Development and application of 'RAFOS Fish Tags' for studying fish ...

INTERNATIONAL COUNCIL FOR THE EXPLORATION OF THE SEA

ICES CM 2006/Q:16 Use of data storage tags to reveal aspects of fish behaviour important for fisheries management

Development and Application of ‘RAFOS Fish Tags’ for Studying Fish Movement by Conrad W. Recksiek, Godi Fischer, H. Thomas Rossby, Steven X. Cadrin, and Prasan Kasturi Abstract The newly developed ‘RAFOS fish tag’ reverses the tracking process of conventional acoustic tags by receiving acoustic signals from moored sound sources, allowing triangulation of geographic position during deployment on fish. We report progress in developing this archival tag for geolocating juvenile and adult demersal shelf fishes. The tag and navigation system are similar in concept to those of isopycnal RAFOS floats, in which arrival times of low frequency tones broadcast from anchored sources are archived and later retrieved for retrospective positioning. The principal differences between the RAFOS fish tag and RAFOS floats is that the tag is small enough to be attached to or implanted in fish about 50 cm or larger, and the tags must be recovered from the tagged fish to download data. Prototype RAFOS fish tags are being deployed on adult yellowtail flounder, Limanda ferruginea, on Georges Bank to study movement in the vicinity of an offshore area that is closed to fishing. Deployment of sound sources will be on or along the edge of the continental shelf where detection ranges appear to be on the order of 100 to 120 km for sources generating a sound pressure of 180 dB re 1 μP. The size of the prototype is governed by dimensions of a cylindrical housing which functions as the hydrophone. Within this is a full-custom 0.5μm feature size receiver chip, memory chip, timing crystal, two batteries, and pressure sensor (temperature sensor is on-chip). The receiver chip consumes 36 μW at 3 V with an expected data storage life of several months to two years. Keywords: acoustic positioning, acoustic tracking, archival tags, data storage tags, electronic tags, RAFOS fish tag, passive listening, retrospective positioning, yellowtail flounder Not to be cited without prior reference to the authors Conrad W. Recksiek: Fisheries, Animal, and Veterinary Science, Woodward Hall, University of Rhode Island, Kingston, RI 02881, USA [tel: +1 401-874-2334, fax: +401-874-6160, e-mail: [email protected]] Godi Fischer: Electrical and Computer Engineering, Kelly Hall, University of Rhode Island, University of Rhode Island, Kingston, RI 02881, USA [tel: +1 401-874-5879, fax: +1 401-782-6422, email: [email protected]] H. Thomas Rossby: Graduate School of Oceanography, University of Rhode Island, South Ferry Road, Narragansett RI 02882, USA [tel: +1 401-874-6521, fax: +1 401-874-6728, email: [email protected]] Steven X. Cadrin: NOAA/UMass CMER Director, School for Marine Science and Technology, 838 South Rodney French Boulevard, New Bedford, MA 02744-1221, USA [tel: +1 508-910-6358, fax: +1 508-9106396, email: [email protected]] Prasan Kasturi: Electrical and Computer Engineering, Kelly Hall, University of Rhode Island, University of Rhode Island, Kingston, RI 02881, USA [tel: +1 401-874-5861, fax: +1 401-782-6422, email: [email protected].

0

Development and Application of ‘RAFOS Fish Tags’ for Studying Fish Movement Introduction We report progress in developing an archival tag capable of identifying sounds broadcast from distant, moored sources where the sounds’ arrival times are used for estimating a (tagged) fish’s position. The tag and sound source system we describe here is designed for geolocating juvenile and adult demersal shelf fishes; we hope to test the system on yellowtail flounder, Limanda ferruginea, of Georges Bank. The system concept could be applied to pelagic nekton offshore and into deep waters; and it could be used, within the physical constraints of the system, for retrospectively tracking most anything in the water column. General principles The technical approach builds on decades of research and development for tracking ocean currents (Rossby 2003) by means of subsurface drifters capable of broadcasting or receiving sound. Beginning in the 1950’s oceanographers used the deep sound or SOFAR (SOund Fixing And Ranging) channel to geolocate subsurface floats (Stommel 1955). Given the time of the float’s sound emission and the time of the emitted sound’s arrival, the float’s range may be estimated. A second receiver provides a fix. A ‘SOFAR float’ system is closely analogous to conventional systems for acoustically tracking aquatic animals, in which acoustic tags transmit to a receiving array (e.g., Urquhart and Smith 1992) or tracking vessel. In the early 1980s oceanographers reversed the tracking system concept by mooring the sound sources and letting the floats do the listening. This led to coining the word, RAFOS, SOFAR spelled backwards, to indicate the opposite direction of acoustic signaling (Rossby et al. 1986). The RAFOS tacking system has been applied extensively in oceanographic studies, e.g., Bower et al. 2002. The RAFOS system was designed for studying ocean currents by retrospective analysis of a listening/recording float’s archived sound source arrival time data. We have simply applied the same idea to studying fish movements, using ‘RAFOS fish tag’ to describe the system. The RAFOS sound source (Rossby 2003, Rossby et al. 1986) amounts to an aluminium, open, 35-cm diameter, 2-m pipe with a transducer inside and a power supply attached. For open ocean work the sources are typically deployed near where sound speed will be at a minimum, the SOFAR channel; in shallow waters at mid-depth. There is usually no surface marker; the float and mooring may be retrieved with an acoustic release or may simply be left where it is (because the cost of recovery exceeds the value of the source). Typically three sources are deployed. RAFOS sound sources, insonifying the deep ocean, may be detected at distances of one to two thousand kilometers, depending upon the sound channel characteristics. Where the thermocline is absent, such as in boreal oceans, minimum sound speed is at the surface. The RAFOS source emits a low frequency tone centered at 260 Hz, having a duration of 80 sec, on a predetermined schedule. For tracking RAFOS floats in the Gulf Stream, for

1

instance, tones would be broadcast twice daily. The tone, known to the community of oceanographers studying currents this way, is referred to as a pong. The tone changes frequency steadily, i.e., it sweeps, over its broadcast interval, viz., increasing from 259.375 to 260.898 Hz over 80 sec. The strength of the signal at 1 m is on the order of 180 dB re 1 μP. The motivation for ramping the frequency, or sweeping the frequency, is that such sound is unique in the ocean; the chance of confusing that pattern, i.e., a ramped, low-frequency sound, with ships and natural sounds, physical and biological, is minimal. The RAFOS float package (or hull) (Rossby 2003, Rossby et al. 1986) is a 15-cm diameter, 2-m PYREX glass pipe. It is ballasted to float vertically and is engineered to enable it to drift along isopycnal surfaces thereby tracking fluid motion (advection and diffusion) – hence the appellation isopycnal. By means of a hydrophone on the bottom end, the float listens and when the unique, ramped low-frequency signal is detected the arrival time is recorded. Knowing when the RAFOS source emits its tone, together with sound arrival time, yields a time difference between float and source; a range may be obtained given an appropriate speed of sound. With sound arrival times from a second source, a fix may be obtained (Figure 1). Note that if three or more signals are available, hyperbolic navigation may be employed to obtain a fix, since the locus of equal sound arrival time difference is a hyperbola (as with LORAN). Besides sound source arrival times, the float records ambient pressure and temperature. At the end of its mission, a ballast weight falls away, the float reaches the surface, and, through an antenna inside the glass pipe, its data are uploaded to an ARGOS satellite. The float track and temperature/pressure records are retrospectively derived from the data set recovered from ARGOS (see http://www.po.gso.uri.edu/rafos/ for additional details on RAFOS technology). The RAFOS fish tag system works the same way. The sound source side of the system is identical. The principal difference is that the tag is small enough to be attached to or implanted in fish about 50 cm or larger, and the tags must be recovered from the tagged fish to download data. The capability of breaking free from the fish at mission’s end is of course feasible but an antenna and RF system would make the device more costly and larger. The size of such a device, i.e., one capable of broadcasting radio frequencies at the surface, would be a function of the power needed to communicate with satellites. RAFOS fish tag system and operational environment A study of fish movements using RAFOS would be, on the ‘tagging side,’ approached and conducted like an archival tag study, e.g., Cadrin et al. 2005, Cadrin and Westwood 2004, NMFS 2005. The logistics amount to tagging and releasing animals and waiting for the fishery to return tags. Clearly, the closest operational environment to a RAFOS-based study of ocean currents would be a fish movement study on pelagic species of the deep sea. Indeed RAFOS technology may ultimately prove most fruitful there because physical oceanic gradients used to geolocate conventional archival tags can be very slight. Consequently we have

2

chosen to implement the prototype system on the continental shelf, on a flatfish species in particular – in our case Georges Bank yellowtail flounder (Figure 2). Our group (SXC) has experience, cited previously, with that species and the region. For yellowtail flounder, archival tags, as in studies of other finfishes, have led to stunning discoveries (and confirmation of some scientists’ and fishermen’s conjectured movements) of offbottom behavior (Cadrin and Westwood 2004, Walsh et al. 2004; for historical evidence of off-bottom movements see Walsh 1991 and Walsh 1998). We consider yellowtail flounder to be a reasonable RAFOS prototype ‘test platform.’ Recent holding studies indicate that tagging-induced mortality of yellowtail flounder (including the capture system and attachment of a tag with a nickel pin) is negligible (Alade et al. 2005). Capture rates of tagged fish in the trawlnet fishery on Georges Bank are relatively high because of intense fishing pressure, and reporting rates are extremely high for a commercial fishery (see http://www.cooperative-tagging.org for additional details). Cadrin and Westwood (2004) deployed 131 archival tags in 2003 and report acquiring 33 of them. An advantage of using a flatfish as a test platform for the RAFOS fish tag prototype is that no endogenous, swim-bladder amplified sounds are expected (acoustic filtering is discussed below). There is a great operational difference between tracking a RAFOS float in deep ocean and tracking a RAFOS fish tag. Sound propagation at ‘RAFOS frequencies,’ 200 Hz – 300 Hz, on the continental shelf is limited to ranges in the neighborhood of 100 km depending upon the particulars, and these may vary considerably depending on water depth and sediment type. Astonishingly, sound propagation at RAFOS frequencies on the continental shelf remains poorly understood. During September 2005 (Rossby et al. 2005, Fischer et al. 2006), we studied propagation loss from RAFOS sound source deployments off the northern edge of Georges Bank and on the continental shelf of southern New England, USA (Figure 3). For the usual RAFOS 1-m signal source strength, 180 dB re μP, Fischer et al. (2006) report a “realistic receiver detection range of 100 km - 120 km,” at least on wide, temperate shelves during early fall. RAFOS fish tag configuration and specifications The RAFOS prototype archival fish tag is a cylindrical device; the first tags will have a diameter of 15 mm and length of 35 mm, with a weight in water of 3 gm. Affixing the tag to the fish and all other archival tagging of fish logistics will be as described elsewhere (Cadrin and Westwood 2004). Upon recapture, the tag will return time, pressure, temperature, and an acoustic record of an identified source sound with its time stamp. This prototype will record data only during a programmed listening window. For instance, the tag may listening, i.e., be ‘awake’ or on, i.e., operational as opposed to being in a standby mode, only during evening hours. This is to conserve batteries and memory space. (Hereafter we use operational to denote the listening mode and standby mode to denote the non-listening mode.) While operational, the tag would record time, temperature, and pressure at programmed intervals. All the while, the signal from the hydrophone would be filtered

3

and then sampled for the arrival of a RAFOS source sound. Should the sampling yield a probable source sound, the time and elements of the sound signal would be recorded. The latter data set, that derived from a probable source sound, will be used in post processing to establish the validity of the source sound and the best value arrival time so as to estimate the most accurate range. Considerable retrospective analysis of the time, temperature, and pressure records and of the sound data is anticipated. Post calibration of pressure sensor and temperature circuit will be required. So will measurement of the timing crystal output so as to correct time. Note that post processing is more cost effective then calibrating every tag deployed; tag returns on the order of a few percent are expected so obtaining calibration parameters on all tags would be effort wasted. For the prototype, archived data transfer will take place by means of a wire connection when the tag is activated after recovery. For activation before deployment, following performing system tests, the sampling parameters would be loaded. Then communications would be terminated, the wires cut, and the device sealed for tagging a fish. While operational, the device consumes 36 μW at 3 V. This corresponds to a maximum supply current of 12 μA. In the standby mode, however, the supply current should not exceed 1 μA. Assuming an active duty cycle of 50 %, i.e., a scenario where the acoustic receiver remains operational for 12 h per day, a 3 V battery with a capacity of 100 mAh would enable an active tag lifetime (defined as the time span during which battery power is available to power the circuitry) of about 460 d. The tradeoffs are battery capacity, the mass of the device being a function of battery size, and the amount of time the device is operational. The latter will depend on the goals of the application. In our Georges Bank yellowtail flounder study, we are considering using two 7.9 mm diameter AgO button cells with a total capacity of 80 mAh and a terminal voltage with two batteries in series of 3 V. Our immediate scientific goal is to understand the details of the species’ departure from the bottom during the night, so we would likely program some tags to be operational then, i.e., be in standby mode for 12 hr during the day – a 50% duty cycle. Data are stored in a dedicated non-volatile memory, so when the batteries are exhausted, after about 410 d for a 80 mAh power supply over a 50% duty cycle, the data remain archived until the tag is recovered. The physical size of the device is governed by dimensions of a cylindrical housing which functions as the hydrophone (Figure 4). In turn, the inside diameter of the casing/hydrophone is a function of the choice of batteries and duty cycle, which, as described above, control the operational lifetime. The dimensions of the circuit board, or electronics unit (Figure 4), follow from battery and hydrophone dimensions. On the circuit board is a full-custom 0.5-μm feature-size receiver microchip, the heart of our system, plus non-volatile memory chip, timing crystal, bias resistor, shunt capacitor, and 5 pads for communication purposes. The pressure sensor is mounted on an end of the cylindrical casing. Temperature is sensed by a circuit implemented on the custom receiver microchip, thus obviating the necessity of one component on the circuit board.

4

For our yellowtail flounder archive tagging, given our power requirements, the prototype measures 15 mm in diameter and 35 mm in length. Engineering design of RAFOS fish tag receiver chip The full-custom 0.5-μm feature-size acoustic receiver microchip performs system control, signal processing, data processing, and temperature sensing functions of the RAFOS fish tag device. Utilizing a state-of-the-art CMOS 1 process having a sub-micron feature-size has enabled the design of a relatively small (therefore less expensive), and low power application specific integrated circuit (ASIC) which consumes on the order 10 μA (12 μA maximum) while fully operational and less than 1 μA in its standby mode. In addition to its function as the tag’s system controller and processor, the ASIC handles all communications with the external world via a personal computer (Figure 4). The architecture of the RAFOS microchip is presented in detail elsewhere (Fischer et al. 2006). The receiver microchip acts upon the programming parameters loaded into it; it controls the timing of operational episodes, i.e., the duty cycle, and the time, temperature, and pressure recording accordingly. While operational, it executes the filtering and signal processing of the ambient acoustic environment ever ‘searching’ for a source sound arriving. Once having identified an incoming signal, it archives time and signal data for postprocessing. Whereas the heart of the RAFOS fish tag system is the ASIC, the CMOS microchip, the critical part is the architecture which permits identifying, and time stamping, a RAFOS source sound on the continental shelf. In contrast to a RAFOS float which receives an 80-sec pong sweeping between 259.375 and 260.898 Hz, to help eliminate environmentally induced signals the RAFOS fish tag is programmed to detect 32-sec pongs sweeping between 263.000 Hz and 264.523 Hz. After received sound energy is changed to electrical energy, the latter is filtered such that frequencies below and above those emanating from the RAFOS sound source are removed. The filtered signal is amplified, digitized and the frequency is divided, i.e., ‘lowered,’ several times for comparison with the expected signature of a RAFOS sound source. An on-chip sampling algorithm compares the incoming stream of data with an internal data set, implemented in the hardware, which emulates the filtered and divided signal from a RAFOS source sound. Comparisons between environmental data and the internal data set are calculated such that enough matches indicate that the incoming sound is from a source. This process, a cross correlation, is very discriminating and disqualifies most of the remaining parasitic in-band signal components. If the correlator yields a good match (which is determined by a numerical threshold, essentially a correlation coefficient), the signal strength and its arrival time are stored in the non-volatile memory chip together with temperature and pressure. The time of that taking place is recorded and the incoming bit stream either side, in time, of that is recorded as well for retrospective analysis. It can be appreciated that signal processing must be going on constantly while the device is 1

The acronym MOS denotes the three most important layers forming a transistor, namely Metal-OxideSemiconductor, while the preceding C stands for complementary (logic).

5

operational; this can consume considerable energy but implementing the design in 0.5μm CMOS has overcome this obstacle. Discussion The success of the RAFOS fish tag system hinges upon the obtaining an accurate fix. That depends on calculating accurate ranges between tag and RAFOS sound sources. At this point in the system’s development, we are confident that the device is capable of recording an arrival time from a source around 100-km distant. Obtaining a fix, as in any navigation problem, depends on there being two or more sources within detection distance. The detection limit depends on the sound transmission quality of the water column between source and tag and the amount of ambient noise, including that provided by the animal. Since the device is about the same size of conventional archival tags we expect the logistics of tagging and recovery to be about the same. Deploying and recovering a sound source is a routine matter; using acoustic releases or even using grapnels to recover sources is practical. Keeping sources from being carried away by commercial fishing gear becomes a serious problem and it is essential that a study’s investigators understand fishing operations in the area. Navigational accuracy is primarily constrained by timing accuracies. Speed of sound is important but an uncertainty in speed of sound of 1% due to temperature uncertainties expected on the continental shelf would imply errors of 100 m at 100 km. This is small compared to time of arrival uncertainties and clock corrections which will be at the 1 sec level (equivalent to a distance of 1500 m). This is a new area for us so experience will be an important guidepost in the next couple of years. Thus, the crystal oscillator requires hyperbolic navigation from time to time to determine clock drift in the tag, but in general range-range geolocation using the best received signals results in higher quality tracking. The more rapid sweep of the RAFOS sound source signal in the fish tag application, i.e., 32 sec in the fish tags versus 80 sec in oceanographic floats (discussed above), reduces time of arrival errors due to swimming-induced Doppler shifts. If during reception of a signal the fish is swimming towards the source (or moving due to currents) at 1 Knot, the estimated arrival time will be 1.6 sec early resulting in a range error of 2.3 km. One would expect more Doppler Effect problems in tracking large pelagic fishes, like tunas, than with flatfishes.

6

References Alade, L., A. Westwood, A. Johnson, S. Cadrin, P. Rago, and E. May. 2005. A pilot study on tag-induced mortality of yellowtail flounder (Limanda ferruginea). Proceedings of the 135th Annual AFS Meeting, Anchorage, Alaska. [Available online at http://www.fisheries.org.] Bower, A.S., B. Le Cann, T. Rossby, W. Zenk, J. Gould, K. Speer, P. Richardson, M.D. Prater and H.M. Zhang. 2002. Directly-measured mid-depth circulation in the northeastern North Atlantic Ocean. Nature. 419: 603-607. Cadrin, S., C. Legault, and J. King. 2005. Cape Cod - Gulf of Maine yellowtail flounder. In Assessment of 19 Northeast Groundfish Stocks through 2004. NEFSC Ref. Doc 05-13: 127-148. Cadrin, S., G. Shepherd, T. Sheehan, S. Kubis, J. Moser, A. Westwood. 2005. The use of information from electronic tags for stock assessment of northeast fishery resources. In Advanced Sampling Technology Working Group Tagging Workshop. NOAA Tech. Mem. In press. Cadrin, S.X. and A.D. Westwood. 2004. The use of electronic tags to study fish movement: a case study with yellowtail flounder off New England. ICES CM 2004/K:81 Fischer, G., S. Lee, M.Obara, P. Kasturi, H.T. Rossby and C.W. Recksiek. 2006. Tracking fishes with a μW acoustic receiver – an archival tag development. IEEE J. Oceanic Eng. In press. NMFS (National Marine Fisheries Service). 2005. Advanced Sampling Technology Working Group Tagging Workshop. NOAA Tech. Mem. In press. Rossby, T. 2003. The evolution of the Swallow float to today's RAFOS float. [Available online at http://www.po.gso.uri.edu/rafos/general/history/index.html]. Rossby, T., D. Dorson and J. Fontaine. 1986. The RAFOS system. J. Atmos. Oceanic Tech. 3: 672-679. Rossby, T., C. Recksiek, S. Fontana, P. Kasturi, Y. Fan, C. Mueller, J. Xu, R. Yablonsky, L. Zhou, A. Meade, D. Wilson, and B. Fanning. 2005. Endeavor Cruise 411. University of Rhode Island, Graduate School of Oceanography report. Stommel, H. 1955. Direct measurements of sub-surface currents. Deep-Sea Res. 2: 284-285. Urquhart, G.G. and G.W. Smith. 1992. Recent developments of a fixed hydrophone array system for monitoring movements of aquatic animals. Pp 342-353 in I.G.

7

Priede and S.M. Swift, eds. Wildlife Telemetry: remote monitoring and tracking of animals. Ellis Horwood, New York 708 pp. Walsh, S.J. 1998. Diel variability in trawl catches of juvenile and adult yellowtail flounder on the Grand Banks and the effect on resource assessment. N. Am. J. Fish. Mgmt. 8:373-381. Walsh, S.J. 1991. Diel variation in availability and vulnerability of fish in a survey trawl. J. Appl. Ich. 7:147-159. Walsh, S.J., and M.J. Morgan. 2004 Observations of natural behaviour of yellowtail flounder derived from data storage tags. ICES J. Mar. Sci. In press.

8

Figures Figure 1. Determining retrospective fixes and track of isopycnal RAFOS float at intersection of sound pulse arrival time loci from three moored sound sources, labeled S. RAFOS sound source signals, or pongs, 180 dB re 1 μP at 1 m, are 80 sec in duration, increasing frequency, or sweeping, from 259.375 to 260.898 Hz. RAFOS float position accuracy at the scale illustrated and in ocean waters, i.e., away from the continental shelf, is on the order of a few kilometers. Figure courtesy of Mark Prater, Graduate School of Oceanography, University of Rhode Island, USA.

9

Figure 2. Study area for deploying RAFOS fish tags on yellowtail flounder to evaluate movement in the vicinity of Closed Area II (indicated by gray shading) on Georges Bank.

10

Figure 3. Sound frequency over time detected at a range of 70 km eastward from a RAFOS sound source at 50 m moored on the continental shelf off southern New England, USA (400 34.5’ N 710 05.56’ W) along the 70-m contour, 26 September 2005 (Table 1). Pong sweep, clearly visible here, from 261.000 Hz to 262.523 Hz. Listening hydrophone deployed 10 m off the bottom at 400 37.6’ N 700 16.2’ W. From Fischer et al. 2006, Figure 10.

11

Figure 4. Block diagram of RAFOS fish tag (above) and (below) packaging of circuit board (Electronics Unit) and battery pack within hydrophone. The device detects and archives time-of arrival of acoustic signals from RAFOS sound sources moored on the continental shelf so as to geolocate a tagged animal. Temperature, pressure, and elements of the RAFOS signal are archived on a programmed schedule when the tag is in operational mode. When in operational mode for incoming signals from moored RAFOS sources, the device consumes 36 μW at 3 V with a maximum current drain 12 μA; when in standby mode, current drain is about 1 μA. Above, ASIC, application specific integrated circuit, see text. Below, Custom IC, 0.5-μm feature-size acoustic receiver microchip, see text.

12