Range tests of acoustic transmitter–receiver combinations are advised, and often conducted as an initial step in research that is implemented to study the movements of mobile aquatic species [14, 15, 20, 38]. We conducted our range tests in order to determine the feasibility of implementing movement studies using acoustic telemetry, attempting to determine the number of receivers that might be required to gate likely movement paths of sablefish and Pacific halibut. Ultimately, our focus was to determine the effort and costs that would be necessary to achieve detection efficiencies (i.e., recapture probability) sufficient to address questions regarding movement among management areas. Our results suggest that high fish-identification rates are likely for deepwater fish species crossing acoustic gates that are placed in waters up to 600 m in depth, with receivers spaced at up to 1200-m distance from one another. However, striking temporal patterns in tag detection that were presumably related to oceanographic variability could markedly reduce detection probability on scales ranging from daily to seasonal and confound the interpretation of fish movement in some contexts. Thus, we strongly support recommendations to include sentinel tags in all active acoustic-tagging arrays (sensu [21]).
The detection range between acoustic receivers and transmitters can be influenced by a wide variety of factors. In addition to the characteristics of the transmitters (e.g., power, ping rate; [39, 40]), these can include ambient noise levels (caused by current, waves, vessel traffic, or marine organisms), and wave bending that can be influenced by factors such as depth, temperature, and salinity gradients [15, 34, 38, 39], animal swimming speed [40], and physical obstructions between receivers and transmitters [20, 41].
Considerations related to gear configuration
Among the variety of factors that can influence tag-detection rates in deepwater environments, our results provide insight regarding the distance between receivers and transmitters, depth of transmitters relative to the sea surface, the relative elevation of transmitters off bottom, the potential effects of deploying transmitters in fixed orientations at fixed locations, and the likelihood of tag signal collisions when multiple tags broadcast in close proximity to one another. With respect to receiver–transmitter distances, of immediate concern is the degree to which we were able to place nodes within each array at their target distances. While we were unable to effectively use SYNAPS to verify node locations after placement, the data from Station 3 confirmed that lateral drift of the nodes during their descent to the sea floor might have introduced substantial variance into the nominally 200-m spacings that were intended. However, inspection of the resulting reception data suggests that this is unlikely to have had a large influence overall. Specifically, were the internode distances to vary broadly around the 200-m target, we would expect to have observed numerous cases in which the reception rates decayed discontinuously over the nominal distances. That is, it should have been common to observe cases in which adjacent nodes displayed nearly identical reception rates due to their close proximity to one another, followed by large stepwise decay at adjacent node(s) due to the artificially large internode distance(s) that would subsequently have resulted. At Station 4 there is evidence of this: the 600- and 800-m nodes appear as if they were relatively close to one another based on nearly identical reception rates, while potentially having been being more distantly separated from the 400- and 1000-m nodes, respectively. In a few other cases, there were “stutters” in the reception rates observed at single transmitter nodes, but not for both transmitter elevations as would be obligatory considering their attachment to the same mooring. Thus, considered among all stations and across the distances that defined their attenuation curves, we conclude that the nature of these results were unlikely to be substantially affected by node placement errors.
The percentages of total detections from transmitters other than those that we deployed were very small, ranging between 0.01 and 0.10% among the six receivers; therefore, collisions with transmitters other than ours were rather trivial. Similarly, although the transmitters deployed at each station transmitted constantly, increasing the chances for acoustic collisions among them, the peak reception rates observed at each station were consistently higher than the overall estimated rates derived from VEMCO’s “Collision Calculator” (https://vemco.com/collision-calculator/). Whereas ten V16-5H transmitters programmed as ours and broadcasting simultaneously in close proximity to one another would be expected to suffer a 27% signal-collision rate (i.e., 73% total detection rate), we observed detection rates that peaked at 74–82% among stations. Although some proportion of this relative difference would have been due to fewer signals reaching the receivers than expected, and therefore being unable to interfere with one another, these generally high peak reception rates argue against the hypothesis that code collisions were largely influential in the observed detection patterns. Additionally, our detection-to-ping ratios were relatively high. Detection-to-ping ratios in the range of 50–60% are common (D. Webber, VEMCO Limited, personal communication), whereas the mean daily ratios in this study ranged from 73 to 79%. Positive noise quotients result from environmental noise that generates more pulses than could have been generated by tags [34]. Negative noise quotients, such as we experienced, are an indication of a high degree of signal collisions. Such high signal-collision rates should probably not be expected for free-ranging animals that only periodically come within range of receivers. As such, our tag-detection estimates are expected to be generally conservative.
Transmitters anchored in fixed positions may experience additional variations in detection rate that would be uncommon for transmitters attached to fish. With little or no current, our transmitters would have been oriented vertically in the water column, along the axis of their tethers. However, with increasing current the tethers would be pushed off vertical. In their study in a riverine system, Clements et al. [42] noted that acoustic power output was lower through the ends of transmitters relative to that which was produced around the sides of the cylinder. Thus, the orientation of the transmitter influences the direction of its greatest acoustic output. For a horizontally oriented tag implanted into a fish, maximum output would occur above and below the animal rather than off its nose and tail; i.e., the signal would be directed upwards toward the suspended receiver instead of along the seafloor. For our vertically oriented tags, their detectability was likely to vary as current conditions caused off-vertical fluctuations in their orientation, as well as changes in receiver angle. Additionally, sudden decreases in tag detection rate without later recovery (e.g., the near-bottom transmitter at the 1200-m node at Station 2, after February; see Fig. 5h) might occur if a tag breaks from its mooring, either in whole or in part.
Additionally, detection of our moored transmitters might decrease in cases where transmitters are depressed by currents to such an extent that a benthic object could occlude the path between the transmitter and receiver. Some of the consistently low detection rates that we observed (e.g., the near-bottom transmitter at the 800-m node at Station 2; see Fig. 5b) may have been attributable to a physical obstruction (e.g., kelp fronds) that partially blocked their transmission pathways. Again, such phenomena are less likely to affect tag-detection in mobile fish, as one would expect the animal to eventually swim past any obstruction and become detectable.
The pattern that was repeated at both offshore stations of decreased total detections from the 200-m transmitters relative to those at 400 m is also noteworthy. One possible reason for this could have been partial or periodic blocking of transmissions by the trawl floats used for transmitter suspension. For the transmitters positioned at 200 m, the estimated angles to the receivers were approximately 18 and 16 degrees at Stations 5 and 6, respectively; for the 400-m transmitter nodes, these angles were about 9 and 8 degrees, respectively. Although these angles seem too low to place the trawl floats in direct lines between the transmitters and receivers, under unusually strong current conditions the floats could have been displaced sufficiently to have caused such interference. Alternatively, reduced detection of the tags closest to the receivers may have represented “close proximity detection interference” [43], in which signal decoding is prevented due to temporally overlapping detection of tag signals and their echoes emanating from reflective surfaces (e.g., the benthos, pycnoclines, or sea surface).
Finally, consistently higher tag-detection rates for the deeper receiver relative to the shallow receiver at Station 2, across transmitter-node distances and transmitter elevations, may be attributable to numerous factors, including relatively shorter linear distances from tag-to-transmitter (i.e., slant distance; e.g., [44]), greater distance from surface-related acoustic noise, and greater distance from pycnoclines (see next section) that may interrupt or deflect signal transmissions. Alternatively, in habitats in which local topography might serve to block the transmission pathway between benthic transmitters and a suspended receiver, placing receivers higher in the water column could improve line of sight and increase detection probability relative to deeply deployed receivers. As such, for deep-dwelling fish, maximizing detection probabilities may involve a tradeoff between minimizing receiver–transmitter distances while maintaining effective line of sight.
Environmental considerations
Among the transient environmental factors that might affect tag-detection rates, daily averaged wind stress as a proxy for resulting sea state did not appear significant. Although others [45, 46] have concluded that wind stress can have a strong influence on acoustic tag detection, we could find no such effect on a daily scale. Differing results may be a function of the cited studies having deployed tags on mobile fauna as opposed to fixed moorings; or due the relatively deeper environments in which our tests took place. Alternatively, inconsistency in our results may have been caused by the nature of the wind data upon which we based our regressions. For most stations we were forced to use data from anemometer stations that were likely imperfect reflections of conditions at our test locations. Specifically, Bartlett Cove is a protected embayment located in the entrance to Glacier Bay, approximately 60 km north of Station 4; Port Alexander is located on the protected eastern side of Chichagof Island. As such, the wind stress estimates for all but Station 3 are likely to have been somewhat decoupled from the true conditions at the range-test stations. Furthermore, analyses based solely on daily mean wind speed will be naïve to relative persistence of those winds and the systems that generate them and, hence, the potential for sea state to build over extended periods without changes in mean stress. In contrast, the strong diurnal pattern that was apparent in the periodograms is wholly consistent with wind influence. Diurnal cycles in tag detection have been attributed to diurnal winds [46] and biological noise [15, 24]. Here, decreasing detection as days progressed and recovery of detection rates overnight suggest the building and relaxing of thermally driven diurnal winds [46, 47] in contrast to observations that biological noise tends to increase either at night [15] or crepuscularly [48].
The lack of strong evidence for tidal effects on tag detections was likely due to our deep operating depths, at which relatively low current speeds would be expected. Semmens et al. [27] conducted acoustic range tests over a monthly tide cycle in a flat-and-channel habitat that exhibited depths no greater than 40 m and found no relationship between tag-detection rates and either tidal phase (i.e., high, ebb, low, flood) or state (i.e., neap, intermediate, spring). In the current study, we did find evidence of the daily ebb-flood cycle via Fourier analysis (sensu [26]), but the signals were weak and the periods over which they occurred were transient. Although not indicated in the periodograms, closer attention to the hourly data from Station 4 suggested potentially strongest influence at the semilunar period. At this location, peak current velocities were expected to be ≥ 3 cm s−1 during spring tides and were probably capable of depressing transmitter and receiver floats, thereby changing angles of incidence associated with the transmissions. Inspection of the station’s receiver revealed abrasions on its casing suggestive of having come into regular contact with the rocky bottom. Current has been shown to be among the factors that can impart rhythmic patterns in detection frequency (sensu Fig. 9), absent of tagged-animal behavior [15, 26].
Stratification of marine waters (i.e., pycnoclines: thermoclines and/or haloclines) can also influence the detectability of acoustic signals [26, 40]. Hydrographic structure of Southeast Alaskan waters is strongly influenced by seasonal precipitation, snow-melt, glacial runoff, and thermal forcing. We hypothesize that the stepwise, seasonal shifts in tag detection observed both within Frederick Sound (Stn. 3) and on the coastal shelf (Stns. 5 and 6), as well as the erosion of diurnal periodicity seen in the Fourier transforms, were influenced by such stratification. Strong seasonal shifts in the detectability of sentinel tags have been observed elsewhere (e.g., [26]), and in our case the onset of decreased detection was consistent with the expected development of summer pycnoclines. Conversely, downwelling, strong winds, and cooling serve to enhance vertical mixing [49] and probably contributed to the improved detectability in autumn and winter (e.g., Stn. 3, Fig. 6d, e). With respect to diurnal periodicity, we hypothesize that in summer the pycnocline may have served as a reflective surface that directed wave-generated surface noise back toward the receivers, amplifying its ability to interfere with transmitter signals. Erosion of this layer in winter would allow such noise to more efficiently disperse and attenuate to depth.
The source of 8-h periodicity in transmitter detections is more difficult to explain. We are unable to identify any obvious atmospheric, oceanographic, or anthropogenic factors that were likely to be operating in these areas during the study period. In theory, pronounced biological noise occurring at dawn and dusk (sensu [48]) should be capable of this at times when day length approximates 8 h. However, in the study region this would occur from approximately November through January; the period over which no such periodicity was observed. Ultimately, additional studies in which corresponding environmental data are collected and measurements of ambient noise are conducted throughout the course of transmitter–receiver deployments may be required to identify the source of this signal.
Variation in tag detectability may also be attributable to factors such as varying, spatially complex bathymetry and shoreline configuration, passage constrictions, and the presence of sills; each of which can influence the direction and speed of wind, currents, salinity and temperature gradients, and vertical mixing. Finally, suspended sediments are known to influence acoustic transmissions [43]. Runoff from the coastal mountains of Southeast Alaska and glacial scouring introduce high particulate loads into the interior waters [49], in addition to the organically derived particulate load.
Feasibility of deepwater acoustic gating
Series of receivers may be oriented in linear arrays to form gates across pathways of known or suspected fish movements (e.g., [15, 38, 42]). Discrete estimates of detection probabilities can be used to quantify detection range, estimate appropriate spacing or density of receivers, and evaluate the utility of acoustic tagging as a method of addressing movement questions. Given the greatly varying and location-specific conditions that can influence detection probability, calculations such as those conducted herein should probably be conducted at each candidate location for acoustic gates or arrays.
Welsh et al. [20], working in a coral reef environment, concluded that the working detection range could be as low as 60 m for a VEMCO V9-1L transmitter with a power output of 146 dB. Studying movement of blackspotted croaker (Protonibea diacanthus) Semmens et al. [27] used VEMCO VR2 receivers in concert with V16-5H transmitters (the same combination that we used) in two coastal areas of Australia. The areas included a channel that was 40–50 m deep bordered on each side by shallows (5–10 m), and a shallow (< 10 m) muddy bay with rocky outcrops. They determined a maximum effective detection range of 200 m, achieving 73–91% detection. Studying cuttlefish (Sepia apama) in South Australia, Payne et al. [15] achieved a detection efficiency of about 46% at a distance of 200 m in water that was 6–10 m deep, using VEMCO V9AP-2L transmitters and VR2W receivers. Monitoring movements of Atlantic cod (Gadus morhua) in deep (65–100 m) boulder fields in the Gulf of Maine, Lindholm et al. [50] estimated a 400-m detection range for VEMCO V8SC-1H-R256 transmitters and VR2 receivers, based on 80% detection efficiency. Egli and Babcock [51] used VEMCO V16 and V8SC tags in combination VEMCO VR2 receivers on a coastal reef and determined a working range of < 500 m.
Here, we estimated individual pairwise receiver–transmitter detection proportions that were no higher than 0.84, but demonstrated that the probability of detecting a tagged fish moving through the gate could often be markedly higher, and a working detection range of 1000 m might be reasonable in many contexts. This would occur because the fish’s transmitter would be expected to transmit multiple times while within the detection band, providing multiple opportunities for tag identification. In addition, the fish-identification probabilities that we estimated assumed that the fish would travel at a constant rate, along the shortest path possible, at maximum average distance from the receivers, with all of its tag transmissions spaced at the maximum nominal delay. This suite of assumptions is quite conservative, relative to a more likely scenario in which transmissions are more frequent and the instrumented fish travels a meandering path, thereby increasing its residence time and coming closer to the receivers than would be experienced by travelling along the MIL.
Pincock [52] suggests that, under worst-case conditions, 50% detection probability can be used as the limit for reliable tag detection. Among our range-test stations, this level was generally achieved at 1000–1200 m from the receivers, potentially allowing for a high probability of fish identification using 2000–2400 m spacing in the areas where we would be likely to establish acoustic gates. This spacing might be insufficient to yield precise estimates of movement rates in applications where such estimates would directly impact quota decisions, and maintaining substantially closer spacing (e.g., 1000 m) may be financially prohibitive across large segments of continental shelf. For example, it may be infeasible to gate the borders between British Columbia and the USA, from shore to 1000-m depth, in order to estimate seasonal biomass redistribution [8] relative to fishery opening and closing dates (sensu [53]). However, using acoustic tagging to address questions that are less rate-dependent appears feasible. For example, the technology could be appropriate for estimating proportional redistribution among areas and for elucidating the timing of dispersal of various demographics within both Pacific halibut and sablefish populations. The receiver spacings that proved viable here are similar to those used by Pecl et al. [40] to study movements of calamary (Sepioteuthis australis) in inshore waters of Tasmania and may prove feasible for gating ecologically important deepwater features of the North Pacific, for example, passes of the Aleutian Island Chain in order to assess the relative isolation of local halibut spawning groups (sensu [54, 55]), or the continental shelf west of Vancouver Island to investigate relationships between sablefish dispersal and environmental forcing functions (sensu [2]). Ultimately, the utility of any technological advancement should be assessed relative to the scale of the questions that are to be addressed. Acceptable detection thresholds will be case-specific and should be evaluated in the context of acceptable levels of failure to detect instrumented animals moving through the chosen array.