Satellite Detection and the Discovery of Bloody Marsh

Abstract

Assessment and mitigation of potentially polluting wrecks in deeper water is a more challenging endeavor due to limitations of divers and operating pumping equipment at depth, but additionally because locating these wrecks is harder. While these wrecks in deep water are more out of sight and out of mind, even when leaking, than those near shore, they are safer from damage by dredges, trawls or other anthropogenic activities. The wreck of the tanker SS Bloody Marsh, reportedly sunk off South Carolina on July 2, 1943 in 560 m of water, was considered a lower risk in NOAA’s PPW study due to its location in deep water, which the NOAA Screening Level Risk Assessment Packages typically considered as less risk: ‘deepwater shipwrecks tend to settle upright on the bottom, and is supported by the conclusions made by the U.S. Coast Guard in 1967 that oil will likely escape from a wreck’s vents and piping long before its hull plates corrode’ (NOAA, 2013b: 6). This, however, is a conclusion that needs to be reconsidered. Bloody Marsh was carrying a cargo of 106,496 barrels of bunker C heavy fuel oil, which is one of the largest cargoes among the vessels on the PPW list. In reviewing the 87 wrecks on the list, we selected Bloody Marsh as a high priority because of its large cargo and the fact that it was struck with two torpedoes and reportedly broke in half while sinking. Only ships with intact hulls would settle upright on the seabed, and therefore presumed Bloody Marsh did not.

9.1 Introduction

Assessment and mitigation of potentially polluting wrecks in deeper water is a more challenging endeavor due to limitations of divers and operating pumping equipment at depth, but additionally because locating these wrecks is harder. While wrecks in deep water are more out of sight and out of mind, even when leaking, than those near shore, they are safer from damage by dredges, trawls or other anthropogenic activities. The wreck of the tanker SS Bloody Marsh, reportedly sunk off South Carolina on July 2, 1943 in 560 m of water, was considered a lower risk in NOAA’s PPW study due to its location in deep water, which the NOAA Screening Level Risk Assessment Packages typically considered as less risk: ‘deepwater shipwrecks tend to settle upright on the bottom, and is supported by the conclusions made by the U.S. Coast Guard in 1967 that oil will likely escape from a wreck’s vents and piping long before its hull plates corrode’ (NOAA, 2013b: 6). This, however, is a conclusion that needs to be reconsidered. Bloody Marsh was carrying a cargo of 106,496 barrels of bunker C heavy fuel oil, which is one of the largest cargoes among the vessels on the PPW list. In reviewing the 87 wrecks on the list, we selected Bloody Marsh as a high priority because of its large cargo and the fact that it was struck with two torpedoes and reportedly broke in half while sinking. Only ships with intact hulls would settle upright on the seabed, and therefore presumed Bloody Marsh did not.

This observation comes from a Coast Guard report from 1967 that indicates that wrecks that settle upright on the seabed will lose their oil through vents, and that ships that sink in deep water tend to settle upright (NOAA, 2013b: 6). While this is often the case due to the streamlined design of a vessel’s hull, and hence a postulated hydrodynamic glide to the seabed, and some ships that have been sunk by torpedoes and lie in deep water have been found to settle upright, for example that of SS Coast Trader off Vancouver (Delgado et al., 2018), the assumption here is that vessels remained intact. Coast Trader was struck by a single torpedo and remained intact. Bloody Marsh was struck by two torpedoes and witness accounts indicated it broke in half on the surface. Therefore, when reviewing the list of PPW wrecks, we suspected that Bloody Marsh would be in sections on the seabed like Coimbra and Munger T. Ball (Brennan et al., 2023), and likely not upright. Another tanker wreck that illustrates this sinking pattern is that of RFA Darkdale, a British ship sunk of St. Helena island in the South Atlantic in 1941, which was broken into two halves, the bow inverted, and the stern section lying on its port side (Liddell & Skelhorn, 2012; Hill et al., Chap. 6 and Lawrence et al., Chap. 11, this volume).

We were able to locate Bloody Marsh at a depth of 465 m off South Carolina due to a fortunate survey of opportunity and that it was at a depth where hull-mounted multibeam sonar could detect it. This linked with reports of oil slicks on the surface nearby directed us to where to survey, which will be discussed in more detail herein. This wreck was also important to find due to its large amount of cargo (>106,000 barrels), and location near the Gulf Stream where a catastrophic leak could oil the shoreline from north Florida to Delaware (NOAA, 2013b: 31). However, due to the assumption about ships in deep water, the PPW study concluded: ‘this is more likely a shipwreck to be aware of and not one considered for in-water assessment… it is unlikely that significant amounts of oil remain on the wreck’ (2013b: 6).

Writing off wrecks in deep water due to the assumptions discussed above is premature, as the case of Bloody Marsh indicates. Numerous vessels were struck by multiple torpedoes and could lie in sections on the seabed, on their sides or inverted, positions in which oil would not escape through vents and piping. This example furthers the need for increased ocean exploration efforts to locate and assess potentially polluting wrecks and determine the exact state of the wreck site to progress past the desktop review initially put forth in the PPW study.

9.2 SS Bloody Marsh

SS Bloody Marsh was a T2-SE-A1 oil tanker built in Chester, Pennsylvania by the Sun Shipbuilding and Dry Dock Company in 1943 (Fig. 9.1). The T2 class tankers were a pre-fabricated commercial design that started production in 1940 by Sun Shipbuilding for the Standard Oil Company. These tankers were designed with standard dimensions and commonly available machinery to fast-track their construction. The T2-SE-A1 tanker was an ‘oceangoing, single screw oil tanker of about 500 feet waterline length that had turboelectric propulsion, carried less than twelve passengers, and had a deadweight of 16,000 tons (approximately 141,000 barrels)’ (Spyrou, 2006: 3). Following the attack on Pearl Harbor in December 1941, the U.S. Maritime Commission ordered additional tankers to supply the war effort and replace those being lost from U-boat attacks. A total of 536 T2 tankers were built of three types, and 525 were nearly identical T2-SE-A1 vessels. This basic type vessel was based on the standard ship designed and built by the private commercial oil company in response to the establishment of an emergency tanker program, which was in response to the signing of the Lend-Lease Act in March 1941 to provide aid to Britain (Sawyer & Mitchell, 1974). The hulls of these tankers were all welded, which was a relatively new method introduced in the late 1930s. Some of these tankers developed hull fractures due to the electric welds and quality of the mild steel used for the hulls in cold temperatures that made the crystalline structure of the metal brittle (Spyrou, 2006). Notable instances of these tankers breaking in half include Schenectady in January 1943 and SS Pendleton, which sank in a storm off Cape Cod in February 1952 and inspired the novel and film, The Finest Hours. Bloody Marsh was sunk in warm summer waters off South Carolina on its maiden voyage, so the structure of its hull did not have the chance to be tested.

Fig. 9.1

Bloody Marsh (U.S. National Archives)

Bloody Marsh was loaded in Houston with 106,496 barrels of Navy fuel with specific gravity 13, and departed for New York under Master Albert Barnes operating for the Cities Service Oil Company. The tanker was not completely full; reports indicate tanks 2–8 were full as were wing tanks 3–8, and wing tank #2 was approximately ¾ full. Not sailing with a convoy, the tanker sailed from Houston in a zigzag pattern through the Gulf of Mexico and around Florida, as was common anti-submarine defense procedure. About 75 min before the attack, the tanker stopped this pattern and ran a straight course (Burch, 1943). The first torpedo from U-66 struck the port quarter at the engine room, causing the ship to immediately begin to settle by the stern, and the ship sunk until the gun platform was ‘nearly awash’ when the second torpedo struck amidships at the #3 tank, breaking the ship in two. The stern sank almost immediately while the bow floated free for 3–5 min before sinking (Burch, 1943: 2–3). The torpedo indicator went off about 30 s prior to when the first torpedo struck. The ship turned hard to port and was swinging in that direction as the torpedo hit. The bridge issued a distress signal ‘SOS SSSS SOS BLOODY MARSH 31.33 N 78.55 W TORPEDOED’ two or three times following the first strike, which was the only transmission the tanker sent during its entire voyage (Burch, 1943: 6). The course steadied after the explosion and gradually slowed, then proceeded at about two knots until the ship was abandoned.

The crew consisted of 50 merchant crew and 28 armed military personnel. By the time the second torpedo struck, all surviving crew had abandoned the ship in lifeboats with the exception of the Commander of the Armed Guard Unit and three enlisted men at the stern gun, who joined the crew after the second explosion. The crew spotted the U-boat before the firing of the second torpedo when it surfaced and moved toward the ship and lifeboats. Two rounds were fired from the submarine’s deck gun at a 5 s interval directly after the release of the second torpedo, but it is not known if that was directed at the sinking tanker or the survivors (Burch, 1943: 4). Following the second torpedo, the submarine struck the stern of the No. 1 lifeboat, which raised it up from the water and knocked several survivors out, one of whom landed on the submarine before falling into the water. The submarine reportedly stayed in the area for 15–20 min but did not open fire on or attempt to capture any survivors. Survivors described the submarine based on their quick glimpses in the dark and a sketch was produced based on their observations (Fig. 9.2). A total of three crew were killed in the attack likely from the first explosion: third Assistant Engineer, Robert T. Winslow; Fireman, James B. Mitchell; and Oiler, Frank B. Robuck, all of whom were on duty in the engine room at the time of the attack (Moore, 1983). The 75 survivors were picked up by SC 1049 and brought to Charleston, South Carolina the following morning. This attack is compared to U-66’s sinking of SS Esso Gettysburg the month before off the east coast of Florida, a tanker that was traveling the same route as Bloody Marsh.

Fig. 9.2

Sketch of U-66 based on descriptions of survivors (U.S. National Archives)

9.3 Initial Expedition

In 2018, NOAA Ship Okeanos Explorer published its planned mapping expedition location off the southeastern US and the first author submitted Bloody Marsh as a potential mapping target. Previously, the Kongsberg EM302 multibeam sonar was run over the wrecks of Gulfoil and Gulfpenn in the Gulf of Mexico from exploration vessel (E/V) Nautilus. Each wreck is at about 500 m depth and this was a test of the detection capability of the hull-mounted sonar over large oil tanker wrecks; they were visible but only as small bumps on the landscape. However, this indicated that the wreck of Bloody Marsh could be detected with hull mounted sonar. Multibeam expert Gary Fabian reviewed the data collected from Okeanos Explorer and found a target approximately the right size for the tanker, however when the ROV dived on it, it was determined to be a natural rock outcrop (Brennan, 2019).

Following this result, the geoscience company CGG reached out to maritime archaeologists at SEARCH through a mutual colleague at the Bureau of Ocean Energy Management (BOEM). CGG had previously licensed BOEM with an offshore natural oil seepage study across the US East Coast, which included satellite-detected oil slicks around the documented site of the Bloody Marsh recorded over several decades. It was hoped this specific oil slick dataset could help refine the search area for future explorations for the wreck.

9.4 Satellite Detection of Sea Surface Oil Slicks

The detection of oil slicks from sunken shipwrecks, and the usage of those oil slicks in the subsequent locating of a vessel is not in of itself a new technique. Following the loss of the bulk carrier MV Derbyshire in 1980, oil slicks were reported by patrol aircraft of the Japanese Maritime Safety Agency a week after the vessel’s disappearance. These oil slick records, paired with ocean current models, allowed the successful discovery of the wreck location with side-scan sonar and an ROV dive in 1994 (Mearns, 1995).

The detection of the presence of oil slicks by satellite imagery is also not a new occurrence. Indeed, CGG were the first to extensively test new satellite systems in the early 1990s, partnering with the British National Space Centre (BNSC) and several oil companies at the time to test theories of oil slick detection. Since then, many new satellites have been launched, adding to the archive of satellite imagery capable of recording oil slicks. CGG has utilised this dataset to build a global database of oil slicks around the world.

The original BOEM natural oil seepage mapping study provided by CGG follows oil slick mapping and classification techniques first developed to document spatially repeating oil slicks across a satellite imagery stack over multiple imaging dates (Press & Lawrence, 1995). By observing a large imagery stack over a location across multiple dates, features can be observed that may be infrequent/ephemeral, whilst allowing for false positive features that may be misidentified on single imaging occurrences (e.g., oil slicks sourced from shipping) to be removed.

In the region of the proposed Bloody Marsh site, the study utilised 76 satellite images recorded from 1992 to 2019, of which 16 documented notable oil slicks (see Table 9.1 and Fig. 9.3). This imagery, a mixture of Synthetic Aperture Radar (SAR) and multispectral, was sourced and processed from the Open Access archives of the European Space Agency (ERS and Sentinel-1), Canadian Space Agency (Radarsat-1), Japanese Space Agency (ALOS-1) and the United States Geological Survey (Landsat-8).

Table 9.1 Slicks observed across satellite imagery

Fig. 9.3

Oil slicks mapped across 16 satellite images from 1992 to 2019, compared to the (imprecise) historically documented location of the Bloody Marsh wreck site. Whilst the majority of slicks originate from the southwest within a relatively constrained region with a radius of ~1.5 km, two notable outlier slicks are documented approximately 10 km away from this repeat location (imaged on 4th December 2008 and 24th May 2019). These outliers may reflect significantly different currents within the water column at the time of imaging or additional seabed oil sources

9.5 Oil Slick Evidence Towards Bloody Marsh Wreck Location

While only 20% of images used in the BOEM study documented oil slicks in the region, the reoccurrence of slicks across 27 years was highly indicative of a persistent fixed seafloor origin of the surface oil (see Fig. 9.3). Recent studies of other fixed point oil release sites around the world, at varying depths, carried out by CGG and others, resulted in the conclusion that these slicks showed strong spatial correlation to a single point source.

Typically, sea surface oil slicks fed from seabed sources (natural or anthropogenic) tend to develop curved ‘corkscrew’ morphologies as the overlying sea surface rotates due to near-inertial oscillation drift. This is due to most satellite scenes that observe oil slicks being acquired at low wind speeds, due to the requirement of relatively calm conditions for the slicks to form. However, the slicks observed in the Bloody Marsh region were observed to drift in the strong Gulf Stream currents off the coast of South Carolina. These currents, averaging 10.3 km/h, act to drift the surfacing oil slicks rapidly away from surfacing location, leading to long linear slicks ranging from approximately 5–110 km. Crucially, the surfacing location of these slicks (origination point) where observed in the same location and the long drifting slicks showed no sign of fragmentation before evaporation. This strongly suggested that the release of oil from the wreck was likely continuous, and those images in CGG’s archive that did not record an oil slick where likely to be due to other mitigating met-ocean conditions. When utilising the multispectral data in particular, that images oil directly, thicker components of the oil slick could be observed at the southwestern end of the slick, in keeping with the theory of a subsurface release, before thinning and drifting in the currents (Fig. 9.4).

Fig. 9.4

Contrast enhanced Landsat-8 image of a sea surface oil slick (linear feature running southwest to northeast) acquired on 27th June 2014. The origin point of the oil slick can be determined as the southwestern end by the presence of thicker oil (silver/metallic appearance) which disperses and thins out into a sheen as currents move the oil slick away from the source point. Landsat-8 image courtesy of the U.S. Geological Survey

Importantly, the site of the recurring oil slicks mapped by CGG was 12 km southwest of the historically documented location of SS Bloody Marsh and previous surveys. Following on from these observations, 3D models of oil plumes in the water column were created based on the prevailing currents in the region. These allowed backtracking of the oil source from the sea surface slick observation locations towards a clear target area on the seabed.

9.6 Discovery

An additional survey request was made to NOAA, and a target was detected only a few miles from the downstream limit of the oil slicks. The target measured shorter than the length of the tanker, and was a single target, where the reports of Bloody Marsh’s sinking have it breaking in half at the surface. However, the target was located in a relatively flat area of seabed two miles directly downstream of the oil slick origins on the surface, so was a promising target.

On October 28, 2021, ROVs Deep Discoverer and Seirios were launched from Okeanos Explorer to locate and identify the multibeam sonar target. Maritime archaeologists participated in the dive in real-time through telepresence capability on board the ship through NOAA’s Ocean Explorer website and directed the dive remotely (Brennan et al., 2018). The ROV approached the target at 465 m depth and first came upon a debris field that was the remains of the destroyed stern and engine room. Immediately adjacent to this was the prominent hulk of the oil tanker completely inverted on the seabed (Fig. 9.5). A small amount of hull remained from the engine room, and a bulkhead between that space and the aft tanks appeared intact. Corroded holes were visible in the hull, but only those over machinery spaces. Like the Coimbra wreck, corrosion was more visible in areas of hull that had seawater on both sides; areas of hull that appeared to contain intact tanks had less corrosion (Brennan et al., 2023). Moving southeast forward along the wreck, the hull plate joinery indicated welded seams, which, along with the location, helped to positively identify the wreck as Bloody Marsh.

Fig. 9.5

ROV image of the broken stern end of Bloody Marsh at the joinery between the engine room and cargo tanks (NOAA Ocean Exploration)

The main hulk of the wreck had few visible features, as the ship was completely turtled and the superstructure and decks buried in sediment. Biologists viewing the stream live commented that the wreck was colonised by what appeared to be a single species of gorgonian coral all of a similar size, indicating a single colonisation event. The ROV documented the starboard side of the wreck, but was unable to move around to the east-facing port side due to a strong current, therefore if any superstructure remains or debris lie on the seabed on that side, it was unable to be seen during this dive. When the ROV reached the forward end of the wreck, the measurement of the hull section indicated approximately 300 feet of hull. Subtracting the destroyed engine room, this would suggest that another 100 feet or so of hull remains. However, the bow section was not visible beyond the wreck and the current prevented further exploration in the area. The break in the hull, based on this measurement, appears to be at Tank #3, which is consistent with historic accounts. Therefore, the bow section could contain intact tanks #1 and 2 as well the as Deep and Fore Peak tanks forward. Future exploration in this area would be needed to locate and document this section.

9.7 Discussion and Conclusion

We had initially suspected Bloody Marsh could be a greater pollution risk than the PPW assessment projected due to its large cargo and the incident of its sinking. Had the vessel settled upright on the seabed, as the PPW report suggests is likely for wrecks in deep water, oil cargo would have escaped through the vents over time (NOAA, 2013a, b). This was the case for Coast Trader off the northwest Pacific coast (Delgado et al., 2018), as well as tanker wrecks in the Gulf of Mexico including Virginia, GulfOil and GulfPenn (Church & Warren, 2008), all of which sank intact in one piece. However, other oil tanker wrecks, such as Coimbra and Munger T. Ball, illustrate that ships sunk by torpedoes that break into multiple sections do not settle evenly on the seabed (Brennan et al., 2023). While those wrecks lie in shallower water, this proves to be the case for Bloody Marsh as well, and potentially indicates that other tankers sunk in deep water that remain undiscovered should also be considered high pollution potential as they may not have righted as they sank, especially if broken. This case study also serves as a proof of concept for utilising surface oil slicks as indicators of leaking oil tankers if oceanographic modeling can assist in back-calculating the wrecks’ locations. As many of the PPW wrecks that remain to be discovered are in deeper water, this presents a promising method for locating other sites, especially those with cargos that are in danger of releasing.

The discovery of Bloody Marsh illustrates the complexity of locating shipwrecks in deep water. The size of the tanker and the depth at which it sank allowed for multibeam sonar to detect it once the satellite data redirected the search to a new area. This work shows the importance of technological advancements and data review to detect leaking wrecks and its potential to locate undiscovered sites. Additional survey in the area of Bloody Marsh is, however, still needed, as the 100-foot bow section remains missing. The second parallel slick identified in the satellite data could be this section, which is large enough to have some of the cargo tanks intact and still containing cargo. Further survey may locate this section potentially downstream of that second slick. The overturned hull of the tanker appeared stable, suggesting large amounts of oil could remain inside, and this site could be one to be considered for ROV-based mitigation in the near future.