Golden Ray: 40 Hours Trapped Inside a Capsized Ship

Four crew members were still inside this ship. For more than sixteen hours, no one outside knew if they were alive. The ship was the Golden Ray, a car carrier two hundred metres long. Leaving the port of Brunswick, Georgia, it heeled over to sixty degrees in less than a minute. No storm. A wind of four knots. Engines and rudder working. The investigators found the problem in numbers entered by hand into the stability computer, before the ship sailed. So how did a modern ship leave port without enough stability, and no one on board know? And how do you reach four people inside a hull lying on its side?

The Golden Ray was a floating car park: twelve enclosed decks of vehicles, stacked inside a steel box. A South Korean company, G-Marine Service, crewed and managed it. In the IMDG Code, the rulebook for dangerous goods at sea, a gasoline or diesel car is UN 3166, class 9. But under special provision 961, cars stowed in ro-ro spaces, with no sign of leakage, are exempt from most of the Code. A 2018 review by the US Transportation Research Board warned that car carriers, with high sides and cargo on high decks, are more prone to losing stability to begin with. What helps keep them upright is weight down low: fuel, and seawater in ballast tanks at the bottom of the hull.

It started with a hurricane. In late August, on the way from Texas, the master learned that Hurricane Dorian was heading up Florida's east coast. To make the ship more stable for the storm, the chief officer pumped about fifteen hundred tonnes of seawater into tanks low in the hull. The ship waited off Key West for the hurricane to pass. Then the plan changed. The ship was sent to Jacksonville, Florida, where the port limited its draft, how deep it could sit in the water. So the chief officer pumped the same fifteen hundred tonnes back out. When it left Brunswick, the Golden Ray carried 4,161 vehicles, about 373 tonnes more cargo than when it arrived. That ballast never went back in. Keep it in mind.

Before the ship left Brunswick, its stability rested on one person. The chief officer told investigators no one else on board knew how to use the stability computer. The computer was called LOADCOM. It could read the tank levels straight from the ship's monitoring system. He entered them by hand. His result: a GM of 2.45 metres. GM, the metacentric height, is a measure of how hard a ship pushes back when it first starts to heel. He told the master the computer said it was OK. The master did not check the calculation. The company's procedures did not require anyone to. There was something else. Since 11:41 that night, two watertight doors on deck five had stood open. They led down to the engine room and the steering gear. On the bridge, a panel would have shown them in red. As the ship left, the crew opened the pilot door on the port side, a hatch about two metres square, ready to drop off the pilot at sea. Then no one stayed with it. Keep that door in mind.

At 1:36 in the morning, the pilot ordered the last turn, toward the open sea. The ship had made two left turns without trouble. Now, at about thirteen knots, it entered a sixty-eight degree turn to the right. Starboard ten. Starboard twenty. Then, eleven seconds after he ordered the rudder back to midships, the ship began to heel hard to port. The pilot later said the ship "just took off". On the voyage data recorder, the bridge crew can be heard in surprise. And the pilot asked: what's the GM on this thing? Then equipment slid across the bridge, and alarms began to sound. Port rudder did not stop it. In the NTSB's reconstruction, the ship heeled past fifty degrees in about fifty seconds, and past sixty in less than a minute. Its turn to the right sped up, from about two degrees a minute to more than forty. At 1:40, the Golden Ray ran aground outside the channel, lying at sixty degrees. The engines and generators had stopped. The emergency generator ran for a few minutes, then quit. The ship went dark. And that pilot door was open. As the ship heeled past about seventeen degrees, the sea came in through it, onto deck five, and down through the open watertight doors.

The bridge was now high above the water. The first Coast Guard boat launched at 1:54. Over the next hour, eleven people, among them the pilot and the master, were lowered from the bridge to the boat on a fire hose. A helicopter lifted two more off the high side. Other responders reached five more. Then, around 4:30, responders saw flames on the vehicle decks. The decks held more than four thousand cars, with fuel in their tanks, tyres and plastic. Thick black smoke and heat kept rescuers out. The fire burned for about a day. At 6:45, the boat crew broke a cabin window below the bridge wing and pulled out two more, among them the chief engineer. That made twenty people safe. Four were still missing: three engineers and a cadet. By 9:30, the hull had rolled onto its side, to ninety degrees. Responders heard loud crashing from inside.

Inside the engine room, there was no light at all. The cadet had a flashlight. After the capsizing, two engineers in the control room tried the main stairwell. Water was pouring in from an open door on deck five. Their way out was blocked. One got through another door to join the other engineer and the cadet. The fourth stayed in the control room. The water kept rising, oily, on the low side. They climbed higher. The heat built up. They told investigators it was hard to breathe. In the end they got into the floodwater, just to stay cool. Remember the sixteen hours? At 6:12 that evening, just over sixteen hours after the capsizing, responders heard tapping from inside the hull. To cut in safely, they planned to use cold-cutting gear, about fourteen hours away. The next day, at one in the afternoon, they drilled a small hole through the hull and made contact. The air inside measured up to 155 degrees Fahrenheit, about 68 Celsius. The rescuers did not wait. They kept drilling, hole after hole, into a square opening. At three o'clock, the two engineers and the cadet came out. The fourth was behind the explosion-proof glass of the control room. Rescuers broke through it, and at 5:51 he climbed out. After nearly forty hours trapped, the last of them was out. All twenty-four people survived. Two crew members had serious injuries, a broken finger and a broken foot.

It was not the weather. Not the engines or the steering, not a cargo shift, not the channel, and not the fire, which likely started after the ship rolled. The NTSB's finding: the Golden Ray capsized because it did not have enough righting energy to resist the turn. Coast Guard engineers rebuilt the loading condition from the ship's own tank data and load plan. Its GM was about 1.76 metres. The chief officer had reported 2.45. The minimum for that condition was 2.54. Other analyses agreed: well below what he reported. The ship's centre of gravity sat too high. Remember the ballast pumped out for Jacksonville? With about 1,492 more tonnes of it, the Coast Guard found, the ship would have met the international standard. Not just on this voyage, but on the two before. In a test the investigators did not observe, the charterer gave the stability computer of the sister ship, the Silver Ray, its own data for the Golden Ray. It showed "Not OK". The NTSB concluded that the chief officer made errors entering the ballast tank levels. What he entered is not known: seawater ruined the computer. Why he typed the data in by hand is also unknown. The NTSB also found a contributing factor in the company. Its procedures left stability to one officer, with no company training on that computer, and no one checking his numbers. The report calls it a single point of failure. And the open watertight doors went against the company's own departure checklist. After the heel, they let the sea in and blocked the main way out of the engine room.

To get the wreck out, salvors cut it into eight pieces, with a chain. A giant floating crane, the VB 10,000, straddled the hull. It pulled a heavy anchor chain through the steel, seven cuts in all. Around the wreck stood a barrier about a mile long, to catch oil and debris. Cutting began in November 2020. On May 14th, 2021, workers were using torches to clear steel from the chain's path. The wreck caught fire. The Coast Guard said it broke out while they were using the torches. Smoke poured from the hull. Tugs sprayed seawater on it for hours, and the thirty to forty workers on the crane were evacuated. No one was hurt. The last section left St. Simons Sound on October 25th, 2021, more than two years after the capsizing. The Coast Guard's federal on-scene coordinator called it the largest wreck removal in US history. When the NTSB reported in 2021, salvage costs were estimated at more than 250 million dollars, and still rising.

On January 1st, 2020, four months after the capsizing, an amendment to SOLAS, the main treaty on safety at sea, took effect. Adopted back in 2017, it extended, according to the NTSB, a rule for passenger ships to cargo ships: before departure, the master must determine the ship's stability, and record that it meets the criteria. G-Marine added stability training, made officers compare the computer's GM with the ship's booklet, and kept ships in port until the final load plan was checked. The NTSB's two recommendations went to the company: verify stability before leaving port, and audit that watertight doors are closed.

The Golden Ray did not meet a storm. It met a routine turn, at a normal speed, in light wind. What it lacked was enough ballast, and a second check that would have caught it. Stability is a calculation, and a calculation needs someone to check it. This is Hazard Manifest. If you want to understand the dangerous side of global trade, subscribe.