Assessing the environmental risk posed by a legacy tanker wreck: a case study of the RFA War Mehtar

The ocean floor is littered with thousands of wrecks containing hazardous fuels, lubes, armaments and cargoes that have polluted, are polluting, or will pollute the marine environment. The UK Ministry of Defence manages the environmental risk associated with its ∼5,700 military wrecks through the Wreck Management Programme. This paper explores the methods used to assess the environmental risk associated with the wreck of the Royal Fleet Auxiliary (RFA) tanker RFA War Mehtar, which lies at approximately 39 m depth, 15 nm east of Great Yarmouth, United Kingdom. The methods employed include archival research, computational oil spill modelling combined with sensitivity mapping, environmental sampling and hydrocarbon concentration analysis, multibeam echosounder sonar surveys and neutron backscatter probe measurements. Each method is described, and its efficacy discussed. The archival research was essential to determine what fuel and cargo the RFA War Mehtar sank with; but only when high-resolution multibeam sonar images were combined with the ship’s plans and neutron backscatter measurements from individual tanks could one confidently conclude that the wreck does not pose a significant pollution risk.


Introduction
Around 8,500 potentially polluting wrecks lie across the sea floor, most of which sank during World War II, including ∼1,500 tankers (Michel et al 2006). The rate at which these wrecks corrode is affected by environmental variables including seawater temperature, salinity and dissolved oxygen concentration, and wreck topography, depth and epifauna (MacLeod 2016). Corrosion rates are likely to increase as the seas warm and acidify in response to increasing atmospheric carbon dioxide concentrations, making the proactive management of oil spill risks ever more urgent.
Potentially polluting wrecks are a global environmental issue, but national wreck management programmes are at varying levels of maturity (e.g. Idaas 1995, NOAA 2013. The UK Ministry of Defence (MOD) first started actively managing the leaking wreck of the battleship HMS Royal Oak in the 1990s through its Salvage and Marine Operations (SALMO) team and this led to the inception of its Wreck Management Programme in 2009, which began with the prioritisation of the MOD's entire post-1870 shipwreck inventory. The Wreck Management Programme proactively seeks wrecks that still contain significant quantities of oil, using a number of tools and techniques, so that remaining oil can be removed in a controlled operation.
Archival research is a useful starting point to determine what fuel and cargo a wreck might have sunk with, how much damage was caused at the time of sinking, and where the wreck now lies (Liddell and Skelhorn 2018). Multibeam sonar is typically used to determine the condition of the wreck and a comparison of point clouds can highlight temporal change that might indicate the wreck is collapsing. Multibeam sonar, as well as observation by diver or remotely operated vehicle (ROV), can also be used to verify the identity of a wreck. ROV video and sonar surveys can also be used to observe the condition of a wreck. Computational oil spill modelling is routinely used to predict the fate of potential oil spills, and combined with geographical sensitivity data to predict potential environmental and ecological impacts  capable of measuring 240 depths per ping, up to 50 Hz, with 0.5°across track and 1°along-track beam width, with a maximum 120°swath angle. The multibeam array could be lowered down to the wreck allowing the collection of high-resolution imagery at a fraction of the cost of an ROV. Twenty-three survey lines were run over the wreck at a towing depth of 25 m.

Historical Desk-based Assessment
Archival research was used to complete an Historical Desk-Based Assessment (H-DBA) of RFA War Mehtar in 2016. Books, websites and archived documents, including Lloyd's Register ship's plans and survivor's reports, were used to build up a picture of the wreck's history and current condition, how much oil she contained when she sank, and how much could remain. Unusually in this instance, multibeam sonar imagery and an associated 3D model already existed from the 2014 multibeam sonar survey which indicated which tanks could remain intact. The H-DBA is usually completed before any on-site surveys.

Environmental Desk-based Assessment
An Environmental Desk-based Assessment (E-DBA) of the RFA War Mehtar was completed according to a method devised for wrecks (Goodsir et al 2019). The E-DBA consists of three parts: (1) an assessment of the likelihood of the wreck releasing oil; (2) oil spill modelling to predict the spatial extent of potential oil spills; and (3) a risk assessment based on the overlap between the oil spill model outputs and the ecological and socioeconomic sensitivities in the area.

Likelihood of an oil release
The likelihood of the wreck leaking oil was assessed according to eight pre-defined criteria of different weighting (table 1) on a scale of 1 (low) to 3 (high) risk, using the information compiled in the H-DBA (Wessex Archaeology 2016). The assessment was given a confidence rating depending on the availability and reliability of information.

Oil spill modelling
A chronic (slow, continuous) and two acute (large volume released over 24 h) oil spill scenarios were modelled using the Marine Environmental Modelling Workbench (MEMW) suite (Aamo et al 1997, Reed andRye 2011). The chronic release of 50 kg/day was modelled using the Dose-related Risk and Effects Assessment Model (DREAM). Worst-case (total oil volume: 7,000 tonnes) and most-probable case (one tank: 1,000 tonnes) scenarios were modelled using the Oil Spill Contingency and Response (OSCAR) component of MEMW. The model was configured and meteorologically forced as specified in Cefas (2017a).
RFA War Mehtar was carrying Admiralty Fuel Oil which was the generic name given to fuel oil at the time and covered a range of oils with different (generally unknown) chemical and physical properties. The modern IF-30 fuel oil was used as a proxy for Admiralty Fuel Oil in the oil spill model. IF-30 is a refinery mixture product consisting of approximately 35% gas oils and 65% Bunker C (Environment and Climate Change Canada 2017). It was adjusted in the model to 25% Polycyclic Aromatic Hydrocarbon (PAH) content to match the composition of modern marine diesel, to avoid underestimating oil spill toxicity.
Predicted oil toxicity and level of contamination on the shoreline, water surface, and in the water column and sediment were output from the model and mapped in Esri ArcMap to visualise the potential spatial extent of Table 1. Criteria used to predict the likelihood of a wreck releasing oil with weighting in parentheses. Adapted from Goodsir et al (2019).

Criterion
Reasoning Vessel depth (2) Shallow wrecks (<30 m) will be more impacted by waves and storms which may increase rate of deterioration compared to deep wrecks in relatively benign conditions. Oil leak history (3) Reports of oil leaks indicate a deteriorating wreck. Wreck integrity (2) Intact wrecks are generally more of a concern, although broken up wrecks can still pose a risk if their tanks are intact. Vessel age when sunk (1) Assumption that older vessels would have been less structurally sound than newer vessels due to operational wear and tear and/or modifications and could therefore deteriorate quicker. Time submerged (3) Extent of corrosion assumed to increase with time submerged. Storage method (2) Bunker and cargo tank bulkheads are thicker and therefore more resistant to corrosion than drums stored on deck. Incident type (1) The more severe the incident, the more damage caused and therefore the more oil released at the time of sinking. Seabed stability (2) An unstable seabed can put strain on a wreck causing fractures.
pollution, and therefore assess the risk to local environmental and socioeconomic sensitivities. Thresholds were applied to the model outputs according to table 2.

Risk assessment
The potential oil spill impact on six ecological sensitivities (coastal and marine protected areas, cetaceans, seals, birds, fish nursery and spawning grounds, benthic communities and habitats) and four socioeconomic sensitivities (infrastructure, tourism and recreation areas, fishing activities, shipping) was assessed based on the level of oil pollution reaching these sensitivities predicted from the oil spill model outputs. The impact could be seasonal if it relied on seasonal metocean patterns and/or seasonal breeding patterns of birds, for example. The impact on each sensitivity was scored as low (1), medium (2) or high (3).
The risk from each oil spill scenario was determined separately for ecological and socioeconomic sensitivities from the equation: =Ŕisk Likelihood of oil release Impact

Environmental on-site survey (2017)
An on-site survey of the RFA War Mehtar was completed between 27th June and 1st July 2017 following a protocol developed for wrecks (Cefas 2017b). The full report (Cefas 2017c) is being published separately so only a summary is provided herein. The survey had three objectives: (1) verify wreck position and integrity, (2) characterise habitat, and (3) measure hydrocarbon contamination in the surrounding seawater and sediment.

Wreck position and integrity
A multibeam sonar survey was done to determine whether the wreck had changed since the 2014 survey. Two systems were used: Kongsberg EM2040 on the drop keel and EdgeTech 6205 mounted in the Moonpool of the vessel. Following calibration, the wreck was surveyed from four different directions along thirteen survey lines spaced at 40 m.

Habitat characterisation
Sediment samples were collected from around the wreck to validate the seabed type derived from the multibeam bathymetry data and to determine local infaunal ecology. A 0.1 m 2 ('mini') Hamon grab collected 28 samples for analysis for sediment grain size and distribution. A 500 ml sub-sample was stored at −20°C and later processed for particle size distribution according to Mason (2016).
The faunal fraction was collected on a 1 mm mesh, photographed, then fixed in buffered 4% formaldehyde. Fauna were identified to the lowest taxonomic level possible, counted and weighed (blotted wet weight) to the nearest 0.1 mg. Abundance and biomass data were analysed in PRIMER v6 (Clarke and Gorley 2006).

Hydrocarbon contamination
Sediment and seawater samples were collected for total hydrocarbon content (THC) and PAH content analysis.
Sediment samples were collected along transects heading north (5), east (5), south (6) and west (5) from the wreck using a Shipek grab that was hosed down and rinsed with pentane between samples to remove traces of oil.
Shipek grabs collect the top ∼10 cm of sediment. Surface samples were transferred to 500 ml glass Beatson jars and frozen at −20°C. After thawing, wet sediment samples were homogenised, extracted using alkaline saponification, filtered and extracted into pentane.
Seawater samples were collected from 2-5 m above the seafloor, and from depths where elevated fluorescence was observed, using pentane-rinsed Go-Flo sampling bottles on a CTD rosette. Roughly 1.5 l of each sample was dispensed into 2.5 l Winchester bottles, topped up with 50 ml pentane and stored at 5°C-10°C. Samples were filtered using a 2 l separating funnel and extracted into pentane.
The pentane extracts (from both sediment and seawater samples) were reduced to 1 ml using a rotary film evaporator and passed through an alumina chromatography column to remove residual polar compounds, re- Table 2. Thresholds applied to oil spill model outputs (Cefas 2017a).

Location
Threshold applied Sea surface 0.1 g m −2 Water column 50 ppb total concentration Sediment 1.7 g m −2 Shoreline 1 g m −2 concentrated to 1 ml and sealed in a glass autosampler vial. Concentrates were analysed for a suite of alkylated and parent PAHs by coupled gas chromatography-mass spectrometry as detailed in Kelly et al (2000).

Wreck integrity ROV video and sonar survey (2017)
The condition of RFA War Mehtar was inspected using a Sub Atlantic Comanche ROV during 17th August to 1st September 2017. Live feeds of video and Teledyne BlueView P900-130 2D forward looking imaging sonar (900 kHz) were used to direct the ROV to perform a systematic inspection of the wreck's hull, looking for evidence of damage and oil leaks.
2.6. Wreck integrity ROV multibeam sonar survey and neutron backscatter (2019) A further multibeam sonar survey of the RFA War Mehtar was done in June 2019; SALMO were planning for a potential oil removal operation so needed a high-resolution model of the wreck and the location and volume of remaining oil. The objectives of this survey were (1) collect high-resolution sonar data to compare with previously collected data looking for evidence of deterioration, (2) overlay general assembly plans on the 3D model to identify which tanks are breached, and (3) deploy a neutron backscatter probe to determine the contents of apparently intact tanks.
2.6.1. Multibeam sonar survey A dual head Reson 8125-H high-resolution multibeam sonar system mounted to a Schilling Robotics UHD 200HP work-class ROV with tether management system collected data to produce a 3D model at the resolution required for planning purposes (CWAVES 2019a). Three lines were surveyed across the starboard side of the wreck to establish her general condition. This was followed by a detailed survey to identify the interface between the wreck structure and the sediment. Fill-in lines were then run to ensure sufficient coverage. A High Precision Acoustic Positioning (HiPAP) Ultra-short Baseline (USBL) system tracked and logged the subsea position of the ROV.

Neutron backscatter
The non-intrusive neutron backscatter technique was chosen over the more traditional method of drilling into tanks to look for oil due to concerns that the pressure from drilling might have caused an uncontrolled release of oil from the old and deteriorating wreck. The neutron backscatter probe contains a source of high energy neutrons and is sensitive to 'slow' neutrons that are returned (CWAVES 2019a).
Areas of hull at pre-defined locations (at the top, middle and bottom of each tank) were cleared of marine growth using wire brushes attached to the ROV. The ROV then held the neutron backscatter probe against the cleaned hull to measure the number of neutrons that were reflected by hydrogen atoms within each tank. The higher the number of reflected neutrons the higher the hydrogen content of the liquid on the other side, so it was expected that a higher number of neutrons would be reflected by hydrocarbons compared to seawater. Measurements taken from tanks known to be breached, and therefore containing seawater, were used as controls to calibrate the instrument and to compare with measurements from seemingly intact tanks.

Results
The various methods used provided varying degrees of certainty in what they were aiming to measure. The different methods and how well they answered questions that helped to assess the environmental risk posed by the RFA War Mehtar are summarised in table 3. The results of each method are presented and discussed in turn.

High-resolution multibeam sonar survey (2014)
The 3D model produced from the sonar imagery collected in 2014 (figure 2) allowed positive identification of the wreck when compared with photographs of the RFA War Mehtar taken in service and the general assembly plans of her sister ship. She is broken in two at mid-ships with a gap of ∼3 m between the two sections. There is complex scouring around the wreck indicative of strong tidal flow in two directions. The wreck lies on her port side so half of her tanks are partly buried. The forward section appears to be in good condition with little obvious damage. The after section is damaged, with obvious collapse of hull plating and large fractures in some areas. The imagery indicates that some, if not all, of oil tanks 4, 5, 6 and 7 may have ruptured.

Historical Desk-based Assessment
The RFA War Mehtar, commissioned in 1920, is a Z class freighting tanker with seven longitudinally divided oil cargo tanks that can hold up to 7,000 tonnes of oil between them. She was torpedoed and later sank while under tow 15 nm east of Great Yarmouth, UK (figure 1) on 20th November 1941. The ship was fully loaded with 7,000 tonnes of Admiralty Fuel Oil when she sank. The captain noted that, although fuel leaked into the engine and oiler rooms after the torpedo attack, he did not see it escape into the sea. However, the engine room crew were 'black with oil' when they were rescued. There are contradictory reports on the scale of the fire at the time of the attack.

Likelihood of an oil release
The weighted likelihood of oil release was calculated as 32 out of a range of 15-45, giving it an overall likelihood of 'medium' (table 4). By 21st November 2031, when the wreck has been submerged for 90 years, the likelihood will be reclassified as 'high'.

Oil spill modelling
The 50 kg/day chronic oil spill scenario was predicted to have a localised impact stretching 5 km east-west and 30 km north-south across the wreck, posing a risk to 1%-5% of the most sensitive pelagic species near the wreck. The level of pollutants in the water column reached an equilibrium between the input and removal of hydrocarbon within 30 days. Most of the oil was predicted to partition into the sediment rather than form surface slicks. The overall predicted risk, calculated as the Predicted Environmental Concentration/Predicted No Effect Concentration (PEC/PNEC) was 1 (medium) in the immediate vicinity of the wreck and 0.2 (low) in the wider area.
The predicted fate of oil 30 days after the acute oil releases is presented as the average of multiple trajectories in table 5. The most-probable case acute oil spill scenario of 1,000 tonnes released over 24 h could impact around 2,650 km 2 of sea surface, with up to 628 tonnes reaching the east coast of England. The worst-case acute oil spill scenario of 7,000 tonnes released over 24 h could impact around 4,600 km 2 of sea surface, with up to 4,240 tonnes reaching the shoreline, predominantly on the east coast of England but also potentially the Netherlands.

Risk assessment
The chronic oil spill scenario could pollute two designated areas (the Haisborough, Hammond and Winterton Special Area of Conservation and the Outer Thames Estuary Special Protected Area) and could have an impact on fish spawning and nursery areas in winter and spring months. The overall environmental impact score was 9 out of a possible range of 6-18, and the socioeconomic impact score was 4 out of a possible range of 4-12. When multiplied by the likelihood of release score (32) the chronic oil spill scenario poses a medium risk to ecological sensitivities (288 out of a range of 90-810) and a low risk to socioeconomic sensitivities (128 out of a range of 60-540).
Both acute oil spill scenarios could cause significant pollution to three marine protected areas: the Haisborough, Hammond and Winterton Special Area of Conservation, the North Norfolk Sandbanks and Saturn Reef Special Area of Conservation and the Outer Thames Estuary Special Protected Area. They could have a high impact on cetaceans, seabirds and the benthic environment, and a medium impact on seals and fish spawning and nursery areas. Both acute oil spill scenarios scored an overall ecological impact of 16 (possible range 6-18). Both scenarios had a high impact on offshore infrastructure and shipping, and a medium impact on tourism and commercial fishing, scoring an overall economic impact of 8.17 and 8.42 (possible range 4-12) for the most probable and worst-case scenario, respectively. When multiplied by the likelihood of release score (32) the acute oil spill scenarios pose a high risk to ecological sensitivities (480 out of a range of 90-810) and a high risk to socioeconomic sensitivities, scoring 261 and 269 (out of a range of 60-540) for the most probable and worst-case scenario, respectively.

Wreck position and integrity
The multibeam sonar confirmed the position and condition of the wreck as seen in the 2014 survey; no noticeable change had occurred. It also presented further evidence of sediment scour at the wreck's bow and stern that had cut across megaripples in the sediment. It is unclear from the images whether the scouring had relocated or deepened between surveys. Total weighted risk score 32

Habitat characterisation
The dominant sediment type in the area is sand (0.125-1.0 mm) with little in the way of very fine or coarse sediment. The extensive sand waves and megaripples indicate highly mobile sediments. The sediment in the scours was gravelly muddy sand. Stations to the west had the lowest diversity and abundance; stations along the north-south transect had the highest diversity and abundance. The first station to the north of the wreck had the highest biomass and joint highest number of taxa.

Hydrocarbon contamination
The highest sediment total hydrocarbon content measured was 25.5 mg kg −1 ; most samples were <10 mg kg −1 . The highest concentration of PAH measured was 404 μg kg −1 ; most samples were in the range 10-82 μg kg −1 . The maximum percentage of PAH was 8% of THC. Sediment hydrocarbon concentrations were typical for the North Sea, exhibiting no noticeable elevation.
The highest seawater total hydrocarbon content measured was 7.85 μg l −1 ; most samples were in the range 0.004-0.027 μg l −1 . The highest concentration of PAH measured was 0.100 μg l −1 ; most samples were in the range 0.004-0.027 μg l −1 . The maximum percentage of PAH concentration was 3% of THC. All the measured PAHs were below the annual average environmental quality standard (EQS) limits from the EU Water Framework Directive except for fluoranthene at one station measuring 0.016 μg (limit 0.0063 μg l −1 ).

Wreck integrity ROV video and sonar survey (2017)
Visibility was poor but numerous large holes were detected in the tank plating and there were no oil droplets escaping from them. The ROV did not have tracking, however, so it was not possible to determine exactly which tanks had breached and where.
3.6. Wreck integrity ROV multibeam sonar survey and neutron backscatter (2019) 3.6.1. Multibeam sonar survey The direct comparison of the 3D model produced from the 2019 survey with that produced in 2014 showed no major change in the overall wreck structure, but one of the masts had collapsed (figure 3). The biggest change was in the region of the break between the two sections, which is not an area likely to release oil.
Overlaying the general assembly plans on the 3D model showed which tanks were damaged (figure 4). Cracks and holes were visible in nine of the fourteen tanks, and the oil fuel bunker had completely collapsed.

Neutron backscatter
The calibration readings from tanks known to be breached gave radiation counts ranging from 40,000 to 45,000 counts/3 s. Therefore, a tank containing oil was expected to give a reading of around 47,200 to 53,100 counts/ 3 s. Most readings were below this range but one tank gave a reading of 47,761 counts/3 s. Because it was so close to the threshold, the cause was investigated and it was noticed that a large section of corrosion had fallen away from the area, leaving bare steel. Another calibration reading was taken from a breached tank nearby and measured 44,195 counts/3 s due to the thinner plating. It was concluded that the previous high reading did not indicate hydrocarbon and, therefore, there was no evidence of hydrocarbons in any of the tanks.

High-resolution multibeam sonar survey (2014)
This survey positively identified the wreck and gave a sense of her condition. Some tanks appeared intact, but they may have had internal ruptures and therefore not contain oil. This survey did not provide sufficient information to decide whether oil needs to be removed. It is worth doing multibeam sonar surveys of wrecks to check their location and identity and get an idea of their condition, because some wrecks are clearly too damaged to hold significant quantities of oil, but it is cheaper to do this using ship-mounted multibeam, where wrecks are sufficiently shallow to yield good results, as part of a multi-wreck survey, which is the approach currently being used by the Wreck Management Programme.

Historical Desk-based Assessment
One of the most important pieces of information from the H-DBA is the type of oil in a wreck as this affects its environmental risk, with lighter, more volatile, oils having a more localised and shorter-term impact than heavy fuel oils (ITOPF 2014). Some tanker wrecks were in ballast when they sank, rendering them relatively harmless. It is important to ascertain oil type and volume, or at least make an educated guess, at the H-DBA stage before spending money on surveys.
Hydrocarbons were a relatively new source of fuel at the beginning of the First World War and, even by WWII when the RFA War Mehtar sank, they were not refined or chemically or physically characterised in the way they are today and were often listed simply as 'Admiralty Fuel Oil'. Consequently, comprehensive oil data is lacking from the era, but some generalisations can be made based on where and when the oil was collected and how it was named on the cargo inventory (Wessex Archaeology 2018).
Historical records are generally publicly available and ships' plans, if not publicly available, can usually be purchased inexpensively. The RFA War Mehtar H-DBA provided enough evidence to highlight it as a high priority wreck, which unlocked funding to do further investigation. Thus, the H-DBA is a cost-effective means of gathering evidence to prioritise wrecks for further investigation and remains the initial assessment of MODowned wrecks.

Environmental Desk-based Assessment
The likelihood of oil release scoring system, though useful, could be improved. For example, the assessment considers age at time of sinking to be a factor but does not account for the variation in build quality. To keep up with demand during WWII, ship builders used lower quality steel and transitioned from rivet to welded construction, with some of the steels and early welds failing (Baxter 1955, Shepheard 1957. Furthermore, there is no mention of local meteorological conditions that could lead to a wreck leaking, such as the typhoon that caused the USS Mississinewa to release oil in Ulithi Lagoon, Federated States of Micronesia (Gilbert et al 2003). However, this assessment is probably the most useful aspect of the E-DBA and it has now been incorporated into the H-DBA as this provides most of the data for it.
Oil spill modelling has major limitations for legacy wrecks. Firstly, the quantity of oil remaining in the wreck is unpredictable without on-site assessment, and subsequent neutron backscatter assessment showed that the RFA War Mehtar did not contain oil, thus oil spill modelling was ultimately unnecessary for this wreck. Secondly, the oil properties are unknown and existing oil spill models do not contain oils representative of Admiralty Fuel Oil, which is not a standardised oil anyway, so other oils from the model's database are used as a proxy. For the RFA War Mehtar oil spill modelling, IF-30 fuel oil was adjusted with 25% PAH content to mimic the oil's assumed properties; the highest concentration measured in the environmental on-site survey was 8% PAH in a sediment sample, which could mean that the 25% PAH concentration modelled is too high or that the PAH component of the oil has already undergone weathering. The oil within a wreck could have biodegraded to some extent which would also affect its chemical, and therefore physical, properties (Rabus 2005), making it even more difficult to model. Incidentally, microbial activity in oil can increase corrosion of steel initially, but offer protection once this has stabilised (Crolet 2005). Admiralty Fuel Oils are likely to be Type 3 oil and therefore bound to persist in the marine environment (ITOPF 2014). Using Geographic Information System (GIS) software to simply consider estimated oil volume and proximity to shore and local vulnerabilities might be as good a measure of risk as oil spill modelling based on assumptions, saving budget for on-site investigations.
The risk assessment section of the E-DBA makes no mention of physical and mental health impacts associated with oil spills, the extent of which will depend on the extent of oiling and the ability or inclination of local people to avoid it. Further, it is a standardised assessment designed to be applied globally, but the vulnerability and experiences of people who could be impacted by wrecks is not globally uniform (Lövbrand et al 2015, Fadigas 2017. It gives a reductionist view of socioeconomic impact; more effort must be made to consider vulnerability of local communities who may lack access to non-marine food or income, oil spill response resources or financial support. For example, wrecks in remote locations should immediately be highlighted as a priority for on-site investigation because a major oil spill cannot be counteracted efficiently.
The E-DBA is not deemed necessary for all wrecks but an improved method may be used to assess wrecks that pose a major risk based on oil volume and location if further evidence is required to secure funding for oil removal.

Environmental survey (2017)
Sediment samples were only taken from the top 10 cm for hydrocarbon analysis. Rogowska et al (2010) measured hydrocarbon pollution in sediments below the surface layer and as deep as 230-240 cm around the wreck of the SS Stuttgart. Given the dynamic conditions at the RFA War Mehtar wreck site it seems likely that deposited oil could get repeatedly covered and uncovered by sediment, which could cause sporadic oil slicks. A core sampling regime would have been more appropriate in determining the full extent of sediment contamination around the wreck.
Seawater and sediment samples containing at or below background hydrocarbon concentrations could be caused by several scenarios. For example, the wreck still holds all its oil cargo, or the oil was lost years ago and the environment has recovered or sediment transport has buried deposited oil, or oil is being released so slowly it does not accumulate in the sediment, or oil was released but got carried away by wind and currents and deposited elsewhere. Conversely, if high hydrocarbon concentrations were measured in the sediment one still would not know how much oil remained in the wreck. Similarly, hydrocarbons in seawater samples only tell us if the wreck is currently leaking oil, since seawater contamination quickly returns to background concentrations (e.g. Boehm et al 2007), but little of the rate of release and nothing of the remaining contents.
From the perspective of the Wreck Management Programme there is little benefit in measurements that give ambiguous results; a reliable assessment of how much oil remains in each wreck is essential for the prioritisation of wrecks for oil removal so this method will not be used again to assess risk. However, the use of core sampling is now being trialled to collect samples from deeper in the sediments to determine whether this technique will provide more useful information on historic oil spills and an alternative explanation for recurring oil sheens.

Wreck integrity ROV video and sonar survey (2017)
It was clear from the ROV video and sonar survey that the wreck had many large holes in her hull, and no oil was leaking from them. Based on this survey the RFA War Mehtar was deemed to pose no major risk, as the bulk of her cargo oil had likely already escaped. However, SALMO were contacted by the Maritime and Coastguard Agency (MCA) in the summer of 2018 regarding some patches of oil sheen that may have come from the wreck. The MCA did regular aerial surveillance of the oil, a thin sheen heading north-eastwards, and collected samples from the slick. Gas chromatography with flame ionisation detection (GC-IFD) and gas chromatography mass spectrometry (GC-MS) analysis showed the oil shared similarities with oil removed from the RFA Darkdale (Fugro 2018), which was lost the same year, suggesting the oil could have come from the RFA War Mehtar.
This release of oil prompted the final ROV multibeam sonar and neutron backscatter survey. So, although the ROV survey had satisfied SALMO that the wreck was full of holes and therefore posed as low a risk as reasonably practicable, when confronted by the MCA there was not sufficient evidence to convince them that the wreck would not leak a more significant amount of oil. Thus, one might conclude that this ROV survey did not give value for money and thus such surveys are now rarely used except as part of a multi-wreck ship-mounted multibeam sonar survey to help positively identify a wreck.

Wreck integrity ROV multibeam sonar survey and neutron backscatter (2019)
This survey provided higher resolution imagery than the 2014 survey so the damage to tanks could be seen more clearly. But it was the data from the neutron backscatter probe showing the tanks were empty of oil that proved an oil removal operation was not required, saving the MOD an estimated £10million. The US spent $3.5million determining that the wreck of the tanker SS Montebello did not require further intervention (Stout and Rubini 2014), so this considerably cheaper survey was good value for money. Non-intrusive methods respect the cultural heritage and avoid an uncontrolled release of oil. However, neutron backscatter readings should ideally be verified by drilling into the tanks.
Neutron backscatter was successful on the RFA War Mehtar because she is a tanker with a simple tank layout and only 14 tanks, all of which are accessible due to her position on her portside. Large tanks reduce the likelihood of taking a reading at a point where there is a stiffener or pipe behind the plate, which would skew the result, and multiple readings can be taken from each tank to reduce error. She also has a single skinned hull, which was typical for tankers at the time, but not necessarily for the fuel tanks of warships (e.g. Hill 2019).
This was the only piece of work that gave a definitive answer to the question of tank contents and will be replicated on similar wrecks highlighted by an H-DBA to potentially contain oil and an initial ship-mounted multibeam sonar survey showing the wreck is intact.

Conclusion
A suite of techniques has been applied to assess the potential environmental risk posed by the RFA War Mehtar, each of which produced useful data to inform how to manage the wreck and others like her. As more work was done on the wreck, we reflected on the utility of previously collected data with the benefit of hindsight. The archival research of the H-DBA was essential to determine the quantity of fuel the RFA War Mehtar was carrying when she sank, and whether she was likely to still contain that oil based on the circumstances of her sinking. The E-DBA provided a good visualisation of potential pollution risk that can help with funding bids for preventative work but did not contribute much to the management of the wreck. Similarly, the first three on-site surveys gave ambiguous results rather than showing whether the wreck contained oil.
The conclusive evidence came from the high-resolution multibeam sonar images overlaid with ship's plans and neutron backscatter measurements. Although the costliest element, it was only at this point that one could confidently conclude that the wreck does not pose a significant pollution risk. Therefore, for similar tanker wrecks it is recommended to move from initial archival research to rapid assessment, preferably by shipmounted multibeam sonar, to determine general condition. If this shows the wreck to be intact, a highresolution multibeam survey with overlaid ship's plans will determine which tanks could contain oil and should be tested for oil using neutron backscatter or other suitable techniques.