At 08.30 am MET (07.30 am UTC) on Monday, 15 February 2010, a commuter train and an intercity train collided in Buizingen, in the greater Brussels region. Initial reports mentioned a “head on” collision, but De Standaard reported (in Dutch) that one train ran into the side of another, presumably at a set of points. The high-speed lines for the Thalys Cologne/Amsterdam-Brussels-Paris service, as well as for the Eurostar Brussels-London service neighbor the accident track and services have remained suspended during rescue and investigation operations. An spokesperson for Infrabel, the rail infrastructure maintenance organisation, is reported to have estimated that it will take a further 16 hours after the rescuers and investigators have finished and debris has been cleared to free the high-speed lines for use.
This commentary is based on the best reports I have found on the accident: New York Times, 15 February and New York Times, 17 February. There is already a Wikipedia article on the collision with some useful on-line news references.
There are two themes I wish to address here. First, the premature announcement of «cause» by public figures, and its consequences. Second, the involvement of the judiciary and its operatives (police) as primary investigators.
Lodewijk de Witte, the premier of the local region Flemish Brabant, was reported Monday after the crash as saying that it was likely caused by one of the drivers running through a signal at danger, in English terminology a Signal Passed At Danger, or SPAD. Local dignitaries seem under pressure to make such cursory judgements. I recall the Lathen Maglev accident in September 2006, in which Alexander Retemeyer, the state attorney responsible for the investigation, said on the day of the accident that “human error” was the likely cause, by which he meant the operators and controllers on duty at the time, the people at the “pointy end”. Even our local newspaper, the Neue Westfalische Zeitung (NW), had figured out a day later that, rather than it being all “human error”, certain technical configurations had played a causal role. Indeed, as became clear, a lot of them: the service vehicles were not incorporated into the technical track safety system for the Maglev, the usual waiting position for service vehicles was not covered by the otherwise-ubiquitous CCTV cameras, the two boards showing location respectively of service vehicles and Maglev trains were in separate places in the control room, the controllers of the service vehicle and the Maglev train were two different people, even though all vehicles operated on the same single track, and the voice communications systems were separate for service vehicles and Maglev, all of which became evident over the next few days (see my contribution to the 9th Bieleschweig Workshop in 2007, linked below, for details). Many of these factors have also played a role in other rail accidents, for example separate control/communication systems for two trains using the same track played a causal role in the Glenbrook accident in Australia
One of the consequences of de Witte’s premature statement is that train drivers have been striking. The strikes are what the Brits call “wildcat”, that is spontaneous and not organised by the union. 75% of the drivers in the Wallonia region (French-speaking Southern Belgium) are reported to be absent, and in the Flemish region it is “more sporadic”. The strikes began in Leuven, the home town of Johan de Keyser, the driver of the intercity train, who was killed.
State railway statements have been more sober concerning cause. Jochen Goovaerts, the SNCB spokesman, was reported as saying it was too early to know the cause, and “we are not going to speculate about that until the investigation has been completed”.
Maybe this large-scale action by stakeholders will help dampen the tendency towards premature speculation on causes by public figures. I believe that this premature speculation is not helpful for safety. Indeed, it may very well hinder appropriate action on countermeasures. My reasoning to this conclusion is as follows. Case studies show that premature speculation is more likely than not to be misleading; however, the misleading “cause” cited at the time is more likely to remain in the “public mind” than the causal factors which are indentified by the investigation, for these come out months or years later when the incident is no longer “news”; the «instant» but misleading factor is thus more amenable to use in political arguments for specific countermeasures (many countermeasures require political action to implement, and thereby a measure of support from stakeholders such as the travelling public); which in turn may lead to non-optimal countermeasures being preferred.
That all sounds very abstract. Let me make it concrete. Let’s look at a possible political development. De Witte says “driver ran a stop signal”; the public remembers this because it is in the same news from which they learnt about the crash; someone starts a campaign for universal installation of “automatic train protection” (ATP) on Belgian rail; this garners public sympathy as people remember what was said (prematurely) about the cause. Whereas, as far as we know at the moment, it may have been that the signal in question was not showing a red aspect for some reason, or that some points were wrongly set. In neither of those cases would ATP have helped: the appropriate countermeasures for the future would rather involve improvements to the technical reliability and security of signalling systems.
Let me consider the issue of ATP as this debate has played out in Britain.
The various ATP systems work as follows. There are trackside status-transmission devices in the neighborhood of signals and sensors on the trains. When the signal is at danger (red) and a train overruns the signal, the status is read by the sensor on the train and emergency brakes are automatically applied, no matter what the driver does. The German «Indusi» system (for «induktive Signalsicherung», inductive signal protection), based on trackside magnets, has been in operation since 1934.
After the 1999 Ladbroke Grove accident in the neighborhood of London’s Paddington station, which bears some superficial resemblance to Monday’s Buizingen accident, there was a call for universal installation of ATP on British rail services. (At Ladbroke Grove, a local train ran a red signal and ran into the side of a Bristol-London express train. The number of people killed lay in the 30’s. There was a public inquiry, chaired by Lord Cullen, who had previously chaired inquiries into the Piper Alpha North Sea oil platform fire and the Kings Cross underground station fire.) Similarly, in our local paper, the NW, there was a sidebar to the Buizingen accident story on Tuesday morning explaining how «it couldn’t happen here» (in Germany) because of the «Indusi» system.
The two main issues with introducing ATP universally into British rail operations is that it is likely to cause more deaths and injuries than it saves, and that it costs an enormous amount of money to install universally. Maybe that money could better be spent on public safety measures which are likely to save more lives than they cost.
How can this be, that ATP is likely to cost lives in net terms? Trackside workers suffer deaths and injuries which are amenable to statistical treatment. Similarly, statistics are available on deaths and injuries to occupants of trains, and the causes of every accident are investigated. It turns out that the statistics suggest that more trackside workers are likely to be killed or injured during installation of ATP than train occupants are likely to be saved by it. (I have been asked by a colleague for a citation to a definitive statement of this phenomenon, and will provide one in due course.)
Following the 1999 Ladbroke Grove accident, the British Health and Safety Executive began publishing monthly public SPAD reports. This task has been taken over by the Office of Rail Regulation, which publishes quarterly SPAD reports. The Rail Safety and Standards Board publishes monthly public SPAD reports amongst its Safety Performance Reports. I have no idea whether the safety benefits this public reporting brings (for example, quick identification of deleterious trends through many pairs of eyes, many of them volunteer) can be quantified. My intuition tells me that extensive public information on safety is almost always a Good Thing. For one thing, it enables public assessment of countermeasures based on established engineering facts, and not on the whimsical reactions to particular accidents by inexpert public figures.
On to my second theme. The driver of the commuter train in Buizingen saw the accident coming, and retreated into the lead car before the collision after applying the emergency brakes. Apparently he is severely injured, but survived. Here comes another well-known negative factor: he will be interviewed first by the police.
The police and judicial system seem to be the primary investigators. This, also, is not necessarily a good thing, for reasons which are explored in detail using many case studies in my talks to the OFFIS Open Day on transportation safety in Oldenburg and at the 9th Bieleschweig Workshop, in 2007. This is a prominent theme in commercial aviation at the moment, since the public prosecution of various people allegedly involved in the Concorde takeoff accident in 2000 has just started in France, and this is one of a series of recent judicial involvements in aviation accidents that eminent aviation safety organisations such as the Royal Aeronautical Society, the Flight Safety Foundation, the Civil Air Naviation Services Organisation, and the Academie de l’Air et de l’Espace have all deplored. My recitation of their reasons, along with some more case studies of my own, may be found in my somewhat lengthy contribution to the 9th Bieleschweig Workshop in 2007.
The debris at Buizingen is all still in place. The rescuers are not sure they have accounted for all occupants; there is one still missing apparently. After the track debris has been cleared, we then have a further 16 hours for the high-speed lines to be freed. It seems Thalys and Eurostar services will remain suspended until that occurs. This is the second time in two months that Eurostar services from Brussels have been suspended for many days. I am scheduled to use Eurostar again in late March, and in late February the Thalys to Paris. It was pointed out in the article that Brussels is a key node, connecting Cologne, Amsterdam, Paris and London, all within two hours or less of Brussels by high-speed train. Indeed so. This infrastructure connecting four of Europe’s capital cities and one of its most significant non-capital cities is more valuable than anybody has yet seen fit to assess.