCP power output over time, leading to a decrease in the signal-to-noise ratio for the acoustic backscatter time series. This effect was most pronounced for acoustic bins farthest from the instrument, with the range of backscatter counts declining by 0.5% near the instrument to 14% at the most distant (shallowest) bin over a typical 12-month deployment. Although the decrease in power output could often be modeled using a 3rd order polynomial (Fig. 3), attempts to fully correct for power loss were hindered by irregular temporal adjustments in the ADCP automatic gain control or other unknown factors associated with the instrument. This effect is clearly illustrated by Fig. 3, which shows the declines in ADCP voltage output (blue line) and backscatter intensity over time for selected bins.

Our inability to accurately correct acoustic backscatter time series for instrument power loss made it impossible to derive absolute backscatter time series for within-deployment and between-deployment Seabreeze mooring periods. However, because the decrease of instrument power over time was very gradual, short-term changes in backscatter intensity are unlikely to have been affected in any substantial way. As we are mainly interested in seasonal and intra-seasonal changes in backscatter intensity and distribution, the next step was to estimate the effect of power loss by smoothing the daily anomaly series with a 90-d low-pass filter. Estimates of the backscatter anomaly time series for time scales <3 months were then calculated by subtracting the low-pass filtered record from the daily averaged data for each bin. The results help mitigate bias from loss in instrument power over time. As discussed in the previous section, acoustic backscatter anomalies associated with zooplankton aggregations are partially reduced in amplitude by subtraction of the bin-averaged values.

The first set of ONC moorings at Endeavour Ridge (sites RCM-NE and RCM-NW) included time series of acoustic backscatter from upward-looking 75 kHz ADCPs obtained via the ONC web-based data access portal. The Endeavour Ridge segment of the cabled observatory is described at oceannetworks.ca/introduction-endeavour and in Table 1. The ADCPs were moored at approximately 1900 m depth at both locations and the original data series sampled at intervals from 2 s to 15 min. For consistency among the records, the more rapidly sampled series were re-sampled to 15 min by averaging over each 15-min interval. Because the moorings were powered through the cabled network, there was no ADCP power output loss over time.

The ample power available from the ONC network made it possible to set the ADCPs to the high-power mode. An unexpected consequence of this decision was that acoustic backscatter records from the two instruments were contaminated by cross-interference. This interference, which gave rise to anomalously high backscatter intensity for periods up to 1.5 h, was prevalent from the beginning of October 2011 to 15 February 2012, after which an adjustment in instrument settings was made to considerably reduce the problem. Data collected during the interference periods were “cleaned” using an acoustic threshold level determined individually for each series as the 100th highest value in the series. This method successfully removed many of the larger interference “spikes”. Although some weaker interference spikes (with peak values below the specified threshold) still remained in the series, calculation of the backscatter anomalies by subtraction of the time-averaged (background) backscatter intensity value for each bin removed any remaining low amplitude interference. However, because of a power output adjustment made to the ADCP settings on 15 February 2012 by ONC, the time-averaged background backscatter intensity also shifted. Therefore, the backscatter anomalies were calculated by subtracting time-averaged backscatter counts for each bin over the deployment period up to 15 February 2012 at 2000, and then separately for the remainder of the deployment. As with the Seabreeze data, acoustic backscatter anomalies are partially reduced in amplitude by subtraction of bin-averaged values.

Vent plume distributions in the study region consist of output from the major vent fields and diffuse seafloor venting sites within the axial valley of Endeavour Ridge, and output from isolated seafloor venting features to the east of the ridge in Cascadia Basin (Thomson et al. 2009). A compilation of transmissometer profiles from a series of summer tows between 1991 and 1996 in the general vicinity of the ridge (Fig. S1) shows that, following their coalescence, the rising particle-laden thermal plumes from the major vent fields reach a level of neutral buoyancy several hundred mab and then spread laterally in and around the axial valley. The level of neutral buoyancy over the ridge depends on the water depth and initial buoyancy of the plumes, but is typically around 1900–2200 m depth. These plume heights are less than those of mega-plumes generated by major seafloor eruptions, which can rise as high as 1000 mab (e.g., Baker et al. 1989; Dziak et al. 2007) as noted in Tow 95G (Fig. S1).

Figure 5 shows selected transmissometer profiles from summer net tows conducted from 1991 to 1996. The profiles are representative of historical conditions proximal to the three mooring deployment areas in the axial valley (north, middle, and south; see Fig. 1). For the northern group, the plume depth ranged between 1600 and 2200 m, depending on bottom depth, with evidence of suspended particulates extending all the way to the seabed. In the center of the axial valley, the plume depth was more confined, ranging from around 2000 to 2100 m depth. The southern grouping tended to have variable plume depths in the depth range 1850–2200 m.

Because mooring N2 is common to all three Seabreeze deployments, we have used data from this site to represent temporal changes in acoustic backscatter intensity within the Axial Valley. The upward-looking ADCPs on mooring N2 revealed acoustic backscatter layers from approximately 1700 m to near-bottom depths (Fig. 6). Supplementary Material 2 includes the daily anomaly backscatter contour plots for all nine Seabreeze deployments. The strongest backscatter anomalies occurred above 2100 m depth, the approximate depth of the laterally spreading plumes, but varied considerably in depth and intensity over time. In some cases, a smaller backscatter peak was observed within the plume or below it. These smaller peaks sometimes appeared concurrently with the stronger anomaly above the plume (e.g., deployments ER03-S1 and S3, ER04-N3, and ER05-N2).

At times, we observed a number of fragmented and diffuse backscatter layers spanning the entire depth range ensonified by the ADCP, whereas at other times clearly distinct and vertically compact positive anomalies were observed separated by negative or low positive anomalies (indicative of relatively clear water). For example, deployment ER04-N2 (Fig. 6a) showed anomalies above and below the plume in early July 2004 and 2005. A similar seasonal pattern was evident in other deployments (Supplementary Material 2), including ER03-N2 late summer, ER04-N2 fall and spring ER05-N2 fall, ER04-Y2 late fall, early summer, ER03-S1 early summer, and ER04-N3 late fall.

The backscatter layers identified by ADCPs on the Seabreeze moorings showed signs of seasonal/ontogenetic vertical migration, a pattern evident to a greater or lesser extent in all of our ADCP deployments. For example, during mid-winter (January–March), the ADCP records show generally much lower overall backscatter anomalies in the deeper portion of the water column, with most of the scatterers positioned at depths <1600 m. Similarly, the swathes of positive backscatter anomaly observed at depths shallower than 1900 m in early fall and late spring are indicative of vertical seasonal zooplankton migration. Figure 6b presents a detailed 50-d duration contour plot of the backscatter anomalies for ER04-N2 from August 2004 through mid-October 2004. The intense scattering layer found just above the plume at 1950 m depth in August is typical of what has been documented in towed summer ADCP profiles from 1991 to 1996 (Burd and Thomson 1994, 1995, 2015). However, by September, this layer appears to have split into two component parts that moved in opposite vertical directions. One layer moved upward in the water column and seemed to disappear above the range of the ADCP, whereas the other layer moved downward, well below the depth of the neutrally buoyant plume.

Comparison of acoustic backscatter time series for different mooring locations within a given year show a striking similarity in the backscatter anomaly pattern above the ridge crests and laterally spreading plume (Fig. 7a), but not below the plume (Fig. 7b). In particular, moorings N2 and N3 had similar temporal patterns, with a slight offset in time between the patterns at those sites at the more northerly mooring, Y2. This suggests that epiplume zooplankton movements tend to be closely synchronized over the entire axial valley, regardless of local differences in plume conditions. Conversely, backscatter anomalies tended to be uniformly low and more randomly distributed near the bottom (>2200 m depth) amongst mooring locations.

A surface plot of the ER04-N2 backscatter anomaly time series (Fig. 8; same data as Fig. 6a) shows periodic peaks and troughs in the backscatter anomaly values throughout the measured depth range. The cyclic pattern has a periodicity of 20–40 d and is present in all contour plots from the Seabreeze data (Fig. 6, and see Supplementary Material 2), but is most clearly visible in surface plots. The pattern is most intense in the 1800–1950 m depth range, but is also present below plume depth. Minimal or reduced intensity of the cyclical patterns are evident in the near-bottom layers below plume depth (>2100 m depth). In aggregate, the patterns are clearly suggestive of temporal shifts in the concentration and dispersal of organisms over the visible acoustic depth range (~1375–2000 m; Table 1) ensonified by the combined set of ADCPs.

Because the upward-looking 75 kHz ADCPs on the ONC (formerly Neptune) moorings were positioned 250 m above the seafloor, the juxtaposition of the deep scattering layers (DSLs) to the hydrothermal plume is only partially observed in the backscatter data. In contrast, the acoustic records from these moorings extend much farther up the water column than the Seabreeze observations, enabling us to examine migrational behaviour in the backscattering layers over the depth range of roughly 1375–1900 m. The backscatter anomalies for the northeast ONC deployment (Fig. 9a) show an aggregation of fauna from 1375 to 1600 m depth in late winter through late spring 2011. This was followed by low backscatter anomalies throughout this depth zone from early summer through early fall 2011. Backscatter layers reappeared above 1500 m depth in mid-fall 2011 and intensified through the winter. By early spring 2012, backscatter anomalies were small throughout the depth range covered by ONC mooring ADCPs. Dispersed backscatter layers re-appeared in mid-spring 2012 and became more pronounced towards late spring (near the end of the deployment). The dispersion of backscatter anomaly layers at different depths suggests that the fauna do not all migrate at the same time, but rather migrate vertically in pulses. The backscatter time series from the shorter northwest ONC deployment (Fig. 9b) shows the same basic pattern as the northeast ONC deployment. Despite the roughly 1 km separation between the moorings, the similarity between the patterns is striking.

As with the Seabreeze moorings, surface contour plots of the backscatter anomalies for the northeast ONC mooring (same data as Fig. 9a) reveal a cyclical pattern of peaks and troughs that is most evident for time-depth regions with higher backscatter anomalies (Fig. 10). This pattern also occurs in the northwest ONC mooring, indicating that horizontal or vertical movement of the organisms occurs on a regular basis with a period of 20–40 d.

Based on biological studies of the North Pacific (e.g., Goldblatt et al. 1999; Kobari and Ikeda 2001a, 2001b), the large-scale seasonal shifts in depth of the backscatter layers we observed above the vent plumes are consistent with expected life-cycle migratory patterns of the dominant upper ocean copepods and their predators. In particular, an upward movement of the DSLs observed above the level of neutrally buoyant plumes was evident in late summer and fall, resulting in concentrations of fauna at much shallower depths in winter months. This is synchronous with the timing for adult maturation and reproductive incubation of the major migratory copepod taxa, which are non-feeding. A downward movement of backscatter layers towards the depth of the neutrally buoyant plume occurred in late summer/early fall, at a time when the pre-adult copepods are known to migrate to greater depths prior to reproductive development (Kobari and Ikeda 2001a, 2001b). Net samples obtained from tows near the vent plumes in the summers of 1991 to 1996 show that the dominant copepods are mainly stage V (pre-adults) (Burd et al. 1992). The active migration of late stage copepods from the upper ocean to deeper water prior to maturation is believed to minimize predation during the reproductive phase. At Endeavour Ridge, there is isotopic evidence that these late stage copepods resume feeding (Burd et al. 2002) near the neutrally buoyant plumes, presumably to increase their reproductive output (Burd and Thomson 1995, 2015).

In accordance with typical life cycles of the dominant migratory copepod species, it is expected that overall water column biomass should progressively decline from early fall through early spring, as the reproductive copepod adults release eggs and die. This pattern has been noted in the northwestern Pacific (Kobari et al. 2003), but is not clearly obvious in our data. Although a decline in backscatter intensity was recorded in fall and winter 2003 and 2005 by the ADCPs on the Seabreeze moorings, there was no apparent decline in intensity in the fall and winter of 2004. Acoustic records from the upward-looking ADCPs on the ONC moorings showed that there were strong backscatter anomalies in the water column during fall and winter at depths too shallow for the Seabreeze moorings to detect. The only time of year that the near-plume backscatter layers were not present was late spring to early summer, when new copepod larvae undergo stage I–V molts in the upper ocean and very few of the adults below 1000 m depth are still alive (Kobari and Ikeda 1999, 2001a, 2001b). Thus, unlike previous studies, we found that there is no season during which the DSLs disappear completely from the lower water column near Endeavour Ridge.

There is no a priori reason to assume that the deep ontogenetic migrators remain at the same depth throughout spawning and senescence. We know from previous studies (Burd and Thomson 1994, 1995) that the predators of the copepods near Endeavour Ridge (chaetognaths, shrimp, jellyfish, and small fish) also move to deep waters in early summer following their prey, so there is no safety for copepods in staying at a single depth during reproductive development. Rather, they would likely continue to migrate vertically to avoid predators. This could explain why there is a clear upward movement of the scattering layers in early fall, away from plume depth and the food source (Figs. 9a and 9b). This movement may also explain why the scattering layers near vent plumes periodically appear to be spread-out in fall and early winter, a pattern which is particularly evident in data recorded by the shallower ONC moorings (Fig. 7).

Maturation and subsequent release of reproductive material from the adult copepods can occur over a period of about 6 months in late fall and winter, with the possibility that some species or portions of populations overwinter or have biannual life cycles (i.e., do not die off in winter; see Tsuda et al. 1999). It is unlikely that overwintering or biannual adults could survive for such long periods without feeding. At Endeavour Ridge, where there is a seasonal food source at the surface and a continuous food source near the vents, the life cycles of zooplankton may be altered from the expected pattern in the open Pacific Ocean. Certainly, results from this study suggest that deep aggregations of zooplankton occur all year round near the ridge, although species composition and life stages cannot be determined from this study.

As high-frequency variations in backscatter intensity have been filtered from the processed ADCP data using a 30-h filter, our focus is restricted to cycles in the backscatter structure at periods longer than a day. Towed ADCP profiles from the Endeavour Ridge venting region clearly show that the intense positive backscatter anomalies found just above the spreading plumes at the ridge in summer months are patchy, both in space and time (Burd and Thomson 1994, 1995). Based on the long-term (decadal) backscatter data analyzed in this study, we believe that this patchiness is consistent with the periodic, 20–40-d variations in backscatter concentration observed at all moorings (ONC and Seabreeze).

The considerable spatial synchronization of cyclical backscatter patterns over the geographic range of the different moorings is evidence that the 20–40-d temporal cycles have large spatial footprints and are not locally isolated over the ridge. At plume depth, the movements of the scatterers at sites N2 and N3 appear to be highly synchronized, whereas movements at the more northerly site (Y2) are slightly offset in time (by hours) and are less coherent than those to the south. This suggests that the fauna at the northern location are responding to slightly shifted exogenous cue(s) relative to the southern locations.

The backscatter layer overlying the hydrothermal plumes in summer is composed primarily of upper-ocean copepods that migrate ontogenetically along with their predators (Burd and Thomson 1994, 1995). However, community analyses also indicate that this DSL often also includes typical deep zooplankton fauna that are not known to have any significant seasonal migration patterns. Small-scale movements in these deep-sea fauna are expected to be opportunistic in response to local stimuli such as food or predation. Presumably these deep-sea fauna also congregate near the plume to feed.

Based on net sampling work from 1991 to 1996 (Burd and Thomson 1994, 1995), the upper ocean migratory fauna typically remain above the depth of the neutrally buoyant plumes. In contrast, the mooring data examined in this study, as well as the historical towed net-ADCP data, indicate the common presence of several distinct scattering layers near the vent plumes at Endeavour Ridge. These layers consist of upper ocean migratory fauna above the plume only, as well as non-migratory deep fauna, which range from just above the plume down to near-bottom (several hundred meters). The near-bottom deep-sea zooplankton may be actively swimming vertically, or may be subject to vertical displacements by internal tides or topographically trapped wave motions, but are seasonally persistent features.

A near-bottom (20 mab) zooplankton/micro-nekton layer has also been observed near Endeavour Ridge by direct submersible observations (Skebo et al. 2006), with concurrent net samples indicating that these are not benthic larval forms or vent-specific fauna. Much of this visible near-bottom biomass is shrimp, eelpouts, and rattail fish, with concentrations of smaller fauna evident in net samples. Skebo et al. (2006) also suggested that the near-bottom aggregations of fauna were absent where high temperature, high volume vents emerge (the major vent orifices), but aggregate away from the direct vent outlets, as found in the Seabreeze and ONC mooring locations, which were deliberately placed away from vent orifices.

In the towed ADCP/net studies near the ridge from 1991 to 1996 (Burd and Thomson 2012, 2015), it was assumed that near-bottom backscatter measured in bins distant from the downward facing towed ADCP were related to acoustic reflections from the seafloor. These near-bottom backscatter layers were sometimes more than 50 m thick. For example, towed ADCP profiles from 1993 show a near-bottom (2700 m deep) DSL (Fig. 11) at several low-temperature venting fields in the Cascadia Basin (Fig. 1 and see Thomson et al. 1995). This was mistaken for bottom reflection at the time, and it was concluded that there was no deep scattering layer in the region. The moored ADCP results in the present study suggest that this assumption may have been incorrect. Based on our analyses of the acoustic backscatter data from the bottom-mounted upward-facing ADCPs on the Seabreeze moorings, positive near-bottom backscatter anomalies in the region appear to be due to deep zooplankton aggregations.

Although seasonally migratory upper ocean fauna have not been found in deeper net tows away from the axial ridge, where there is no detectable venting (Burd and Thomson 1994), the deep-sea fauna have been found aggregated in scattering layers at plume depth up to 50 km off-axis to the northwest of the ridge axis. Thus, unlike the migratory fauna, the endemic deep fauna appear to be advected to considerable distances off axis by currents. One possible explanation is that the upper-ocean migratory fauna are more adept at staying near the axial valley and the plume-related food source than typical deep-sea fauna. However, it is also possible that deep-water fauna are not as dependent for food on the plumes as upper-ocean animals, and might even be avoiding competition or predation by staying farther away from the plume.

Although it was not the purpose of this study to examine the circulation at Endeavour Ridge measured by the ADCPs, it is likely that currents play an active role in the zooplankton movement and distribution in the region. As noted previously, passive vertical displacements of up to 100 m due to semidiurnal tides generated over the abrupt ridge topography are expected based on earlier current meter measurements in the area (Mihaly et al. 1998) and from numerical simulations of tidal motions over nearby Axial Seamount (Xu and Lavelle 2017). It is further likely that low-frequency currents modified by the ridge topography may be responsible for passive advection of the zooplankton within the study region. This is especially true for currents at periods of 3–6 d (Fig. 12), an energetic counter-clockwise frequency band with current speeds of order 0.01 m/s known to arise from trapped waves over the ridge (e.g., Cannon and Thomson 1996). Maximum horizontal water parcel excursions due to these motions are around 1–2 km in both the along- and across-ridge directions, indicating that the zooplankton could be retained in the vicinity of the ridge with limited active swimming. In contrast, the 0.01 m/s cross-axis currents observed at periods of 20–40 d, which may be due to long wavelength topographically trapped waves or to passing mesoscale baroclinic eddies, would lead to off-axis advection of the zooplankton to distances in excess of 10 km, making it difficult for the organisms to remain in the vicinity of the ridge without active swimming.

As a result, velocity variations over periods of days or less due to diurnal and semi-diurnal tides, wind-generated inertial currents, and trapped 3–6-d topographic waves, are commensurate with retention of zooplankton over the ridge. In contrast, motions at periods of weeks to months due to passing eddies or to large-scale topographic waves over the ridge would cause the animals to be swept far off the ridge axis, unless they are actively swimming back along a food gradient to the ridge. A separate analysis of the currents observed during the two backscatter periods is planned for a future publication.

The finding of periodic 20–40-d cycles in zooplankton backscatter in all ONC and Seabreeze records suggests that some endogenous or exogenous factors are contributing to the concentration and dispersal of organisms throughout the lower water column. These factors may include: