Nothing on the Railroad Could Tell That the Bridge Had Moved

This is the first finding of a federal investigation into the deadliest accident in the history of Amtrak.

The train crew's qualifications, and the condition of the track, the signals, and the train equipment did not contribute to this accident.

That is not a defense of the railroad. It is the whole problem, written down in one sentence on page fifty-nine.

Everything on that railroad worked exactly the way it had been designed to work.

And forty-seven people died anyway.

This is the record of what a machine cannot notice.

Just after midnight on the twenty-second of September, nineteen ninety-three, a towboat named the Mauvilla left a barge fleet on the Mobile River.

She was eighty-five feet long and she was pushing six barges, in two rows of three.

Four men aboard. A captain, a pilot, and two deckhands. The pilot had the watch.

The National Transportation Safety Board never prints his name in this report, and neither will this video. He is the pilot.

Fog was not forecast that night. The weather report at the time of the accident called for four miles of visibility.

Fog formed anyway. It does that on this water. Quickly.

The pilot's own testimony. As he passed Catfish Bayou, quote, it started getting foggier and foggier on me.

He called another towboat ahead of him and asked about the fog. The answer came back in two words.

It's shut out. Zero visibility.

The other operator kept going. The pilot of the Mauvilla decided to tie up and wait it out.

That was the right decision. On this river you tie a line around a tree on the bank and you stop.

He could no longer see the head of his own tow. It was more than four hundred feet in front of him.

He sent a deckhand forward onto the lead barge with a searchlight to snag a tree.

Two attempts. Neither caught. The pilot could not see the man he had sent out there.

So he ordered him back, afraid the deckhand would be hurt or would go over the side in the fog.

Every decision in that sentence is a decent one. Hold on to that.

While he was looking for that tree, something happened that he noticed and did not understand.

His swing meter showed the tow turning to port. Two banks appeared on the radar screen as the barges straightened out.

He had turned out of the Mobile River and into a side channel called the Big Bayou Canot.

Here is the geography. The Safety Board printed it as figure two.

The bayou leaves the river and runs west, and it is a natural channel, and it looks like a river.

The Coast Guard would put this in writing two months later.

There are few natural landmarks that allow mariners to visually distinguish between this waterway and the Mobile River.

And: vessels transiting the Mobile River may enter Big Bayou Canot in error.

That is the Coast Guard, in November of nineteen ninety-three, describing a hazard it had not marked.

No warning sign. No waterway marking. No aid to navigation of any kind in that bayou.

The pilot's own words: at no time did he see the intersection, visually or by radar.

He believed he was still on the river, heading for a cutoff where he knew there was a tree.

Then an object appeared on his radar, lying straight across the water in front of him.

He assumed it was another towboat with barges that had swung out from the bank.

He never considered that it might be anything else. He decided to tie up alongside it.

It was a railroad bridge.

A single-track bridge belonging to CSX Transportation, with a vertical clearance of about seven feet above the water.

This is the last photograph taken of it. A routine inspection, before.

He kept glancing at the radar. He testified he did not know what scale it was set to.

He idled forward at one or two knots. He tried the radio twice. He shone his lights at it.

Then he put the throttles in reverse, and gave the Safety Board the sentence this entire episode turns on.

I was still trying to figure out what was in front of me. And I never could.

He felt a bump. He thought he had run aground.

The deckhand felt it too, sitting in the galley, and noted the time. Two forty-five in the morning.

Two of the lead barges struck the piers. The center barge, which stuck out about five feet further ahead, struck the girder between them.

And displaced the south end of that girder span thirty-eight inches to the west.

The girder now stuck out into the space that trains travel through.

Nobody on the towboat knew a bridge had been hit.

The captain woke up from the bump and came to the wheelhouse. The pilot told him where he thought they were.

Somewhere between Twelve Mile Island and Thirteen Mile Marsh, in that neighborhood. He was in neither.

Barges had broken loose. Trying to recover them, the Mauvilla backed into the east bank of the bayou and stuck there.

She stayed pinned against that bank for about eight minutes.

This is the entire margin. Eight minutes, and nobody on earth knows the bridge has moved.

Those same eight minutes were being crossed by a train with two hundred and twenty people on it.

Amtrak train two. The Sunset Limited, Los Angeles to Miami.

She had left New Orleans thirty-four minutes late, held up for repairs to an air conditioner and a toilet.

She stopped at Mobile at half past two in the morning. The conductor helped people off and on.

She left Mobile at two thirty-three with two hundred and two passengers aboard.

The conductor described the run from Mobile as uneventful.

The locomotive crew called every signal aloud over the radio, and the conductor repeated each one back.

The last signal he remembered them calling was at Bayou Sara, one and seven tenths of a mile back.

The call was clear.

The train accelerated. Seventy-two miles an hour, in seventy mile an hour territory.

At two fifty-three in the morning she hit the girder.

Three locomotives, the baggage car, the dormitory car and two of the six passenger cars went into the water.

The fuel tanks ruptured and burned. The lead unit ended up buried in about forty-six feet of mud.

This is the Safety Board's photograph of the site.

Forty-two passengers and five crew members were killed. One hundred and three passengers were injured.

The four men on the towboat were not hurt.

The bridge had moved thirty-eight inches out of alignment.

And the signal one and seven tenths of a mile back was showing clear.

The rails did not break.

A wayside signal system works by running a low current through the rails themselves. The rail is the wire.

Break the rail and you break the circuit, and the signal behind it goes red. That is the entire safety principle.

The barge moved the bridge sideways. The track went with it. The steel bent and stayed whole.

Because the rails remained intact and rail continuity was maintained, the signal circuitry was not interrupted.

The signal displayed for train two on the night of the accident was clear, or proceed.

According to tests performed after the accident, the signal system operated as designed.

It was never broken. It was answering a different question than the one that mattered.

It could tell you whether the rail was continuous. It could not tell you whether the rail was where you left it.

Somebody had noticed that. Fourteen years earlier.

November seventeenth, nineteen seventy-nine. Devil's Slide, Utah.

A Union Pacific freight train doing fifty miles an hour struck a displaced bridge and derailed five locomotives and fifty-six cars.

Damage over five million dollars. No passengers on a freight train.

Out of that wreck the Safety Board issued Safety Recommendation R-eighty dash thirty-six.

It asked the Federal Railroad Administration to study the feasibility of installing a mechanism which can be incorporated in the automatic block system to indicate when bridges are displaced.

That is it. That is the whole request. Study whether a signal system could be made to notice a bridge that has moved.

The FRA answered on the sixth of May, nineteen eighty-one.

Ninety-eight thousand route-miles of track sat inside signal territory in the continental United States.

Those route-miles contained some eighty-five thousand bridges.

Installing displacement detection on every one of them would cost about eight hundred and fifty million dollars.

Plus eighty-five million dollars a year to maintain.

Of the forty-one thousand six hundred and twenty-seven railroad accidents between nineteen seventy-six and nineteen seventy-nine, only twenty were caused by displaced bridges or bridges that failed under load.

Of those twenty, four resulted from misalignments that the device would have caught.

The FRA weighed the cost against the number of accidents and concluded that projected costs far outweighed the benefits.

That is a defensible calculation. It is also the only calculation anybody ran.

The Safety Board wrote back in August of nineteen eighty-one and questioned the arithmetic.

Ten thousand dollars to install one and a thousand dollars a year to maintain it struck the Board as excessive.

So the Board asked the FRA for a copy of its analysis.

The FRA did not respond to the Safety Board's request.

Three years later the agency reported it had looked at nineteen eighty through eighty-two and found six bridge accidents, none of them detectable by a device.

On the twenty-fifth of February, nineteen eighty-five, the Safety Board closed the recommendation.

Closed, acceptable action. The Board asked the FRA to revisit it periodically as signal technology improved. The FRA agreed to.

That file had been shut for eight years, seven months and twenty-eight days when the Sunset Limited left Mobile.

Now go back to the towboat, because the same shape shows up there.

Since nineteen seventy-two, anyone standing a navigation watch on a towing vessel over twenty-six feet has needed a federal license.

The Mauvilla was eighty-five feet. Both her captain and her pilot held one.

Read the next sentence in the government's own report slowly.

The licensing regulations recognize the limited formal educational background of many towboat operators by reducing the scope of knowledge required to obtain a license.

An applicant supplies a letter from his employer saying he received training or had duty in the wheelhouse.

The regulations do not prescribe the level or type of wheelhouse training in vessel handling or equipment use.

Then a written test in four parts, which you may retake without limit.

Six point four percent in nineteen eighty-nine, twenty-six of two hundred eighty-three candidates.

Twenty-six divided by two hundred eighty-three is nine point two percent. The next line does not work either.

The report does not explain the difference, and the Coast Guard, it says, does not compile data on how many times a candidate retakes the test.

To meet requirements for the license, an operator does not have to be trained in the use of radar.

And the Mauvilla pilot was not so trained.

Radar training was required. For men working the ocean. Not for the man in the fog on the river.

He learned it the way almost all of them learned it. On the job, from another man, with no examination at the end.

There were two other things on that towboat that would have told him where he was.

The first is a chart. The Mauvilla had none.

She was not required to carry one. Equipment rules for uninspected towing vessels cover life preservers, fire extinguishers, cooking and heating.

They do not cover navigation equipment at all.

The pilot had his own set of charts once. He told the Board what happened to them.

I used to carry my chart with me all the time. I had it about a month, I guess, and I started leaving it at home.

Once you know the river, you don't have to have it. Nobody requires you to have a chart because you don't navigate by the chart.

His employer's general manager confirmed it in testimony. Charts are not required as standard operating equipment.

The second thing is a compass.

But you cannot look at it in fog. And the Safety Board did the one piece of arithmetic that nobody had done.

The difference in compass headings between the Mobile River and the Big Bayou Canot is about ninety-five degrees.

Ninety-five degrees. A quarter turn, near enough. On any compass on earth, that is not a subtle reading.

The Board had already asked the Coast Guard to require compasses on towboats. Nine months before this wreck. After a different accident.

At two fifty-six, the assistant conductor put out a Mayday on the railroad radio.

He said he thought he was at the Mobile River bridge. He was three point two miles from it.

A yard in Mobile heard him and called the dispatcher in Jacksonville, Florida, at two fifty-seven.

Then somebody went to call the Coast Guard.

The telephone number for the Coast Guard in the Mobile telephone directory was incorrect.

Two separate people tried it. Both failed. Somebody had to go through the operator.

The nine one one operator was told the accident was at the wrong bridge.

Eighteen minutes went by while police, fire and the railroad worked out where it actually was and whose jurisdiction it was in.

The Mobile Fire Department's map did not even show the Big Bayou Canot, due to an omission during photocopying.

The bayou existed. The copier had cut it off the edge of the page.

The bridge itself had no name plate and no mile marker on it. Boats searching for it went to the wrong bridge first.

They found it by seeing the fire.

The men who hit it did not learn where they were until three fifty-five in the morning.

In the water, in the dark, people were getting out of two submerged double-decker cars.

Once the cars went under, the emergency lighting stopped working.

Only the train crew had flashlights, and flashlights were not part of Amtrak's emergency equipment.

A few passengers had penlights. That is what people used to find the exits.

Emergency responders could not get an adequate list of who was on that train until the next day.

The county had run disaster drills. It had never run one for a railroad accident.

Representatives of Amtrak, CSX, the county and the Coast Guard testified that they had never discussed the possibility of an accident on that bridge.

Given the remoteness, the fog and the limited ways in, the response was well coordinated and effective.

Towboat crew, a passing vessel's crew, train crew and volunteers pulled people out of that water.

The failures in this story are not failures of the people who showed up.

On the nineteenth of September, nineteen ninety-four, the Safety Board adopted its report.

The probable cause names the pilot's lack of radar competency, and his company's failure to make sure he had it.

And third, the U.S. Coast Guard's failure to establish higher standards for inland towing vessel operator licensing.

A federal agency named as a cause of forty-seven deaths, in a document that agency's own department published.

Contributing: the lack of a national program to assess whether bridges can survive being hit by boats.

Then twenty-nine recommendations, to nine different recipients.

Radar training. Radar aboard. Charts aboard. Compasses. Alcohol testing time limits. Bridge markings. A national bridge register.

Every one of them addresses something real in this record.

Read all twenty-nine, to all nine recipients, and count how many concern a way to detect that a bridge has moved.

None of them.

The Board spent two pages on that question, laid out the whole history, and issued nothing.

It closed the section by noting that the FRA is evaluating the need for bridge alignment protection.

Five railroad signal suppliers had told the Board's investigators, by telephone, that no such device had been designed.

Ten weeks after Mobile, the same thing happened in Tennessee.

The first of December, nineteen ninety-three. A towboat and two barges struck a Norfolk Southern bridge in downtown Knoxville.

The pier moved twelve to eighteen inches. The track moved nine.

The towboat operator called it in and gave the wrong bridge. It took an hour to work out which bridge it was.

Every failure from Mobile, repeating exactly, ten weeks later.

No train hit that bridge. Here is why.

Someone looking out his office window had seen the accident and immediately called Norfolk Southern Railroad.

Which was able to prevent a train about four miles from the damaged bridge from crossing it.

In Knoxville it was daylight, and a man happened to be looking out a window.

At Big Bayou Canot it was quarter to three in the morning, in fog, in a bayou with no houses on it.

The difference between forty-seven deaths and none was a person who happened to be looking.

Things did change. Slowly, and in the places the Board had aimed.

In May of nineteen ninety-eight, the Federal Railroad Administration issued a rule on passenger train emergency preparedness.

It opens by describing this accident.

Emergency plans. Marked emergency windows. Safety information for passengers. Accurate passenger counts for responders. Drills.

Every one of those is a finding from Big Bayou Canot, turned into a federal requirement, four years and seven months later.

In November of nineteen ninety-nine, the Coast Guard rewrote licensing for towing vessel officers.

That rule opens by describing this accident too.

Watch the wording the Coast Guard chose for the finding against itself.

In part, the Board maintained, because of the U.S. Coast Guard's failure to establish higher standards.

The Board maintained. Six years on, still held at arm's length.

But the rule itself is real. An approved course, with radar in it, and a practical demonstration of skill before a designated examiner.

Six years and two months after the wreck. And it was published as an interim rule.

Which leaves the bridge.

In July of twenty-ten, the FRA finally issued federal bridge safety standards for railroads.

Seventeen years later. It requires a special inspection whenever a bridge is involved in an event that might have compromised it.

Including, in the text, vessel impact.

And it says the track owner should have in place a means to receive notice of such an event.

Receive notice. From someone who saw it.

That is the whole architecture, and it has not changed. The rule presumes a person knows.

At Big Bayou Canot, for eight minutes, no person on earth knew.

Not the dispatcher in Jacksonville. Not the railroad. Not the Coast Guard.

Not the crew of the Sunset Limited, calling a clear signal aloud and repeating it back to each other over the radio.

Not even the man who hit it, standing in the wheelhouse four hundred feet away, still trying to figure out what was in front of him.

In nineteen eighty-one somebody priced the machine that would have known. Eight hundred and fifty million dollars.

The number was probably right. The file was closed in good faith. Nobody in it did anything wrong.

That is what makes it worth showing you. This is what a system looks like when every part of it is working.