Positioning aquatic animals with acoustic transmitters

Geolocating aquatic animals with acoustic tags has been ongoing for decades, relying on the detection of acoustic signals at multiple receivers with known positions to calculate a 2D or 3D position, and ultimately recreate the path of an aquatic animal from detections at fixed stations. This method of underwater geolocation is evolving with new software and hardware options available to help investigators design studies and calculate positions using solvers based predominantly on time‐difference‐of‐arrival and time‐of‐arrival. We provide an overview of the considerations necessary to implement positioning in aquatic acoustic telemetry studies, including how to design arrays of receivers, test performance, synchronize receiver clocks and calculate positions from the detection data. We additionally present some common positioning algorithms, including both the free open‐source solvers and the ‘black‐box’ methods provided by some manufacturers for calculating positions. This paper is the first to provide a comprehensive overview of methods and considerations for designing and implementing better positioning studies that will support users, and encourage further knowledge advances in aquatic systems.


| INTRODUC TI ON
The study of aquatic animal movement ecology has emerged as a major research field with implications for understanding life on the planet, and how it can most effectively be managed and conserved against human interference (Nathan et al., 2008).However, the study of movement depends on efficient tools for animal observation and resolving where, when and how animals are moving.Importantly, movement is a fractal process (Turchin, 1996); therefore, the scale (both space and time) at which movement is observed will directly influence the outcome of the observations.High-frequency positions yield greater power to detect diverse behaviours, and continuous time series with fixed interval positions are important to yield consistent and comparable estimates of movement (Brown et al., 2012;Nathan et al., 2022).The most common way to position an object on land is with global navigation satellite systems (GNSS), which connect to orbiting satellites.Animal tags designed for movement ecology in terrestrial environments will link to these satellites and log positions at fixed intervals, providing a series of positions that can be leveraged to understand individual behaviours and ecologies (Kays et al., 2015).However, radio signals attenuate quickly in water (especially saltwater), and therefore limit the capacity with which scientists can obtain the locations of aquatic animals with GNSS, unless signals can be transmitted from the surface (e.g. by buoys).The efficient transmission of sound through water, coupled with the use of sensitive hydrophones to monitor and record sound has, however, provided new frontiers for underwater communications (Taraldsen et al., 2011).In aquatic environments, acoustic telemetry transmits the identities of tagged animals to logging stations (i.e.receivers or hydrophones) as well as additional data about the animal's behaviour and physiology or the surrounding environment depending on sensor integration in the tag (Hellström et al., 2022;Hussey et al., 2015).
Acoustic telemetry does not inherently provide accurate estimates of animal positions; detection of a tagged individual by a passive data-logging hydrophone confirms its presence within a detection polyhedron (i.e.dynamic, but typically within 200-1000 m; Kessel et al., 2014).Gridded receiver deployments have been used to calculate centres of activity based on the number of detections within a given time slot as pseudo-positions (Simpfendorfer et al., 2002;Winton et al., 2018).However, for precise 2D positioning, one signal must be detected by a minimum of two or three receivers (depending on the positioning methodology), and the clock drift inherent among independent stations must be accounted for to yield more precise calculation of positions based on time-differenceof-arrival (TDOA: Smith, 2013) and time-of-arrival (TOA; Baktoft et al., 2017;Nathan et al., 2022; Figure 1).In addition, for 3D positioning, receivers either have to be distributed across the tridimensional space for TDOA/TOA-based positioning, or tags need a depth sensor.
Since the first research on acoustic positioning (Kuroki et al., 1971), there have been major advances in the field and remarkable applications of the technology to achieve a better understanding of animal ecologies (Krause et al., 2013;Nathan et al., 2022;Orrell & Hussey, 2022; Box 1).Positioning data have proven valuable for testing fundamental ecological hypotheses about movement ecology and the underlying variables determining the extent and periodicity of movement, as well as developing applied strategies for protecting species and critical habitats, managing invasive species and developing fisheries management tools (Box 1).However, there is currently no overview of positioning in acoustic telemetry, including the development and availability of positioning methods and how the methods are being used.Furthermore, users lack a comprehensive guide to better understand how to conceive, design, implement and improve positioning studies using open, accessible and interoperable infrastructure and digital analytical tools.There is therefore a need to identify the present state-of-the-art in aquatic positioning and forecast the needs of the community to enhance the applicability, and grow the use of these valuable methods.In this paper, we consider the past, present and future of aquatic acoustic positioning for ecology.The paper is intended to provide an overview of the methods and best practices, standardize terminology (Table 1) and identify research avenues perceived by the tracking community that will continue to help advance the field by making it more accessible, user-friendly and open to new developments and innovations.

| PA S T-HIS TORY OF P OS ITI ONING TECHNOLOGY
Positioning of telemetry transmitters began with triangulating positions by obtaining directional position fixes using yagi radio antennas (Heezen & Tester, 1967) or directional hydrophones (McCleave, 1978).Arrays of fixed telemetry receivers can also be used to position fish, predominantly using acoustic telemetry.The including both the free open-source solvers and the 'black-box' methods provided by some manufacturers for calculating positions.
4. This paper is the first to provide a comprehensive overview of methods and considerations for designing and implementing better positioning studies that will support users, and encourage further knowledge advances in aquatic systems.
geolocation, multilateration, positioning solver, reverse-GPS, synchronization, telemetry earliest paper we are aware of that used fixed acoustic telemetry stations to generate fish position solutions, presents tracking of crescent sweetlips (Plectorhinchus cinctus) and rainbow trout (Oncorhynchus mykiss) in both fresh-and saltwater (Lake Okutama and Sagami Bay, Japan; Kuroki Kuroki et al., 1971).Interestingly, in addition to the horizontal position estimation, the transponders (transmitters allowing two-way communication) relayed information about fish swimming depth and water temperature.A triplet of receivers moved by boat was used to detect signals from the transponders and an elaborate apparatus allowed real-time estimation and plotting (on paper) of the horizontal track and swimming depth of the tagged fish.Almost concurrently, Young et al. (1972) were soonafter tracking brown trout (Salmo trutta) in a loch in Scotland using directional acoustic receivers that could be controlled automatically.Fish were tracked and positioned for as long as 24 h.The system was somewhat cumbersome and yielded only modest accuracy.This system was not widely adopted or commercialized but, nonetheless, is a fascinating example of early efforts to generate fine-scale positions for fish using autonomous acoustic apparatus reminiscent of what is conducted today.

F I G U R E 1
Illustration of the process of positioning acoustic transmitters within an array of receivers.Receivers must be deployed at fixed locations measured with high precision (i.e.differential global positioning system [GPS] from above water).Deployment of synchronization tags that are detected across the array is necessary for the user to later adjust the clocks of each independent hydrophone, unless they are cabled and set to a common clock.Data downloaded from the receivers are then fit to a positioning solver to calculate the position.After, users may interpolate missing positions using random walk or state-space models to infer missing positions and generate paths.
Receivers are deployed at known location, with sync tags ensuring clocks can be synchronized 1.

Unique pings of sound sent from acoustic transmitters and decoded by receivers
The next major innovation occurred in the mid-1970s, when Hawkins et al. (1974) deployed omni-directional acoustic telemetry receivers to position Atlantic cod (Gadus morhua) swimming freely in a Scottish sea loch.The researchers used differences in TOA of acoustic pulses on three to five receivers to estimate fish positions.
In this study, the hydrophones were wired together and linked B O X 1 Examples of positioning studies that highlight the various uses of acoustic positioning.Note that the positioning method shown is the method reported in each respective reference.but was also somewhat cumbersome and required that stations had access to the water surface to enable radio transmission of signals.
To overcome clock drift on receivers, cabled arrays were the norm in early days (e.g.Juell & Westerberg, 1993) and it was not until Lotek Wireless (Newmarket, Ontario, Canada) developed a cabled acoustic telemetry system using code-division multiple access (CDMA) technology that it was possible to conduct high resolution tracking (with <1 m accuracy) of many individuals in a small area (glossary in Table 1; Baktoft et al., 2015;Cooke et al., 2005;Cote et al., 1998).
CDMA emerged from the cell phone industry and was adapted by Lotek for fish tracking.The Lotek CDMA technology minimized code collisions, allowing signals from multiple transmitters to be detected at the same receiver in a short duration, and enabled studies of both marine (Cote et al., 1998) and freshwater (Baktoft et al., 2012;Hanson et al., 2007;Nakayama et al., 2018) fishes.This technology is still commercially available using non-cabled battery powered units.
Similar technology (also cabled) was developed by Hydroacoustic Technology Inc. (HTI; now part of InnovaSea) and was applied to the tracking of smolts (Salmonidae) in rivers (Steig, 1999) and Atlantic cod in aquaculture net pens (Rillahan et al., 2009).Nielsen et al. (2012) created a towed array system that could be used to position acoustically tagged fish in 2D providing opportunities for tracking fish without fixed infrastructure.There may be future opportunities using remote uncrewed vehicles as well (Masmitja et al., 2020).
It is easy to forget that the fine-scale telemetry studies of today and tomorrow are not all that different from what was accomplished by the pioneering studies in the 1970s.What differs is that the tools for doing so are now commercially available, smaller, more versatile and cheaper; thus, more widely embraced.There is now potential for a golden age of fish positioning to generate big data sets (Nathan et al., 2022), but it is key to acknowledge and appreciate the historical work that has led us to this state.

| PRE S ENT-B E S T PR AC TICE S IN ACOUS TI C P OS ITI ONING
3.1 | Tag and receiver specification

| Tags and ID code systems
To position tags, and by extension the animals carrying tags, depend on the physics of underwater sound.The acoustic transmitter used in these studies consists of a sound transmitting element (transducer), electronics to control the signal emission, a battery providing power, sensors (optional) and a protective housing.The transmission interval (s) and power output (dB re 1 uPa at 1 m) are determined by the programming and electromagnetic properties but limited by the battery.Lower frequency (e.g.69 kHz) demands a larger transducer and more battery power than higher frequency alternatives (i.e.>=180 kHz).Although higher frequency tags are comparatively smaller than lower frequency tags, the detection range is shorter at higher frequencies and performance may differ more by temperature and conductivity of the water (Pincock & Johnston, 2012).
There are two different main code categories widely in use, based on how the code is constructed.These are Pulse Position Modulation (PPM) codes and the aforementioned CDMA codes (Table 2).The PPM codes consist of unmodulated fixed-frequency pulses emitted in a 'pulse train', with the code information being defined by the composition of the time intervals between each pulse in the pulse train across several seconds of transmission (Ehrenberg & Steig, 2003).With CDMA, however, the full ID may be encoded within a single, short pulse by modulation of the pulse (Ehrenberg & Steig, 2003).There are two main categories used for such modulation in acoustic telemetry; binary frequency shift keying (BFSK), and binary phase shift keying (BPSK).The unique code ID is binary (0 and 1), and the shift between 0 and 1 is made either by a small frequency shift (BFSK) or by a phase shift (BPSK) in the signal (e.g.McMichael et al., 2010).Common to these categories is that the pulses consist of an initializing part (aka Barker), a part that constitutes the transmitter ID, and a trailing part for signal verification/error detection.For a sensor tag, a part containing the sensor value may either be included in the pulse coding for the ID, or sent as a separate pulse immediately after.The number of pulses depends on the desired number of unique IDs, and if the signal should include sensor data transmitted along with the ID but the full signal code generally consists of one to three pulses emitted within a very short time (<1 ms to a few ms).Sensor tags therefore generally have longer codes and hence shorter battery life than non-sensor tags for a given burst interval, though this can to some extent be mitigated by reducing the number of unique IDs.There have been quite a few different code spaces developed for PPM codes, but a code typically consists of 7-12 pulses emitted within 3-5 s (Reubens et al., 2021).
The long code duration and the fixed frequency make PPM-based systems vulnerable to code collisions, and also to multipath (the signal code is corrupted due to reflection by surface, bottom, thermocline or some object of the signal from the same transmitter).A different coding category from PPM and CDMA has been used by HTI, where the code ID is based on the exact interval between consecutive pulses.Each ID has a small difference in burst interval (Ehrenberg & Steig, 2003, 2009).Theoretically, this code system may work better than BPSK with a lower signal-to-noise ratio, however the system requires large efforts for code deciphering when multiple tags are present, and in particular where multipath signals are an issue.Identifying tags from code sequences can be facilitated using artificial intelligence (Medisetty et al., 2021).This coding system cannot be used for sensor data requiring many sensor levels.
TA B L E 2 Overview of available acoustic transmitters for each of the brands.frequencies and/or code sets, enabling a higher number of tags in the system, and even a combination of different signal systems.
There are two main categories of acoustic receivers: cabled receivers and autonomous receivers.Cabled receivers, where all hydrophones are coupled to the same receiver unit, enable all detections to be recorded and stored based on the same clock that is synchronized using the physical connection among receivers.Cabled systems are limited by cable length, and will typically not cover more than a few 100 m.Autonomous independent receivers, which operate without any cabling, have an internal quartz clock, and their measurement of time is affected by precision of the quartz oscillator and temperature-dependent time drift.This time drift must be corrected for by clock synchronization during post-processing, before positioning algorithms can be applied (see the Section 3.2).This may be assisted by using linear interpolation on each receiver prior to running the synchronization model, if the time drift is recorded.
Synchronization signals from fixed beacon tags are necessary to aid in such time synchronization.Autonomous receivers containing built-in beacon tags are becoming more common to support 2D and 3D telemetry.Onboard temperature sensors are increasingly common on receivers, and may be important for positioning as the speed of sound changes with temperature (Simpfendorfer et al., 2008).
Other sensors that are available for some receiver models are ambient noise, pressure and tilt, factors that may affect detection range (Kessel et al., 2014).Some receiver models are offered with acoustic release mechanisms, increasing efficiency of retrieval, removing the need for a surface buoy or grapple lines, and allowing deployment in deep waters or challenging environments (e.g.entanglement risk to marine mammals).Other features also include online retrieval of data, allowing near-live updates of positioning data by the use of a surface data modem (Baktoft et al., 2017;Manicacci et al., 2022).
Such features may be relevant for event driven manipulation studies, active management of fishways or other installations where seasonal choices of activity can be made.

| Synchronization
Modern acoustic receivers are typically autonomous batteryoperated units that allow for modular and flexible receiver array configurations and deployments.Receivers are usually independent but may be cabled to an on-shore station for data collection and clock synchronization.For independent units, each autonomous receiver contains a quartz crystal-based internal clock.Because position estimation is based on extremely small differences in time of detection of an acoustic signal at multiple locations, these internal clocks need to be synchronized so that timestamps of detections are accurate to the milli-second or better to achieve sub-metre spatial precision (Figure 1).When receiver units are modular and operate independently of each other, this poses a challenge because each clock drifts as much as 1 s per day, depending on temperature experience and unique characteristics of each clock crystal.
Synchronization of independent receivers is typically based on internal or external beacon transmitters/sync tags colocated with all/several receivers (Baktoft et al., 2017;Smith, 2013; Table 1).In the simplest case where the exact position of receivers is known, receivers remain stationary throughout the study, and all receivers continuously detect the same beacon transmitter, correcting the clock drift is relatively uncomplicated.In this case, betweenreceiver distances are fixed and equivalent to signal travel time between pairs of beacon transmitters and receivers when accounting for the effect of water temperature on speed of sound.Field studies using acoustic telemetry often entail more complexity as receiver location might be uncertain (e.g. if deployed near the bottom in deep water) and receivers might move during the study period (e.g.waveinduced drift or being moved inadvertently by nets or anchors).Such complexity can be accounted for during the synchronization, if the model used allows for estimation of receiver positions.Additionally, areas of interest and thus receiver arrays, can be relatively large and of complex geography, and it is often impossible for all receivers to detect the same beacon transmitter.In such cases, synchronization can be done sequentially (which entails propagation of uncertainty), or using a more complex model allowing synchronization of the entire array at once.

| Positioning models
Positions in acoustic arrays are estimated from multilateration of detections on multiple receivers at known locations and with synchronized clocks.Regardless of the underlying positioning model and assumptions (see Table 3), position estimation is ultimately based on the principle that the distance between transmitters and receivers is directly proportional to the time it takes for the signal produced by the transmitter to travel to each receiver, which are at known locations.The precise time of a signal transmission is generally unknown, and position solvers have frequently been built on pairwise differences in TOA to estimate transmitter position for when the signal was sent.This TDOA method relies on solving sets of hyperbolic equations to determine the position (Juell & Westerberg, 1993).However, this method requires that the signal is detected on at least three receivers.Whereas current TDOA methods often rely on estimation of independent locations, it is possible to apply a more holistic approach and estimate coherent tracks, thereby utilizing all available information to inform the estimation model.Furthermore, a track-based approach allows estimation of the time of signal transmission and thereby allows position estimation based on TOA, which is more robust to suboptimal receiver array configuration and tagged animals being outside the receiver array footprint.Additionally, basing the track estimation on TOA and using all collected data, makes it possible to estimate tracks of tagged animals in cases where the number of receivers detecting each signal is less than three (

| Tag and attachment choice
In cases where the species is highly resident, or a specific aspect of the movement of the individual is of interest (e.g.nesting black basses Micropterus spp., coastal wrasse species Labrus spp.), PPMbased systems may be a poor choice.The long code trains and code collision risk associated with PPM-based systems will limit minimum burst interval to higher values than desired, so CDMA-based systems may be preferred or needed.A PPM-based system will be strongly limited in the number of tags that can be in the study area simultaneously due to the potential for code collisions.It is possible to simulate the yield of the whole system in different configurations (burst interval vs. number of tagged animals vs. detection range).The risk of code collisions can be reduced by increasing the mean burst interval, and reducing signal power, but doing so will also decrease the positioning rate, and thereby reduce the tracking performance.PPM-based transmitters that also have sensors will have longer code trains that are more vulnerable to collisions and are overall more challenging to consistently calculate positions for.Although PPM systems may have a larger detection radius than CDMA, the latter will have higher positioning rates while within detection range due to the shorter burst intervals (Leander et al., 2020).
Acoustic tags without sensors may be sufficient in some situations, but in other cases, a sensor may be important for interpreting the animal behaviour in relation to the research objectives.For example, it may be very difficult to discern if the transmitter signals come from the animal that was tagged, or a predator that has ingested the tagged animal.In such cases, predation event data can be very useful, using temperature or predation sensors (Hanssen et al., 2022;Klinard & Matley, 2020;Lennox et al., 2023).If the prey and the predator exhibit different behaviours, information from a pressure or acceleration sensor may allow the identification of predation events (e.g.Halttunen et al., 2018).Temperature sensors could also detect predation by mammals who have higher internal temperatures than fish.A pressure sensor, accelerometer or a mortality sensor (a sensor that registers if movement has stopped or if the carrier has lost the ability to maintain orientation for more than a defined period) can also aid in determining if the animal is not moving at all (presumably dead), or if it is moving just very little (alive).
Moreover, a decision key should be used to probabilistically assess the fate of tagged individuals based on objective, repeatable criteria (e.g.Halttunen et al., 2018;Lennox et al., 2023).The timing of the fate event (e.g.predation) is important to assess, to separate tagged animal behaviour from other data in the subsequent analyses.
Methods for tagging have been extensively covered elsewhere and do not bear repeating here (Brown et al., 2011;Jepsen et al., 2015), except for specific notes relevant to positioning.
External tags are likely to have a larger detection radius (Dance et al., 2016), particularly for animals with thick abdominal walls that will more significantly attenuate the signal, but internal tagging should be considered if external transmitter attachment poses extra drag to the animal, increases the visibility of the tagged animal to predators, increases the potential for entanglement, or moves too freely around the animal and comlicates positioning.The internal transmitter may have negligible effects on growth if the tag: animal size ratio is not too large (e.g. but see Hühn et al., 2014), but what this critical ratio will be is likely to depend on species and its size.
The current smallest available acoustic transmitter is the ELAT (Eel-Lamprey Acoustic Tag), made for the JSAT system measuring only 2 × 12 mm, and may be used for fish as small as 70-80 mm (Mueller et al., 2019).However, smaller tags are liable to have lower power and will be harder to design arrays upon which a single transmission will be detected at multiple points.

| Mooring design, biofouling and retrieval
Currents, unstable substrate and human interference can all shift a mooring's location during the study period (Goossens et al., 2020).
Many of these problems can be mitigated by simply adding more weight and using denser materials.Depending on water temperature and salinity, and the density of the mix, concrete typically weighs 40% less under water than in air.Steel-reinforced moorings may therefore be more appropriate.For large moorings (>1 m 3 ), hydrodynamic shapes like pyramids may mitigate the effect of currents on drift.If currents occur in a prevalent direction (as in a river or dam tailrace), use of an upstream anchor may provide enhanced stability.
On bedrock or other smooth, hard substrates, a wider mooring base may provide enough friction to mitigate sliding along the bottom.On sandy or soft bottoms, auger anchors screwed into the substrate are one of the most efficient systems to maintain acoustic receivers at fixed locations.
Biofouling can also significantly affect positioning system performance.Biofouling is the growth of organisms on the hydrophone structure, such as algae and macroinvertebrates that settle and can occlude sound signals from being detected.Different treatments can be used to prevent or mitigate it, but few studies have tested their performance (see Heupel et al., 2008).
Preventative measures (e.g.anti-fouling paint) can vary in their efficacy, and regular cleaning or equipment replacement may not guarantee an increase in performance (Mathies et al., 2014).More work is required to test the effects of biofouling and the efficacy of various treatments.
How receivers are deployed and retrieved can affect positioning performance simply in that difficult retrieval may increase receiver loss and subsequently array coverage.Upon deployment, investigators should also strive to minimize the lag time between deployment of the receivers to minimize clock drift associated with differential temperature experience.Depending on the conditions in the study area (i.e.current, depth, noise, etc.) and the array geometry and mooring weight, several solutions may be practical.Certified professional SCUBA divers can safely and consistently recover receivers at shallow depths.Acoustic releases (which are available as integrated within receiver hull, or as a separate unit) allow for remote retrieval when diving or grappling is impossible or too dangerous, but re-deployment may be challenging without a rope canister to maintain a connection with the mooring (Goossens et al., 2020).In some cases, a new mooring may need to be deployed each time the receivers are downloaded.
Furthermore, acoustic releases may malfunction in noisy waters, making retrieval impossible and breaking an acoustic array; investigators may need to design their arrays with redundancies as a contingency against loss so that one missing unit does not compromise an entire study area, with considerable cost ramifications.
Finally, remote data offloading is now possible with wireless technology, which can mitigate retrieval problems, though receiver battery life still limits the duration of a single receiver deployment event.
The way receivers are attached to the mooring can affect positioning accuracy.For instance, one of the most popular methods of receiver deployment is to attach the receiver to a rope that is fixed to a mooring weight on the bottom and a surface or subsurface buoy (Figure 1).Depending on the rope length and current velocities, the receiver, in these situations, can oscillate around the mooring point up to several metres.Movement of receivers deviates the receiver's listening position from the original coordinates, affecting the grid geometry and the ability of positioning algorithms to estimate an accurate position due to varying distances between receivers detecting an acoustic signal.

| Array geometry
The distance between gridded receivers has to be selected according to the detection range at that location, which in turn, will depend on the power output, sound attenuation and working frequency of the transmitters.The habitat characteristics, such as bottom topography, determine the possibility for direct signal travel between transmitter and receiver.More dynamic factors such as biological (Payne et al., 2010) and industrial noise (Ingraham et al., 2014;Simpfendorfer et al., 2008), macrophytes, biofouling, temperature gradients, wind, wave action and tide will cause detection range to vary over time (Gjelland & Hedger, 2013;Huveneers et al., 2016;Winter et al., 2021).Tags can even be overpowered in the vicinity of reflective surfaces (hard substrates), in which case reflections or echoes (multipath) self-destructively collide, reducing detection range at short distances but not at long distances (Kessel et al., 2015).
Independent of other factors, a positioning array's geometry greatly determines the positioning precision (e.g.Kraus et al., 2018;Welsh et al., 2012), especially when applying hyperbolic multilateration positioning algorithms.The decrease in the accuracy of the positioning can be estimated by calculating the dilution of precision (DOP, including the HPE), which depends on the relative location of the transmitter and the receivers that detected the signal (Niezgoda et al., 2002).percentage of signals detected by at least three receivers) and accuracy (estimating DOP), to select the optimal design that adapts to the specific experimental conditions (Aspillaga, Arlinghaus, Martorell-Barceló, Follana-Berná, et al., 2021;Kraus et al., 2018).
Developing open-source tools and packages to facilitate these tests for researchers will be helpful to support decision-making.Lastly, knowledge on the bathymetry of the study site and the bottom substrate can be crucial to prevent violation of the direct line of sight between receivers.In the case where tags are not in direct line of sight (i.e.there is an underwater hill or corner), but transmissions are detected by the hydrophones, the calculated position will deviate from the true position.
Another key aspect that may affect the estimation of animal positions is the error associated with the positions recorded for the different acoustic stations within the array.In this regard, there are at least two important aspects to consider: the ability to record precise and accurate receiver positions, and the potential movement of the acoustic receivers during the study period.The position of the acoustic receivers is typically recorded at the start of the study by means of a handheld GNSS, and should ideally be supplemented with an additional position at the end of the study for comparison.However, conventional GNSS has a positioning error of ±2-3 m in good weather conditions, but shows much higher error distances on rainy days (Yeh et al., 2009).A better alternative is the use of differential GPS (dGPS) that records positions with a <1 m error.

| Performance of the experimental setup
Once the receivers' geometry has been established, the next important step is to evaluate its positioning performance in terms of accuracy, precision and expected frequency of positioning (Baktoft et al., 2015).It is strongly recommended to perform detection range assessments in the full range of environmental conditions the experimental site could face (e.g.strong dominant winds, winter/summer, the presence of macrophytes in some periods that hinder the acoustic signal propagation; Thiemer et al., 2022), before and after the deployment of the final array of receivers to ensure proper positioning, regardless of the positioning system used.Given that whole-track TOA-based methods can calculate positions without detections on three receivers (Baktoft et al., 2017), they may be more robust than single-point TDOA-based methods to fluctuating asymmetry in detection distances.
Over the life of the experiment, several fixed tags whose positions (and measurement errors) are known should be deployed in the study area, synchronizing tags should be tied to (at least some) receivers, and reference tags, similar to those used for tagging animals, should be spread at different known locations in the experimental setup, separately from receivers to ground truth position estimates on an ongoing basis.Both these tags, whose positions can be estimated by the positioning algorithm as well, inform on how the setup performs, can detect potential anomalies (Binder et al., 2016;Huveneers et al., 2016;Winter et al., 2021), and inform development of filter criteria (Meckley et al., 2014).
In addition to these fixed sentinels, moving tests should be performed, ideally simulating similar movement patterns to those expected from the study species.Best practice is to tow a transmitter whose position is continuously recorded by dGPS through the array, and then to compare the dGPS positions with those estimated by the positioning algorithm (e.g.Aspillaga, Arlinghaus, Martorell-Barceló, Follana-Berná, et al., 2021;Baktoft et al., 2015;Baktoft et al., 2017;Leander et al., 2020;Roy et al., 2014).Ideally, the test should be performed at different depths and across different habitats, representing what is expected from the targeted species (e.g.benthic/ pelagic).Importantly, and especially for species known to prefer nearshore habitats, it is recommended to perform tests inside and outside the array, as performances can vary between both (e.g.Roy et al., 2014).The speed and turning angles of the towed tags should also be in agreement with the movement characteristics of the targeted species.

| FUTURE-NE W HORIZONS AND OPP ORTUNITIE S FOR AQUATIC P OS ITIONING SYS TEMS
Positioning acoustic tags was one of the most significant developments in aquatic movement ecology, because it allows the generation of high-throughput underwater tracking data (Nathan et al., 2022).Overcoming the challenges of the water-air boundary to acquire knowledge of where and when animals are present in aquatic ecosystems has allowed major advances in the field (Box 1).
However, an obvious limitation has been the reliance on stationary receiver array grids, which need to be relatively closely spaced and Investment in community-centralized computing resources may be a solution to reduce the burden on research groups using positioning.
This challenge could be addressed by changing the paradigm from a traditional model-driven approximation towards a more data-driven oriented approach to transform the large amount of positions into real knowledge.The integration of artificial intelligence, and deep learning in particular, in aquatic high-throughput movement is a good candidate to bridge this gap, but is still in its infancy (Maekawa et al., 2020).In addition to computing power, improved connectivity between receivers and computers that run positioning algorithms designing experiments and interpreting their data including glatos (Holbrook et al., 2021) for study design and data management and YAPS (Baktoft et al., 2017) for receiver synchronization and position calculation.Supporting new open tools and advances in the field will be key to increased uptake of positioning and higher relevance of this tool for aquatic ecologists.

| CON CLUS IONS
The aquatic realm is mysterious and studying the movement of animals underwater at biologically relevant resolution and scale has been a challenge for centuries.Acoustic telemetry positioning systems increasingly enable us to do just that.The basic concepts behind positioning animals underwater using acoustic telemetry have remained fundamentally the same since the early experiments of Kuroki et al. (1971), Young et al. (1972) and Hawkins et al. (1974).
ReceiverA unit with an integrated hydrophone and electronics to process and store acoustic signals from transmitters.Typically distinguished by its serial number Reference tag, sentinel tag A stationary transmitter deployed within an array for verification or evaluation of positioning error (i.e. to compare estimated and true positions) Signal (coded or non-coded) One or more pulses intended to be interpreted as one detection by the receiver Station Location where a receiver is deployed Sync tag, beacon tag A stationary transmitter deployed within an array to aid the receiver time synchronization during post-processing Time-of-arrival TOA Signal arrival time at the receiver Time-difference-of-arrival TDOA Difference in signal arrival times between pairs of receivers & Hussey, 2022; Table Overall, signals detected from contrasting angles (e.g. a signal detected by three or four surrounding receivers) produce better position estimates than signals detected from similar angles (e.g.receivers placed in a row at 180 degrees).The acoustic range and the DOP can be incorporated into computer simulations to compare the efficiency of different grid designs.In these simulations, the positioning efficiency of different configurations can be tested by simulating trajectories based on a movement model (e.g.random walks) and then estimating the positioning efficiency (i.e.
thus occupy a finite space limited by detection range and number of receivers.Near real-time positioning faces additional limitations of data transmission rates between mobile receivers and a central computer.As ambitions for studying animals scale up to major lakes or marine areas, such innovations will greatly expand the applicability of positioning beyond what has been possible to date.Computing power needed to calculate positions is a clear limitation to scaling up positioning studies.Running synchronization models and positioning algorithms on these large datasets is time consuming, computationally expensive, and may be inaccessible to smaller research groups without access to servers or centralized computing.How to store, analyse, use and transform these massive amounts of movement data into useful knowledge is already a challenge for many acoustic positioning studies.Positioning is unlikely to become simpler in the future and open-source solutions are complex.
could improve the quality of positional telemetry data by allowing feedback in near real time.In most contemporary positioning studies, positions are only derived after the conclusion of a study, preventing opportunities to identify and address performance issues as they arise.Synchronizing clocks on independent receivers is one of the most challenging steps to overcome in acoustic positioning.Cabled receivers eliminate this challenge but limit coverage area and impose significant logistical constraints, so independent receivers are more common.Use of more accurate timepieces (e.g.alternatives to quartz crystals) or other methods to synchronize clocks in near real time (e.g. a periodic correction from a GPS clock) would represent a significant advancement.Accurate positioning of large numbers of individuals offersthe opportunity to create genuine wild laboratories to measure responses of aquatic animals to abiotic factors.The combination of acoustic positioning with continuous collection of fine-scale hydrodynamic and meteorological factors can provide new avenues for testing hypotheses about external drivers of movement.Such information is highly valuable for managers, for instance in their ambitions to better guide migratory fish around barriers and reduce entrainment in hydropower facilities through improved passage design.Understanding how individuals respond to environmental variables can also help predict the effect of future climate scenarios on fish populations, especially using smaller systems such as whole-lake studies.Furthermore, knowledge on the relations between hydrodynamics and meteorological factors (among others) could also improve operations of infrastructure through predictions of fish behaviour based on continuously monitored environmental factors.Hence, humans can better adapt operations of infrastructure (e.g.water level management in reservoirs through hydropower operations) to fish movement, migration and habitat use(Koster et al., 2018;Westrelin et al., 2018).Positioning algorithms are common throughout modern technology, from use in cell networks to wifi networks to GNSS-based tracking to pinpoint positions in space, and provide a coordinate that can be mapped for a large variety of use cases.The field of underwater movement ecology must closely follow and borrow from new advancements in positioning solutions from other related fields to fully capitalize on the potential of acoustic telemetry.For example, the use of CDMA technology in radio signal communications such as mobile phones and WLAN have essentially eliminated the problem of signal collisions, while also reducing the battery power required for transmissions due to the shorter transmission time required.In addition to borrowing new signal encoding methods, opportunities exist for developing improved signal detection methods within receivers, with respect to detecting quiet signals at further ranges, picking out transmission from noisy environments and sorting true detections from false multipath generated signals.Acoustic positioning has matured to be a widely used method of underwater geolocation, but we believe the present state-of-the-art is now at a plateau, short of its potential peak of performance.The next decade will hopefully bring new modes of synchronizing receiver clocks, better methods of measuring receiver positions, improvements to open-source positioning algorithms, and better ways of archiving and sharing positioning data within the community for meta-analyses and comparative research.Many of these developments will rely on engineering and cross-fertilization of the technology with other fields of communications.There is a clear need for more bottom-up development in positioning.Open-source tools have been important to ecology and development of open hardware and software can help advance the field towards open science and reproducibility.Tools that support different parts of designing positioning studies, such as: simulations that help investigators choose the right number of tags or receivers, tools that help synchronize receivers and tools that help calculate positions are key to increase accessibility of positioning to new users.Moreover, open and accessible tools maintain a high level of reproducibility, which has repeatedly been emphasized as important to ecology (Powers & Hampton, 2019).Development of open tools for positioning requires investment and interdisciplinary work with data scientists and statisticians to optimize code and expand on the existing tools.The acoustic telemetry field is already benefiting from such tools, which have been developed for assisting users with However, the collective experience, including successes and failures, of the scientific community involved in positioning aquatic animals now leaves us with the knowledge necessary to routinely plan and execute successful fine-scale tracking projects across a range of aquatic environments (from inland rivers and lakes to coastal marine waters) according to best practices, generating rich datasets of animal movement.Advances in computing power, battery life, clock synchronization techniques, positioning solvers, signal encoding methods (CDMA-variants) and array design knowledge have been key contributors towards reaching the high data yields currently achievable.Fish can now be positioned with remarkable accuracy and precision, provided that care is taken in the development of tracking systems and the interpretation of findings.Future improvements towards signal detection across longer ranges, accuracy measuring receiver locations, and incorporation of environmental variability (e.g.flow, currents) into positioning methods will enable large gains in spatial accuracy and consequently our understanding of the behaviour of underwater organisms, with corresponding improvements in the ability to draw inferences about the movement ecology of aquatic animals in high resolution and in relation to their environment.
Glossary of key terms related to acoustic positioning.
PPM ID code is defined by the composition of the time intervals between each pulse in pulse train of a fixed-frequency signal To avoid repetitive code collisions, PPM-based tags are pro- (Cooke et al., 2005;Ehrenberg & Steig, 200321)al (Niezgoda et al., 2002)ansmissions) within a lower and upper limit (e.g.40 and 80 s, with mean 60 s).Thelma Biotel (Trondheim.Norway) has developed a PPM-based receiver system that listens to more than one frequency, thus allowing tags at different frequencies (e.g.67, 69, 71 and 73 kHz) to be detected, thereby reducing the potential for code collisions and thus increasing the number of tagged fish it is possible to track simultaneously.Of the two types of CDMA coding, the BFSK coding has been used in Lotek systems at 76 and 200 kHz (e.g.Říha et al., 2022;Szabo-Meszaros et al., 2021)and BPSK codes are used by the Juvenile Salmon Acoustic Telemetry System at 416.7 kHz (JSATS, a non-proprietary system developed by the US Army Corps of Engineers, seeMcMichael et al., 2010)and by the Innovasea HR2 system at 180 kHz (e.g.Leander et al., 2020).The coded pulse and short signal duration of CDMA technology enables many tags with short burst intervals to be monitored simultaneously without code collisions being a problem(Cooke et al., 2005;Ehrenberg & Steig, 2003, 2009;Niezgoda et al., 2002).Whereas PPM code detection efficiency is often reduced by code collisions and/or noisy environments, false detections (i.e.registering a signal code that was not transmitted) are infrequent for newer code sets.On the other hand, CDMA systems may suffer more from frequent false detections.Removal of false detections can partly be done automatically during logging, but it should always be considered to do this during post-processing to avoid missing true positives.
(Ehrenberg & Steig, 2009;Leander et al., 2020)e of the tags to record detections and calculate positions.Over the years, several frequencies in the range from 30 to 416.7 kHz were used, with PPM-based systems working at 69 kHz being typical for larger regional and global networks(Reubens et al., 2021).Frequencies near 70 kHz have been chosen over others because of a good detection range in saltwater environments, and relatively small tags that can be used for many species capable of travelling far.It should be noted though, that this is inadequate for long-term large-scale monitoring of smaller specimens capable of travelling far, such as salmon smolts.Higher frequencies may give better time resolution and precision than lower frequencies, and also more precise positioning, but come with the cost of shorter ranges and potentially shorter battery life(Ehrenberg & Steig, 2009;Leander et al., 2020).Most high-precision positioning studies work on scales from <1 to several square kilometres, where higher frequency systems may be feasi- a Codesets in italic have compatibility across brands.It should be noted that some of the codesets are no longer available at specific manufacturers.bRequiresmanufacturer to process the data.3.1.2|ReceiversThe ble.Receiver configurations are increasingly important to specify for conducting a successful positioning study, as marketed options are increasing.Whereas most receivers listen to a single frequency, others may incorporate additional features such as listening to multiple

Table 3
Johnson et al., 2008)oni, 2018)15;Leander et al., 2020)tructures in the positioning model (e.g.mixture of Gaussian and t distributions) allowing for better handling of multipath propagated detections and noise with TOA.Langrock et al., 2012) and Step Selection Functions (SSFs, Potts   et al., 2014), rely on regularly sampled trajectories with negligible positioning error, requirements that are rarely satisfied by the raw positioned data from acoustic telemetry.A series of positions for each individual will be combined to form a track, but positioning methods are liable to generate outliers, positions that are improbable givenet al., 2021;Baktoft et al., 2015;Leander et al., 2020).Different positioning algorithms provide different measures of uncertainty that can help with the process.The Vemco Positioning System (Table3) provides an estimate of expected positional error at the estimated location for each pair of coordinates (e.g.horizontal positioning error[HPE]or dilution of precision [DOP]).HPE is calculated based on the geometry of the receiver network and users have in the past used HPE values to remove positions that exceed a threshold(Meckley et al., 2014).However, filtering on HPE can potentially introduce a spatial bias in the data.Previous studies in highly reflective environments (e.g.concrete walls or sheer rock faces) indicated the absence of a relationship between HPE and a real error measure in metres obtained through fixed reference tags for which the actual position was known, highlighting the limits of this filtering method(Vergeynst et al., 2020).There are no standard alternatives to HPE that are publicly available, meaning that position filtering with the Vemco Positioning System can benefit from validations in each study system.Other positioning solvers such as YAPS (Table3) output estimated standard deviation of both coordinates for each position analogous to the familiar standard errors from other statistical models, but less rigorous validations have been conducted to test the relidetections.Despite the availability of simple methods to estimate positions at non-observed time stamps (e.g.linear interpolation and smoothing,McLean & Skowron Volponi, 2018), the application of movement models such as the Continuous Time Correlated Random Walk (CTCRW,Johnson et al., 2008)are used to predict plausible and regularly spaced trajectories that maintain the general To date, however, no systematic experiment has evaluated the increase in positioning accuracy derived from using dGPS to record receiver locations and calculate positions.Even with a dGPS, it can be challenging to obtain accurate coordinates from a receiver deployed on the bottom or close to it when measuring from a boat at the surface, especially in deep waters or when flows are rapid.Some positioning methods can estimate the position of errant receivers (e.g.VPS will re-estimate receiver positions prior to fitting, YAPS can re-estimate receiver positions if individual receivers have moved; Table3), but it is best if the majority of receivers remain at a fixed and accurately measured position.Once the coordinates of the receivers are recorded, they should ideally remain fixed over the course of the study.Locations should be measured at the beginning and end of deployments to verify.