Pacific Coastal & Marine Science Center
U.S. Geological Survey
Open-File Report 92-178
Note: This is the text portion of the report. The appendices are not included.
See the Full Report online at the USGS Publications Warehouse:
Between 1946 and 1970 approximately 47,800 barrels (55-gallon drums), concrete blocks and other containers of low-level radioactive waste were dumped on the continental shelf and slope adjacent to the Farallon Islands offshore of San Francisco Bay (Noshkin et al., 1978). Three specific sites were chosen for disposal of the drums (Figure 1). Approximately 150 drums were deposited at the shallow (90 m water depth) site, 3600 at the mid-depth (900 m) site, and 44000 at the deep (1800 m) site. Owing to inclement weather and navigational uncertainties, many of the drums were not disposed of at the specific sites and it is more likely that the drums litter a 1400 km2 area of sea floor, the Farallon Island Radioactive Waste Dump (FIRWD), defined by the irregular polygon shown in Figure 1 (Noshkin et al., 1978). Consequently, the true location and distribution of the drums on the sea floor is unknown. Not knowing the distribution of the drums has impeded attempts to sample the sediment and biota around concentrations of the drums and to retrieve individual drums for study. An unmanned submersible was used to explore the 900-m site in 1974 and three clusters of drums were located (Appendix I); a barrel was retrieved from this site in 1977 with a manned submersible (Dyer, 1976; Columbo and Kendig, 1990). In order to use submersibles efficiently and to design good experiments to sample from surface vessels and submersibles in the FIRWD, it is necessary to have a map of the distribution of the drums.
Much of the FIRWD now lies within the boundaries of the Gulf of the Farallones National Marine Sanctuary (GFNMS) (Figure 1). In summer 1990 the National Oceanic and Atmospheric Administration (NOAA) provided funds for the U.S. Geological Survey (USGS) to survey part of the FIRWD with a sidescan sonar system as part of a major multi-federal agency multi-purpose research cruise (Karl et al., 1990). The purpose of the search was to determine whether the drums could be be detected with sidescan sonar and, if so, to locate clusters of drums and plot their distribution. Other sidescan sonar surveys have been conducted by USGS in the FIRWD, but the purpose of these was not to detect drums. Although these other surveys are mentioned in this report for completeness, only the two surveys conducted in cooperation with NOAA are described in detail.
The principal sidescan sonar surveys designed specifically to detect the barrels of low-level radioactive waste were conducted during cruise F7-90-NC in July 1990 (Karl et al., 1990). These sidescan data were collected with the SeaMARC 1A, operated by Williamson & Associates under contract to USGS. SeaMARC 1A is a deep-towed sidescan-sonar system that operates at a frequency of 27 to 30 kHz and that can be towed at speeds up to 5 knots, although speeds of 1.5-3.5 knots are more typical. The USGS supplied the winch, armored conducting cable, and shipboard data acquisition and computer processing equipment for the SeaMARC 1A. The USGS research vessel, FARNELLA, was used for the survey.
Data were collected in a 70 km2 area around the shallow (90 m) radiation waste site on the continental shelf and in a 120 km2 around the 900 m site on the continental slope (Figure 2). The SeaMARC 1A can be set to ensonify swaths of sea floor that are 5, 2, 1, and 0.5 km wide. Narrower swath widths provide greater resolution. The 90 m site was surveyed using a swath of 0.5 km and the 900 m site using a swath of 1 km. Tracks in both areas were spaced so that adjacent swaths overlapped by 10-20% to obtain a continuous sonographic image of each survey area (Figure 2). Typically the tow vehicle is kept a distance that is 10% of the total swath width above the sea floor for optimum data quality and processing results. Both areas were surveyed in about two days and approximately 15% of the FIRWD was surveyed at these swaths.
SeaMARC 1A was used for two other surveys as part of the multi-federal agency cruise (Karl et al., 1990) (Figure 2). A survey was done for the U.S. Environmental Protection Agency (USEPA) and the U.S. Army Corps of Engineers (USACE) over a large part (3300 square kilometers) of the continental slope with the system set to ensonify a 5-km swath (Figure 3). This survey incidentally included approximately 60% of the FIRWD. Another survey (200 square kilometers) was done for the U.S. Navy (USN) with the system set at a 2-km swath that incidentally covered about 10% of the FIRWD immediately north of the 1800-m site (Figure 2). The 120 km2 NOAA and 200 square kilometers USN surveys are within the boundaries of the 3300 square kilometers USEPA survey (Figure 3).
The USGS collected another set of sidescan data on the shelf and upper slope that included about 5% of the FIRWD with an AMS-120 sidescan sonar, a 120 kHz system, in 1989 (Figure 3). The 30 kHz and 120 kHz surveys combined mapped approximiately 65-70% of the FIRWD.
The analog signal from the sidescan tow vehicle (both the SeaMARC 1A and AMS-120) was acquired with and stored on a QMIPS data acquisition system manufactured by Triton Technologies, Inc. The data were transferred from the QMIPS to a Masscomp computer for processing (see Danforth et al., 1991 for a description of processing techniques). All geometric and radiometric corrections were done in realtime at sea and a hardcopy record of the processed imagery produced on a Raytheon 850 thermal printer. A true plan-view digital mosaic of each survey area was constructed while at sea (Figure 4 and Figure 5). Identical features in the zone of overlap on adjacent sonographic swaths were matched and the hardcopy from the thermal printer fixed to a stable base thereby progressively building the mosaic of the sea floor.
Three types of acoustic-reflection data were collected as part of the SeaMARC 1A survey. Bathymetric data were collected with a 10 kHz acoustic reflection system simultaneously with the SeaMARC 1A data. These data were recorded graphically in analog form on a wet-paper recorder at scan rates of 2 seconds (s). The SeaMARC 1A tow vehicle includes a 4.5 kHz subbottom profiler allowing the simultaneous collection of 4.5 kHz profiles with the sidescan data. Additional subbottom data were collected with a 3.5 kHz seismic-reflection profiler along selected tracks; these data were not collected simultaneously with the sidescan data. Both the 3.5 kHz and 4.5 kHz data were collected at scan rates of 1 s and displayed graphically in analog form on line scanning recorders.
Optical images of the sea floor were obtained along a transect on the continental slope with a remotely operated camera/video system (Figure 6). The camera system was towed for 4 hours along 4 transects each about 1-hr in duration. The transects were parallel with the isobaths and at a nominal depth of 1000 m. The system consists of a 35 mm still-camera that can be programmed to take photographs at a fixed-interval and a video camera programmed to operate through a VCR. The system is powered by batteries and can operate for up to 5 hours and is towed at speeds of 1-1.5 knots. The system does not provide realtime images of the sea floor.
Four systems were used to navigate the ship: (1) Global Positioning System (GPS); (2) LORAN-C, either hyperbolic or rho-rho; (3) shore-based, line-of-sight transponder net (Del Norte system); and (4) long baseline bottom transponder net. The primary system used for real-time positioning was chosen either manually by the navigator or automatically by the computer. Steering of the ship was aided by a trackline-following program displayed on a CRT screen both at the helm and at the navigation station. Positional accuracy of navigation tracks varied between a few meters when within range of the Del Norte or long baseline system to as much as 100 m of the preplotted tracks when using LORAN-C and GPS. The long baseline system was used to navigate only during a small part of the 5-km swath survey.
The navigation coordinates entered into the SeaMARC 1A data acquisition/processing computers were those of the ship's position and not the position of the tow vehicle. The position of the tow vehicle relative to the ship is a function of the length of cable deployed and the speed of the ship. Since we could not range acoustically on the tow vehicle owing to acoustic interference (ship noise), it was necessary to estimate the position of the tow vehicle with respect to the ship. This difference between ship and tow vehicle is called "layback". The shallower the water, the less the layback difference. The positional difference between the ship and the tow vehicle at the 90 m survey site is on the order of 100 m or less, whereas at the 900 m site the difference is on the order of 0.5-1 km. Because of the necessity to estimate layback, specific features on the sonographic mosaics are probably offset from their true geographic position. The amount of offset at the shallow site is within the accuracy of the navigation system used to position the ship (from a few to 100 m). The positional offset of features on the 900 m site mosaic is potentially as large as 1 km. However, we estimate that offsets of 200-300 m are more probable.
The shallow dumpsite is located on a flat (0.2°) area of the continental shelf just east of the shelf break (Figures 1, 4, and 7). According to our navigation and bathymetric charts, the site is situated in a water depth of about 105 m. The area around the site is featureless except for a low ridge of outcropping rock to the north (Figure 4). Based on two grab samples collected near the site, the substrate consists of a uniform blanket of fine and medium sand (Maher et al., 1991).
In contrast, the mid-depth site is located at a depth of about 975 m on the rugged and steep (regional slope of 6° with slopes as steep as 17° locally) upper continental slope (Figures 1, 5, and 7). The specific location of the site is on the side of a submarine canyon (Figure 5). Except for the small triangular area on the continental shelf, most of the FIRWD encompasses a rugged terrain that consists of a series of ridges and canyons (Figures 1, 5, 7, and 8). No cores have been collected immediately adjacent to the 900 m site. The substrate at similar depths (about 1000 m) to the south of the site consists generally of coarse silt (Booth et al., 1989; Karl et al., 1990). Although the slopes are steep in the area of the 900-m site, very little evidence of downslope mass movement of sediment has been detected on the sidescan sonar images and high-resolution seismic-reflection profiles.
The sidescan-sonar mosaics described herein are acoustic images of the sea floor; acoustic energy transmitted from the sidescan tow vehicle is backscattered from the sea floor. These acoustic data have been computer processed so that the mosaics represent a true plan view of the sea floor. That is features on the sea floor seen on the mosaic are in their correct spatial position and their true geometric shape. The shades of gray ranging from black to white that define the features of the sea floor on the mosaic represent varying energy levels of acoustic backscatter. The darker shades correspond to high backscatter levels. Many complex factors determine how sound is backscattered and reflected from the sea floor. Steep slopes and rough bottom are just two elements that backscatter more acoustic energy. We assume that the sidescan sonar is imaging only the surface of the sea bed. This is probably true for the high-frequency systems. However, sound transmitted by the mid- and low-frequency systems is capable of penetrating below the surface of the sea bed under certain conditions. Therefore, some features seen on the mid-range (30 kHz) mosaic may not represent features on the sea floor, but may represent features buried at an unknown depth (typically a few meters) beneath the surface. Consequently, interpretation of the acoustic mosaic is not as straightforward as viewing and interpreting an aerial photograph or satellite image (see Johnson and Helferty, 1990 for a discussion of sidescan sonar). Interpretation of sonograph images is an art as well as a science. Other data sets must be used to supplement and complement the sonar data so that the sonar images can be interpreted as accurately as possible. By so doing, the sonar image can be verified or "ground-truthed". For that reason other data such as high-resolution seismic-reflection profiles, bottom photographs, and sediment samples must be collected in the sidescan sonar survey area.
Resolution, the ability to distinguish closely spaced or small objects on the sea floor, of a sidescan sonar system is a complex function of several variables that include but that are not limited to pulse length, frequency, and pixel size (see summary in Johnson and Helferty, 1990). Fifty-five -gallon drums are very small objects (about 0.6 x 0.7 m) and theoretically beyond the resolution of a 30 kHz system even at the narrowest swath setting of 0.5 km. However, eventhough a feature is smaller than the theoretical resolution of the sidescan sonar system, the object still can be "detected" with that system and make a visible record on the sidescan sonar image (Johnson and Helferty, 1990). In general, in order for an object to be recorded or recognized as a target on the sonograph, the ambient background level of backscattered acoustic energy should be low and uniform and the acoustic energy backscattered from the object must be sufficiently high so that the object contrasts with the sea floor (see summary in Johnson and Helferty, 1990). When surveying over flat, uniform sea floor, Williamson & Associates have detected 55-gallon drums on numerous occasions with the SeaMARC 1A system (oral communication, M.Williamson, 1990). The acoustic energy from the SeaMARC 1A system excites the modal resonances of many targets which are barrel sized and smaller (oral communication, A. Wright, 1991).
The interpretations in this report are based on visual inspection of the hardcopy sonographic mosaics. Because objects as small as 55-gallon drums are "detected" and not "resolved" with the 30 kHz system, the visible record produced on the hardcopy image is often subtle and indistinct especially in areas of high relief, hard substrate, and coarse-grained or rippled sediment. In many cases it is difficult to differentiate the visible record of the small non-geologic targets not only from small geologic features, such as boulders, but also from noise (acoustic artifacts on the images) on the hardcopy records. The interpretation of the images is as much an art, a culmination of skill through experience, as a science. For example, patterns of objects provide a great deal of information. Consider the following: drums would be either dumped at a fixed location while the vessel is stationary or dumped over a variable straight-line distance as the vessel is transiting. Consequently, the drums would accumulate on the sea floor either in isolated clusters or in linear trends depending upon whether the vessel was stationary or transiting, respectively. Patterns such as these are characteristic of a process and, thus, aid in differentiating among noise, geologic and non-geologic objects. Examples of these patterns are illustrated on Figure 10.
When the sidescan data are displayed on a CRT monitor, it is much easier to differentiate between targets (real objects on the sea floor) and noise not only visually but also by using several techniques of image enhancement and analysis. With the aid of notes taken while observing the CRT monitor in realtime during the sidescan survey, we have identified several areas on the mosaics that we confidently believe represent clusters of 55-gallon drums (Appendix I; Figure 9 and Figure 10). Indeed, as discussed below, we verified our interpretation of one area by observing drums on the sea floor with an underwater camera system.
Sea floor conditions are excellent to detect drums at the shallow site. The sea bed is a monotonously flat and featureless blanket of uniform fine and medium sand. Numerous small targets are visible on the sonograph (Figure 9), many of which we have interpreted as non-geologic. Only 150 drums were reported to have been dumped at this site. The area immediately surrounding the specific dumpsite is devoid of targets. This site is located within a ship transit lane and, undoubtedly, much of the debris on the sea floor represents material thrown overboard from passing ships. Because of this possibility, we did not invest any shiptime to identify targets at this site and do not know if any represent drums of radioactive waste.
Part of this area was surveyed by USGS with an AMS-120 kHz sidescan sonar system in August 1989. The AMS-120 data and SeaMARC 1A data overlap in a 15 square kilometer area. USGS provided Williamson & Associates with processed imagery to compare the two sidescan systems with respect to non-geologic target detection. A. Wright (Williamson & Associates) has analyzed the coincident AMS-120 and SeaMARC 1A images for barrel-size targets using modal resonance as a detection and classification aid when inspecting the hard copy sonographs and the images on a CRT monitor. Using this technique, barrel-size non-geologic targets are easily discerned by a skilled operator on the 30 kHz images; many of these targets are poorly discerned or not discerned at all on the 120 kHz records (Wright, 1991). None of these targets were visually verified with underwater camera systems.
Small objects are much more difficult to detect on the rugged continental
slope. Ambient levels of backscattered acoustic energy are relatively high
and sound paths complex owing to the intricate morphology and steep slopes.
Unless a strong signal is received from an object, it may not make a visible
record on the sonar image. Even under these non-optimum conditions, numerous
small targets were detected on the sonographs (Figure
10; Appendix I). Many of these are interpreted as 55-gallon drums. This
interpretation, however, has been verified at only one location (Figure
11). This location, a small canyon just to the south of the 900-m dumpsite,
was chosen for three reasons: (1) numerous small objects occur over the
area, (2) some of the objects are arranged in a linear pattern, and (3)
the sea floor relief is sufficiently subdued so that risk to the underwater
camera system is minimal. Five 55-gallon drums were observed with the underwater
video/camera system (Figure
11 and Figure
12). The drums are in various states of deterioration. One of the drums
observed on the video tape has imploded in the center. Four of the drums
are clustered within a very short distance (100-200 m) of each other. The
video system images an area of about 4 square meters. The fact that the
camera randomly captured 5 drums in so small a field of view suggests that
many more drums were grouped in the area. In fact, a cluster of 28 drums
was found in a 30x60 m area during the 1974 survey sponsored by USEPA (Noshkin
et al., 1978). Because we could not view the images in realtime, we could
not do a detailed search in the vicinity of the 5 drums. The characteristics
of the drums observed on the video tape and 35 mm film are consistent with
the descriptions of the drums containing radioactive waste reported in the
literature (see eg., Columbo and Kendig, 1990) and prove that they are part
of the consignment of 47,800 containers of low-level radioactive waste.
As at the shallow site, no large concentration of targets was detected in the immediate vicinity of the specific dumpsite; targets are distributed over the entire 120 square kilometers area of the sonograph (Figure 10). The targets are not distributed uniformly over the area but concentrations of drums are clustered in discrete areas. Most of the visible targets are in canyon floors and on gently sloping plains. Targets likely litter the steep ridge slopes but probably were not detected owing to the high levels of acoustic energy backscattered from the slopes. Many targets were observed in realtime on the waterfall display on the CRT monitor that are not visible on the hardcopy sonographs. For example, an extremely high concentration of targets that cannot be seen on Figure 10 was observed on the CRT monitor in Area 4, a zone of very high acoustic backscatter. Owing to the combination of steep slopes and limited shiptime, we did not attempt to identify these targets with the underwater camera system.
Most of the 1-km survey overlapped the 5-km survey conducted for the USEPA and the USACE. Although numerous targets are visible on the 1-km mosaic, no targets are visible on the 5-km mosaic in the zone of overlap. Moreover, no targets unequivocally interpreted as 55-gallon drums were observed on the CRT monitor during the 5-km swath survey. Large non-geologic objects objects were identified on the 5-km mosaic. One target is the SS Puerto Rican and another 270 m long target is possibly the USS Independence scuttled in 1951 or a dry-dock scuttled in 1985 (locations marked with "x" on Figure 1).
Some large non-geologic (?) targets are visible on the USN 2-km mosaic, but we did not interpret any targets as drums. Apparently, the SeaMARC 1A is not capable of resolving or detecting objects as small as 55-gallon drums when operated at swaths greater than 1 km. More conclusive computer analyses of the data are necessary to verify this conclusion.
Drums containing low-level radioactive waste that litter the sea floor in the GFNMS are detectable with a deep-towed 30 kHz sidescan sonar system. The surveys of the 90-m and 900-m dumpsites show that the 55-gallon drums are not concentrated at the designated dumpsites, but that the drums are scattered over a wide area. In order to adequately map the distribution of the drums over the entire FIRWD, it is necessary to survey the entire area with the 30 kHz sidescan operated at a swath width of 1-km or less. We have verified our interpretation of the sonograph with an underwater camera system at one location. Camera surveys need to be done at other sites not only to verify the sidescan interpretation but also to establish the condition of the drums.
H. Chezar and J. Vaughan spent many hours maintaining the camera system that enabled us to verify the interpretation of the sidescan sonar images. N. Maher drafted many of the figures for this report. W. Danforth and T. O'Brien led the team that processed the sidescan sonar data. We appreciate the skill and professionalism of the officers and crew of the R/V Farnella.
Booth, J.S., Winter, W.J., Nieheisel, J., and Dyer, R.S., 1989, Geotechnical,
geological, and selected radionuclide retention characteristics of the radioactive
waste disposal site near the Farallon Islands: Marine Geotechnology, v.
8, p. 111-132.
Colombo, P. and Kendig, M.W., 1990, Analysis and evaluation of a radioactive waste package retrieved from the Farallon Islands 900-meter disposal site: EPA Report 520/1-90-014, 65 p.
Dyer, et al., 1976, Environmental surveys of two deep-sea radioactive waste disposal sites using submersibles: Proceedings of the International Symposium on the Management of Radioactive Wastes from the Nuclear Fuel Cycle, IAEA-Su/65. Vienna, International Atomic Energy Agency, p. 317-338.
Johnson, H.P. and Helferty, M., 1990, The geological interpretation of sidescan sonar: Review of Geophysics, v. 28, p. 357-380.
Karl, H.A., Drake, D.E., and Schwab, W.C., 1990, Preliminary cruise report for federal agency cooperative cruises F7-90-NC and F8-90-NC, Farallon Escarpment, 19 July to 3 August and 5 to 17 August, 1990: U.S. Geological Survey Administrative report, 8 p.
Maher, N.M., Karl, H.A., Chin, J.L., and Schwab, W.C., 1991, Station locations and grain-size analysis of surficial sediment samples collected on the continental shelf, Gulf of the Farallones during cruise F2-89-NC, January 1989: U.S. Geological Survey Open-file Report 91-375A, 4 p., 5 figs., 1 appendix; floppy disc of digital grain-size data OFR 91-375B.
Noshkin, V.E., Wong, K.M., Jokela, T.A., Eagle, R.J., and Brunk, J.L., 1978, Radionuclides in the marine environment near the Farallon Islands: Lawrence Livermore Laboratory, University of California Report No. UCRL-52381, 17 p.
Wright, A. St. C., 1990, Target search with deep towed sidescan sonars: Williamson & Associates technical report 3 p., 7 figs.