The Difference Bridge

A few days ago I told part of this story inside a piece about McKinsey. This is the fuller version.

Laufenburg is a medieval town that sits on both banks of the High Rhine, one half on the Swiss side and the other on the German. The two halves share a name, a river, and very little else. In 2003 they decided to build a bridge between them: a prestressed concrete road bridge that would take the heavy lorries out of the narrow streets of both town centres. It was the kind of infrastructure project that is supposed to be routine.

Each country would build its own half. The two halves would meet in the middle. The total budget was around twelve million Swiss francs. The target opening was summer 2004³.

Where sea level is

The first thing that went wrong was geography, and it had gone wrong two hundred years earlier.

Switzerland and Germany do not agree on where sea level is. They cannot. Neither country has a sea. Both had to adopt the zero point of some other country’s coast as a reference, and they adopted different ones.

Germany’s zero is in the Netherlands. The Dutch have been measuring the level of the North Sea at Amsterdam since the seventeenth century, and they codified their measurement as the Normaal Amsterdams Peil, the Amsterdam Ordnance Datum. When Germany’s surveyors needed a reference frame for their own elevation work, they imported the Dutch one. It has been the German zero ever since.

Switzerland’s zero is in France. In the early 1800s, when Swiss surveyors were laying out the country’s elevation network, Geneva was a French-speaking scientific hub, the Republic had been briefly part of France, and the obvious reference was French. The French reference was the Mediterranean, measured at the tidal gauge in Marseille. The Swiss imported it. But they also needed a local calibration marker, something they could physically touch without walking to Provence. They found it in Lake Geneva, embedded in a rock that sits just above the water a few metres off the shore at Eaux-Vives. The rock is called the Repère Pierre du Niton. It is the primary vertical benchmark of the entire Swiss surveying system. Every elevation marker in Switzerland is ultimately referenced to a rock in a lake that was calibrated against a tidal gauge a thousand kilometres away⁴.

Neither country chose its system because the other one was wrong. They chose the systems they chose because of who their neighbours were in 1820, and the choices froze into infrastructure, and infrastructure becomes permanent.

The Mediterranean and the North Sea are not at the same height. The Mediterranean is an almost-enclosed basin with weak tides and high evaporation. It sits lower, on average, than the North Sea by about twenty-seven centimetres. So the Swiss zero and the German zero are twenty-seven centimetres apart, and that gap is not an error. That gap is geography. The engineers who designed the Laufenburg bridge knew this. Their drawings accounted for it. Every number on every plan was correct.

The minus sign

Something happened during implementation that nobody has ever been able to fully explain.

Instead of adding twenty-seven centimetres to the Swiss side of the design to close the gap, the compensation was inverted. Twenty-seven centimetres was subtracted instead. The correction ran in the wrong direction. The gap between the two sides was not closed. It was opened further.

The new gap was fifty-four centimetres. Twenty-seven centimetres of real geography, plus another twenty-seven centimetres of mathematical direction error, stacked on top of each other. A full half metre of empty air where the bridge was supposed to meet itself.

Beat von Arx, the project manager for the Swiss canton of Aargau’s Civil Works department, described the situation with unusual clarity once the error was discovered. “The difference of 27 cm was certainly known,” von Arx told the press, “and everything had been drawn up correctly on paper.”¹

Everything was correct on paper. The error was in the transition from paper to concrete. The minus sign went into the implementation instructions, the implementation instructions went through an endless chain of engineers and approvals, and nobody caught it. The check was not missing because the engineers were lazy. The check was missing because everyone upstream of the check assumed someone downstream would perform it, and everyone downstream assumed someone upstream already had.

The error poured itself into concrete.

Christmas Eve

Nedeljko Madzarac is the Swiss supervising site manager for the bridge. On the afternoon of Christmas Eve 2003 Madzarac walked out onto the half-built structure with the team for the last scheduled site inspection before the holidays. Two stages of the Swiss half had already been concreted. The German half was coming toward them from the opposite bank.

Madzarac looked across.

The German side was visibly higher than the Swiss. Not by a small amount. Not by the kind of tolerance that could be absorbed into a transition plate. By fifty-four centimetres. The two halves were not going to meet.

“We had already concreted two stages when it became apparent that something wasn’t right with the height,” Madzarac told the Berner Rundschau afterwards¹.

The team stood on the bridge and discussed the situation among themselves. There was not much to discuss. The concrete was not going to move. The geometry was not going to change. The holidays were about to start. Madzarac picked up the phone and called the Swiss engineering office.

What happened next was, by the standards of international construction disputes, astonishing. There was no drawn-out negotiation between the two sides. There was no round of finger-pointing across the Rhine. The Swiss consortium took responsibility within days. Beat von Arx, representing the Aargau canton, stepped in front of the press and said the sentence that became the public memory of the whole affair.

He was able to deliver on it. The full cost of the correction was absorbed by the Swiss engineering consortium’s liability insurance. No public money was spent.

The press

The press had a field day. It was Christmas, there was no other news, and the story was irresistible: two countries build a joint bridge and the two halves do not meet.

On the fourteenth of January 2004, SPIEGEL ONLINE ran a piece mocking the project as a “bridge over the river Rhine with steps.”² The next day, the Weser Kurier in Bremen published an article under the headline Neue Rheinbrücke ist ein Reinfall¹. Reinfall means a flop. It is also, to any German speaker, one consonant away from Rheinfall, the Rhine Falls. A new Rhine bridge is a Rhine-flop. The pun was unimprovable and the headline wrote itself.

Six days later the Stuttgarter Zeitung ran a longer piece on the “embarrassing discovery” and the international dimensions of the story. The coverage eventually faded, as coverage does, but the local memory did not. People in Laufenburg, on both sides of the Rhine, started calling the bridge die Differenzbrücke. The difference bridge. They have been calling it that ever since.

The fix

The physical fix was less exciting than the story of the discovery. The engineers went back to the drawings, found the sign error, applied the correct compensation this time, and adjusted the structure to match. The German abutment on the north bank was modified. The superstructure in the middle was equalised. Parts of the already-poured Swiss concrete had to be broken out and re-poured. The work took a few months.

The bridge opened to traffic in the summer of 2004, on schedule. The two halves met in the middle. The difference was gone from the physical structure, but the nickname stuck. Twenty-two years later, die Differenzbrücke is still what the locals call it.

What the bridge is for

The engineers who built the Laufenburg bridge were not ignorant. They had full awareness of the problem they were solving. The twenty-seven-centimetre offset was drawn up correctly on paper. Every document that mattered accounted for it. Von Arx was right that everything had been done correctly on paper.

What they lacked was the small piece of process that would have asked, before the concrete was poured: the correction was supposed to make the gap smaller; did it?

That piece of process does not take long to run. Walk onto the bridge with a rough theodolite. Compare the two levels before the second pour. Look, across the Rhine, at the thing you have built so far. Does it match what the drawings said it would match? No? Then something has gone in the wrong direction. Stop pouring.

The check did not exist because everyone upstream of the concrete assumed someone downstream would perform it, and everyone downstream assumed someone upstream already had. The check did not exist because it lived in no one’s job description. The check did not exist because the awareness of the problem had been so thorough, so carefully drawn, so completely documented, that the possibility of getting the sign backwards on the implementation had been crowded out by the confidence that came from knowing what the offset was.

Knowing what the offset is does not tell you whether you applied it in the right direction.

This is the reason I care about the bridge. The specific shape of the mistake, the sign error hiding inside complete awareness, is the most familiar shape in the world to anyone who has worked inside a modern organisation. It is not the shape of ignorance. It is the shape of overconfidence produced by documented knowledge. The most dangerous position in any complex system is the one where you are sure you know enough.

The bridge at Laufenburg met itself in the end. Most bridges do not.

Sources

1. Nedeljko Madzarac quote as reported in the Berner Rundschau, January 2004. Beat von Arx quotes as reported in the Stuttgarter Zeitung, January 20, 2004, and the Weser Kurier, January 15, 2004.

2. SPIEGEL ONLINE, “Rheinbrücke mit Stufe,” January 14, 2004

3. Hochrheinbrücke Laufenburg, German Wikipedia

4. Vertical datum background: Amsterdam Ordnance Datum (NAP), Marseille tidal gauge, Repère Pierre du Niton, Lake Geneva. See the Normaal Amsterdams Peil archive and the vertical datum entry on Wikipedia.