The recent accident on the Washington Metro, Washington, D.C.'s subway system, is an all-too-rare and all-too-tragic black mark on public transit safety. News is still coming in fast and furious, and some of the details have changed significantly in the time since I started assembling notes for this post on Monday night. What we know for sure is that one Metro train rammed into another with enough force to stack cars like fallen dominoes, and that at least nine people are dead and eighty more injured. It's an accident that has such scope that people can only step back and be stunned by its scale - a train wreck in the clearest sense of the term.
This isn't the first time there's been such an accident on the Washington Metro - an accident involving a derailment in January 1982 killed three people and injured twenty-five - but it is the most deadly. Speculations as to the cause have been many, and it'll be some time before a consensus is reached as to what happened and why. For now, in the interests of similarity and contrast, I thought it might be instructive to compare this crash with the 1995 Russell Hill subway accident. That was the deadliest accident in the Toronto Transit Commission's history, and I hope that record stands for a long, long time.
On the face of it, the two accidents are broadly similar - one train, moving at standard subway speed, plowed into the back of a stationary train. In both cases the worst damage was in the front car of the colliding train, though Russell Hill was less deadly in this regard; the operator of the colliding train at Russell Hill survived to give his account of what led up to the accident, while the operator of the colliding Washington Metro train is among the deceased.
From there, the two situations continue to differ. The Russell Hill accident was deep underground, between St. Clair West and Dupont stations, which at 1.2 kilometers apart are among the more widely separated stations in the system, and in an area where the tunnel was not arrow-straight. Between Fort Totten and Takoma stations, where the Washington accident took place, the separation is even wider - 2.988 kilometers (1.85 miles) according to the Gmaps Pedometer, give or take a few meters (feet). The difference here, though, is that no portion of the track between Fort Totten and Takoma is underground. Viewed from Google Maps, it more closely resembles the insulated right-of-way the Toronto subway enjoys within the Allen Expressway, but without a highway on either side.
Weather wasn't a factor either. The accident occured at 5:03 PM ET; Weather Underground's records state that at 4:52 PM, a mere eleven minutes earlier, conditions in the Federal City were mostly cloudy and that no precipitation had fallen during the day. Overcast skies can make for depressing vistas, but they're not nearly enough to hide something as large as a subway train until the last minute.
The issue of operator control is one that's been back and forth. At Russell Hill, the operator of the colliding train threw on the brakes as soon as he saw the stopped train ahead of him. According to a Washington Post story I read on Tuesday morning, "it appeared that the operator of the train that crashed did not apply the emergency brakes," but on Tuesday evening a second story cited early results of an NTSB investigation that indicated the opposite. In any case, there was not nearly enough time to prevent the collision once the brakes were applied, either in Washington or Toronto.
In Toronto, the accident was chalked up to human error and design failure - the operator misread his signals, and the train's emergency brake equipment wasn't up to the task. In Washington, it's something different. Not only do we have likely human error, but computer error as well.
It wasn't until I started putting this post together that I learned the Washington Metro is one of a number of heavy rail systems that uses Automatic Train Operation (ATO). What this means is that the operator is, effectively, a spare part, and while the train is under ATO they operate mainly the doors. On the TTC it's completely different - subways have a crew of two, with an operator in the front car driving the thing and a guard in a middle car who's responsible for opening and closing the doors.
Trains on the Washington Metro are fully capable of driving themselves, and at the time of the accident, that's exactly what the colliding train was doing. It's not the first time computer-controlled trains have got into trouble in the District of Columbia, but never this badly. In 2005, two Metro operators had to override their computers to avoid a crash. This time - not so fortunate.
What's this all mean for Toronto? Nothing, objectively - the Toronto subway is still manually controlled, and in the aftermath of the Russell Hill accident, the signalling system was modified to, hopefully, prevent a similar collision. In the fourteen years since there's been nothing to even remotely compare to it in this system. Still, there are potential issues.
One of the present blue-sky projects for Toronto subway expansion is an extension of the Yonge subway line from its current terminus at Finch station to Richmond Hill Centre, a major transit hub in southern York Region. The problem with this is that it's expected to draw significant ridership to a corridor that's already stressed well before it reaches downtown. Automatic Train Control has been touted as a possible solution, one that would enable separation between trains to be decreased to as little as ninety seconds. Sure, it would increase the capacity of the line, but it would decrease the margin for error.
Note that Automatic Train Control doesn't necessarily mean a conversion to driverless operation as on the Washington Metro - from what I understand, it's primarily an improved signalling system. Nevertheless, three-quarters of any success is in the name, and Automatic Train Control conjures up visions of some kind of robot at the controls.
Pure driverless operation may be ideal one day, once the system has been further refined. For now, though, it seems to me it might be better in the end to leave people at the controls and space things out. They can build more lines if they have to - I'd much rather have the opportunity to ride a Downtown Relief Line than never have to wait more than a minute and a half on the platform in rush hour.
I used to think an interesting concept for traffic control in a future city would involve a central coordinating mainframe that kept all the cars in order on their guideways, but accidents like this really make one stop and consider if that's the best way after all.