The evolution of the Swallow float
to today's RAFOS float

     Fluid motion can be described in two ways, either as a flow past a point such as measuring winds with an anemometer, or by marking a fluid parcel and tracking its movement. In oceanography we have seen a tremendous growth in the use of the latter technique to study the circulation of the ocean. A major appeal of the method is that it gives the researcher good insight into the structure or pattern of flow. A point measurement with a current meter, although precise, tells the observer very little about the subsequent movement and fate of the waters flowing past it. In this note we give a brief overview of the evolution of drifter techniques to study the subsurface circulation of the ocean.

     The first neutrally buoyant float for tracking water movements at depth was developed by John Swallow, a British oceanographer. It consisted of an aluminum pipe with a battery and timer circuit that would excite a magnetostrictive transducer, a "pinger", hanging underneath (Swallow, 1955), Figure 1. The float remains stable at a given depth because its compressibility is less than that of seawater. To see this, suppose it is displaced down from its equilibrium depth (pressure) to a greater pressure. Because of its stiffness, it will not compress as much as the surrounding waters, making it slightly buoyant. This causes the float to rise back to where its weight equals that of the water displaced (Archimedes' principle). The high frequency and low power level of the acoustic pinger on Swallow's float were such that it could only be tracked from a surface vessel nearby, Figure 2. Despite this limitation the float concept was a great success and contributed to several studies including the discovery of a south flowing deep western boundary current (Swallow and Worthington, 1961) that had been predicted by Stommel (1957). The Swallow float was also used in a major study of deep currents west of Bermuda. The premise had been that these currents would be rather weak revealing the interior broad scale abyssal circulation closing the rapidly flowing deep western boundary flow. Instead, very energetic mesoscale motions on time scales of a few weeks were observed, precluding any accurate estimate of the mean circulation at depth (Crease, 1962). It was this study that increased our awareness of an energetic deep ocean. But it also made clear the high cost of tracking floats for any length of time from surface vessels, however small or inexpensive.

     It had been known for some time that sound can propagate great distances through the ocean and that this acoustic 'transparency' of the ocean could be used to track subsurface drifters. Stommel, in a letter to the editor in Deep Sea Research in 1955, proposed to deploy subsurface drifters that would be located from time to time by generating explosive signals that would be picked up at distant hydrophones (underwater microphones) using the deep sound or SOFAR (SOund Fixing And Ranging) channel (see next paragraph). Stommel had in mind a float that would release a small explosive device that would detonate when it had reached a certain depth or pressure. At the same time it would release a compensating buoyant element so that the float from which these were released would remain at the same depth as it continued to drift. Actually, he had already brought this idea up in an earlier paper on turbulent diffusion (Stommel, 1949) where he suggested using the SOFAR channel to study eddy diffusion - the dispersion of fluid parcels - on the gyre scale. Nothing came of these ideas, but they do reflect an awareness of the potential for the newly discovered deep sound or SOFAR channel for tracking subsurface drifters over great distances for extended periods of time.

     The SOFAR channel, an acoustic waveguide, was discovered during World War II at the Woods Hole Oceanographic Institution as part of its research into the acoustics of the ocean, a very important aspect of antisubmarine warfare (Ewing and Worzel, 1948). The SOFAR channel is an acoustic wave guide situated below the warm upper ocean. The speed of sound is a positive, nearly linear function of both temperature and pressure such that the speed of sound first decreases with increasing depth due to the rapid decrease in temperature, but beyond a certain depth below which temperature only slowly further decreases, the pressure effect dominates such that the speed of sound increases with further increase depth. This minimum in sound velocity, typically about 1490 m/s around 1000-1300 m below the surface in tropical and subtropical waters, gives rise to a sound channel which tends to trap sound near these depths, in short an acoustic wave guide. Figure 3 (widely reproduced from the 1948 Ewing and Worzel paper) illustrates the propagation of sound through the ocean thanks to this wave guide.

     In 1965-1966, at Stommel's initiative, M. J. Tucker, on sabbatical leave from England at MIT and Douglas Webb at the Woods Hole Oceanographic Institution (WHOI), developed a low frequency piezoelectric transducer, small enough for neutrally buoyant float applications (Webb and Tucker, 1970). It was tested from the RV Chain in early 1966. Figure 4 shows Gordon Volkmann, a researcher at WHOI, with the ring-shaped transducer prior to lowering it to the SOFAR channel from the research vessel Chain. It was heard at a hydrophone at sound channel depth (~1200 meters) at Bermuda 270 km away. This was a major breakthrough for it confirmed that purposeful acoustic signaling through the ocean over hundreds of kilometers was possible. Questions remained about the attenuation of sound in the ocean, i.e. how far one could transmit, but this was a crucial first step.

     Later that year (1966) Tom Rossby joined the Stommel/Webb initiative to develop and deploy a first long-range Swallow float or SOFAR float as it became known. The first one was deployed in January 1968 and could clearly be heard at 846 km distance. Specially built signal detection and recording equipment using existing SOFAR hydrophones at Eleuthera (in the Bahamas), Puerto Rico and Bermuda had been set up to listen to and track the float. Unfortunately, the float failed in only a week's time. A second deployment a few months later lasted only 2 days. After a substantial redesign effort, a third float was deployed in October 1969 and was tracked for four months when it failed. One possibility for the first two failures may have been that the cables, which hung loosely between the aluminum sphere and the transducer, were vulnerable to biological attack (fish bite), but this was only a guess. It was for this reason the float was redesigned to make it more rugged, Figure 5. The transducer sits inside a polypropylene boot only the top part of which is visible in the photograph (due to the protective frame used for deployment).

     Regardless of the reasons for the short life lengths, it became clear that the high cost of fabrication for the machined aluminum sphere and the piezoelectric transducer hanging underneath would preclude their use in large numbers. To address this, Doug Webb suggested using a resonating tube instead. It would be excited by means of a thin piezoelectric bender plate at the closed end of the tube, the length of which would be carefully trimmed to achieve resonance at the desired operating frequency. Instead of spheres, flotation would be provided by long, commercially available aluminum tubes which would also provide the housing for the battery pack. True, the floats would be much larger than the compact spheres seen above, but this could be countered by the use of suitable handling tools. The first design, Figure 6, had two resonator tubes, one to each side of the center flotation tube. Later, a single tube was mounted end-fire to the main tube, Figure 7. Figure 8 shows an array of SOFAR floats ready for deployment in the Mid-Ocean Dynamics Experiment (MODE) in 1973. Figure 9 shows a float in the process of being deployed.

     In the early days of SOFAR float work, the floats were tracked using a network of hydrophones around the western North Atlantic that had been deployed in the 1950s as part of system to determine the splashdown point of missiles fired from Cape Canaveral. The missiles would release a small explosive charge that detonates once it reaches the depth of the SOFAR channel. From the arrival time of this signal at three or more hydrophones one could determine the splashdown point to considerable (<1 km) accuracy. While these hydrophones simplified the early work with floats, they precluded working in other areas of the ocean. In the late 1970s Dr. Al. Bradley (1978) at WHOI developed an autonomous listening system (ALS) for use on moorings. This very successful development enabled major research programs with SOFAR floats all over the Atlantic ocean. The data are stored in the ALS until it is retrieved when the mooring is recovered at the end of the study.

     The SOFAR float was a clear technical and scientific success, but it was not inexpensive to deploy and use. The tracking data stored in the ALSs had to be retrieved before the analysis of the experiment could begin. This necessitated a second cruise to retrieve all moorings. Figure 10 shows conceptually the arrangement between the SOFAR float and the listening system.

     In the early 1980s our group reversed the tracking system concept by mooring the sound sources and letting the floats do the listening. The floats consist of 1.5-2.2 meter long glass pipes that contain the hydrophone and signal processing circuits, microprocessor and battery. This new float is known as a RAFOS (SOFAR spelled backwards to indicate the opposite direction of acoustic signaling) float (Rossby et al., 1986). These floats weigh only a few percent of the SOFAR float reducing the materials cost substantially as well as making them easier to handle. Figure 11 shows a float just about to be launched with another waiting nearby. Figure 12 shows schematically a RAFOS float in one of its earlier configurations. The float is about 1.6 m long and weighs 10 kg. Figure 13 shows the largest version of a RAFOS float ever built, about 2.2 meters long. It includes a pump mechanism that forces it to float up and down to neighboring density surfaces (Rossby et al., 1994). At the end of their missions the floats drop a ballast, return to the surface and telemeter their data to a satellite system, Service ARGOS, designed to collect data from small transmitters anywhere around the globe.

     The moored SOFAR floats provide the acoustic navigation for the drifting RAFOS floats. We usually deploy at least four (and often more) sound sources widely separated to provide good navigation over a wide area while also providing some redundancy in case of premature failure. Figure 14 illustrates how the RAFOS navigation system works. Three sound sources, each one transmitting an 80 second long tone at 260 Hz on a precisely timed schedule. The signals spread out radially from the source and are detected by the drifting float. The time of arrival of the signals gives us the distance from the float to each of the sources. From this information we can determine its position to within a few kilometers accuracy.

     The glass pipe deserves special discussion. Borosilicate glass has a number of very attractive properties: it is transparent to radio waves, it is completely resistant to saltwater corrosion, as an industrial material it is widely available at low cost, it is strong, and it has a very small coefficient of thermal expansion. Its transparency means that the satellite radio antenna can sit inside the pipe, simplifying construction and increasing reliability; its extraordinary strength means that it can operate at great depth (as deep as 4,000 meters). The glass pipe is also very pleasing to the eye!

     But it is the small coefficient of expansion of the glass that led to the realization of the isopycnal float, a float that would stay with waters of a given layer as they move about. To follow a fluid parcel as it moves vertically it needs to have the same compressibility as the ambient waters (canceling out the pressure restoring force used by Swallow for his constant depth floats as mentioned earlier). We achieve this by adding a compressible device consisting of a spring-backed piston that moves in or out in proportion to the ambient pressure. By carefully adjusting the diameter of the piston we can match the compressibility of the float package to that of seawater to within 1%. Figure 15 shows a sketch of a compressee and how it hangs underneath the float. But we also need to make sure that the parcel stays with water of the same temperature as it moves up or down. This is taken care of automatically because the float has a much smaller coefficient of thermal expansion than that of seawater (~1/10th): Imagine a float bobbing up into warmer waters. These waters are less dense because of their greater temperature. But the borosilicate glass float doesn't expand, thus making it heavier than the surrounding waters. It will therefore sink back to its equilibrium temperature. This argument can be generalized to include salt since the float is completely impervious to salinity variations (Rossby et al., 1985). The result is that the float will always stay at the same layer of water regardless of temperature or salinity since neither affect the volume (density) of the float. It is a natural water parcel follower where the waters are stratified.

     Another type of float is what we call the COOL float (COastal Ocean Lagrangian). This float is designed to measure fluid motion across density surfaces in cases of strong upwelling. For example, off the Oregon coast in summer, fluid parcels upwell from depth along the shoaling bottom and flow back out to sea along the surface. As the waters upwell, there is also downward mixing of lighter surface waters, i.e. the waters become lighter as they shoal towards the surface. The isopycnal float discussed above can't follow through since it will remain at the same isopycnal (more accurately the same specific volume anomaly layer). The COOL float resolves this by sensing the presence of upwelling motion. It is equipped with slanted vanes that cause the float to rotate if there is any vertical motion past it, Figure 16. The float senses the rotation using the earth's magnetic field. A future development will add a small pump so that the float could then increases its volume, i.e. buoyancy, just enough to ensure that this rotation does not take place. In other words the float will rise with the waters so that - on average - there is no vertical velocity relative to the float. This added buoyancy equals exactly the buoyancy that is mixed down from the surface. The hydrodynamics of this float has been tested extensively but the this new version of the float has not yet been used in a field program.

     Our group has also developed a float that follows overflow waters as they spill across a submarine ridge from one ocean basin to another. In this design the float drags a very light line along the bottom keeping it at a fixed elevation, Figure 17. As the waters flow downslope the float remains within that bottom layer. This concept has been tested successfully in the northern North Atlantic where waters from the Nordic Seas spill into the deep North Atlantic. Unfortunately, the probability of the line snagging on the bottom is rather high. We hope to develop a float with an acoustic altimeter, a device that measures the distance to the bottom, and a pump to adjust its buoyancy to stay within a certain distance. Such a float will have many applications.

     In summary, the range of float activities has exploded since the days of John Swallow. The variety of designs is enormous. Here we have focused on those that use the SOFAR channel for acoustic tracking. But there are many other concepts: floats that profile the temperature and salinity structure of the upper ocean as they drift, floats that sample the internal wave field and mixing processes, and floats designed to study convective processes.

     Much of the early developmental work was funded by the Office of Naval Research. In the 1970s the National Science Foundation supported our participation in the major international field programs MODE and POLYMODE. We thank ONR and NSF for their continued support and encouragement. Besides those mentioned in the text, several key people played an important role in the development of these float techniques, Jim Fontaine and Don Dorson at URI, Ken Fairhurst, Al Bradley and Pierre Tillier at WHOI, and Bruce Magnell at MIT. There are many others who have helped in various ways along the way.

Bradley, A., 1978. Autonomous listening stations. Polymode News (4). Unpublished manuscript.

Crease, J., 1962. Velocity measurements in the deep water of the western North Atlantic. J. Geophys. Res.,67,3173-3176.

Ewing, M. and J. L. Worzel, 1948. Long-range sound transmission. Geol. Soc. America,27,35 pp. plus 4 pl.

Rossby, T., D. Dorson and J. Fontaine, 1986. The RAFOS system. J. Atmos. Oceanic Tech.,3,672-679.

Rossby, T., E. R. Levine and D. N. Conners, 1985. The isopycnal Swallow float - A simple device for tracking water parcels in the ocean. Prog. Oceanogr..,14,511-525.

Rossby, T., J. Fontaine and E. C. Carter, Jr., 1994. The f/H float - measuring stretching vorticity directly. Deep-Sea Res.,41,975-992.

Stommel, H., 1949. Horizontal diffusion due to oceanic turbulence. J. Mar. Res.,8,199-225.

Stommel, H, 1955. Direct measurements of sub-surface currents. Deep-Sea Res.,2,284-285.

Stommel, H., 1957. A survey of ocean current theory. Deep-Sea Res.,4,149-184.

Webb, D. C. and M. J. Tucker, 1970. Transmission characteristics of the SOFAR channel. J. Acoust. Soc. America,48,767-769.

Swallow, John, 1955. A neutral-buoyancy float for measuring deep currents. Deep-Sea Res.,3,74-81.

Swallow, J. C. and L. V. Worthington, V., 1961. An observation of a deep countercurrent in the western North Atlantic. Deep-Sea Res.,8,1-19.