Industries / Transport and vehicles / Maritime and shipping / Marine salvage
Refloating and towing
This is the work of getting a grounded or disabled ship free and towing it to safety: pulling a stranded hull off, lightening it or dredging around it, and making a towline fast to a ship that has lost its engine or steering.
5
Milestones achieved, from 1978 to 2022.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1978
A tow parts off Brittany
On 16 March 1978 the tanker Amoco Cadiz, loaded with 223,000 tonnes of crude oil by Cedre's figure, lost her steering off Brittany. The German salvage tug Pacific, of 10,000 horsepower, began to pull shortly after 14:00 GMT. By the tug master's account four and a half hours passed between his first offer and 16:00, when the tanker, after approval from her owners in Chicago, accepted Lloyd's Open Form, the standard "no cure, no pay" salvage contract. At 16:15 the tow chain snapped at the tanker's end in winds of force 8 gusting 9 to 10. A second line was made fast at 20:55, and at 21:04 the tanker grounded near Portsall. All her cargo spilled and polluted 360 km of coast.
Lessonin a gale a tanker that has lost steering can be ashore within hours, so the towing gear and the terms have to be ready before help arrives.
Credit
French Senate commission of inquiry, as summarized by Patrick Lagadec; Cedre
1994
Towing gear required on tankers
On 20 May 1994 IMO's Maritime Safety Committee adopted SOLAS amendments requiring new and existing tankers of 20,000 tonnes deadweight and above to carry emergency towing arrangements, fittings at bow and stern to which a tug can quickly connect. Its guidelines set a working strength of 1,000 kN (about 100 tonnes of pull) for tankers under 50,000 tonnes deadweight and 2,000 kN above that, and ask that the stern gear be deployable in harbour in 15 minutes and the bow gear in an hour. A 2008 amendment required every passenger ship by 2010, and every cargo ship by 2012, to carry its own emergency towing procedure. In May 2024 IMO extended the towing-gear rule to ships other than tankers of 20,000 gross tonnage and more built from 1 January 2028.
Credit
IMO (resolutions MSC.31(63), MSC.35(63), MSC.256(84) and MSC.549(108))
2011
Britain retires its emergency tugs
From the time of Lord Donaldson's 1994 report, Britain's Maritime and Coastguard Agency (MCA) chartered emergency towing vessels, tugs kept on standby for ships in trouble. It ran four until 2011, when the government removed them on the assumption that commercial tugs would fill the gap. France, whose navy charters its own rescue tugs such as Abeille Flandre and Abeille Languedoc, had shared the charter of one British tug, Anglian Monarch, in the Dover Strait, and its minister wrote to London with France's concerns on 13 January 2011. A 2020 study for the MCA found the commercial towage market "has not responded in the way it was originally envisaged" and noted the "extreme paucity of sufficiently-powered tugs" in many regions; one tug had by then been reinstated in Scotland.
Lessona market for standby tugs did not appear on its own when the state stopped paying for them.
assets.publishing.service.gov.ukassemblee-nationale.frassemblee-nationale.fr
Credit
UK Maritime and Coastguard Agency; the French government
2019
No tug for the largest
Clear Seas, an independent Canadian research centre, used 2016 AIS data to map the commercial tugs on Canada's Pacific coast and to test whether a tug of opportunity, one that happens to be nearby, could hold a drifting ship. It found that in 2016 no tug of opportunity could rescue the largest ships studied, such as very large container ships, LNG carriers and cruise ships, in winds above 33 knots, and that such tugs are usually towing barges and must first hand off their tow. The two emergency towing vessels leased by the Canadian Coast Guard covered all ship types except the largest container ships in the worst 1% of weather, and the study suggested permanent vessels of 213 tonnes bollard pull, the steady pull a tug can exert.
Lessona coast busy with tugs can still lack one strong enough for its biggest ships.
Credit
Clear Seas Centre for Responsible Marine Shipping
2022
Ever Forward lightened and dredged
On 13 March 2022 the 1,095 ft (334 m) container ship Ever Forward grounded along her whole length outside the channel in Chesapeake Bay. Two attempts to pull her off, on 29 and 30 March, failed, and salvage experts concluded they could not overcome the ground force, the part of the ship's weight carried by the seabed, with 4,964 containers aboard. Dredgers removed 206,280 cubic yards (about 158,000 m³) of sediment to a depth of 43 ft, crane barges lifted off 500 containers, and on 17 April two pulling barges and six tugs from Donjon-SMIT, Moran and McAllister freed her after 35 days.
Lessonwhen pulling fails, the ship has to be made lighter or the seabed lower.
Credit
US Coast Guard; Maryland Department of the Environment; Evergreen Marine; Donjon-SMIT
Watch for
Towing gear for large ships
In June 2025 IMO's Maritime Safety Committee approved interim guidelines for emergency towing arrangements on ships other than tankers (MSC.1/Circ.1691), with revised guidance on shipboard towing and mooring equipment and on emergency towing procedures. They go with the SOLAS rule adopted in May 2024 that requires such gear on non-tankers of 20,000 gross tonnage and more built from 1 January 2028.
Achieved whenships built under the 2028 rule are delivered with approved towing gear, and a flag state or classification society reports a real tow made with it.
Claimed by
IMO Maritime Safety Committee, June 2025
Open
01
Drift forecasts checked against AIS
A public test of drift forecasts for large ships that have lost power. Collect AIS tracks of ships that drifted, re-run each drift with the ShipDrift model in OpenDrift from the moment power was lost, using archived wind, wave and current data, and publish how far the forecast was from the real track after 6, 12 and 24 hours. Responders use such forecasts to judge how long they have before a ship reaches the shore, and so how fast a tug must come.
The briefCredit · unclaimed02
Pull-off tests on model hulls
A bench test that measures the force needed to pull a model hull off sand, mud and gravel. Rest a 1 m model on each bed with a known ground reaction (the share of its weight carried by the bottom, set with added weights), pull it with a winch through a load cell, and repeat after cutting a trench along the hull as a dredger would. Publish the ratio of pulling force to ground reaction for each bed, the figure salvors estimate when they plan a refloat.
The briefCredit · unclaimed
Open 01 · the brief
Drift forecasts checked against AIS
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test of drift forecasts for large ships that have lost power. Collect AIS tracks of ships that drifted, re-run each drift with the ShipDrift model in OpenDrift from the moment power was lost, using archived wind, wave and current data, and publish how far the forecast was from the real track after 6, 12 and 24 hours. Responders use such forecasts to judge how long they have before a ship reaches the shore, and so how fast a tug must come.
Why it is still open
OpenDrift's own example compares ShipDrift only with another model, Leeway, not with observed tracks. DNV's Emergency Response Service launched a drift prediction service built on OpenDrift in 2022 and gave no error figures in its announcement. A 2023 study for the MAIB used OpenDrift for a small boat in the Dover Strait, not a ship. This research found no published check of ShipDrift against real drifts of large ships.
opendrift.github.iohellenicshippingnews.comassets.publishing.service.gov.uk
Built with what exists
OpenDrift, the open-source Python framework of the Norwegian Meteorological Institute; NOAA's MarineCadastre AIS files for US waters; ERA5 reanalysis winds and Copernicus Marine currents and waves; drift positions printed in investigation reports, such as the BSU's MSC Flaminia chronology, for cases outside US waters.
First step
Search one month of NOAA AIS data for large ships moving at under 2 knots for six hours or more far from anchorages, and check each candidate against Coast Guard news releases.
Done looks like
30 drift cases, each with the forecast error in kilometres at 6, 12 and 24 hours, set against a simple rule of thumb (drift at a fixed share of the wind speed); code, inputs and cases public.
Open 02 · the brief
Pull-off tests on model hulls
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A bench test that measures the force needed to pull a model hull off sand, mud and gravel. Rest a 1 m model on each bed with a known ground reaction (the share of its weight carried by the bottom, set with added weights), pull it with a winch through a load cell, and repeat after cutting a trench along the hull as a dredger would. Publish the ratio of pulling force to ground reaction for each bed, the figure salvors estimate when they plan a refloat.
Why it is still open
The nearest work is a 2017 University of Maryland thesis that pulled models of three ship sections out of saturated sand, but only as a baseline for freeing them with air blasts, and a 2025 DTU study of a model ship running onto a sandy slope, which measured the grounding rather than the refloat. At full scale the stakes show in the Ever Forward case, where two pulls failed before salvors decided to lighten and dredge. This research found no open table of measured pull-off ratios by type of bottom.
Built with what exists
A water tank or flume; a 3D-printed or fibreglass model hull; sand, clay and gravel from a builders' merchant; an S-type load cell with an HX711 amplifier and an Arduino to log force; a hand or stepper-motor winch; Zenodo to publish the data.
First step
Build a 1 m model, set it on a sand bed in a tank, and log ten pulls through a load cell.
Done looks like
At least 10 pulls for each of three beds at three ground reactions, with and without a trench, each with its force trace, and the ratio for each case reported with its spread.