It’s 7:45 a.m. I’m bolting down the escalator at the subway station at Wilshire and Vermont, and I see a screen that says, “Union Sta <1”. I have less than 1 minute before the train leaves without me, so I keep running, desperately. Miraculously, the estimate is right—this time. In my experience, the countdown clocks are frequently wrong, if not broken.
The times displayed on these digital countdown screens, found on most Metro rail station’s platforms, once determined just how sweaty I’d be when I arrived at work. I’ve since relocated to Boston, but for the last four years, every Monday through Friday, I counted myself among the roughly 7.9 million people taking Metro rail during the work week, commuting from Koreatown to East Pasadena.
Living without a car in LA, I relied on Metro, and that reliance eventually blossomed into a begrudging love. But also in that time, I developed a grating pet peeve that threatened to sour my Metromance: The reliability of Metro’s estimated arrival times.
It’s hard to trust public transportation for your everyday commute when buses run late or when you don’t know when the next train will arrive.
To-the-minute accuracy matters when it comes to catching your last connection and punching in on time. Faulty estimations have screwed up my schedule and weakened my trust, and pride, in the system.
A 2016 analysis by KPCC found that since 2010, just 1 percent of Metro stops were made late. Its definition of “late,” however, is generous; it gave trains a five-minute grace period. Five minutes can make or break a transfer to less frequent lines. While trains are generally on time, the countdown clocks aren’t reliable.
I hurtle off the escalator onto the Wilshire/Vermont platform and squeeze into a packed car, just as the doors are closing. My sweaty pits were worth it: the estimation was correct. The train lurches forward, and, underneath our feet, circuits embedded in certain sections of the tracks are sending signals back to Metro about our location. That information is used to generate the next ETA displayed on countdown screens.
If our train were to stall on tracks not equipped with circuits, the car wouldn’t be able to update Metro on its position, so people waiting on the next platform would see an estimate countdown all the way to arrival, and then start back up again—with no train arriving.
After taking transfers at 7th Street/Metro station, the train chugs northbound uninterrupted, pinging Metro roughly every 20 seconds to say, “A car is here!” At Union Station, I immediately hone in on the countdown screen to see how many minutes I have to make my Gold Line connection. I have six minutes—just enough time to duck in for a coffee.
How is Metro generating these estimates, and why are are the countdown clocks so inaccurate so often?
For every minute I’m waiting in line at Starbucks, nervously counting down those six minutes, Metro’s Supervisory Control and Data Acquisition system is gathering raw data from the train track circuits and packaging it for crunching by a contractor called NextBus. It gets updates from the data center every two seconds and sends an updated estimate to the countdown displays every minute.
It’s NextBus that predicts the minutes to arrival by comparing it to Metro’s set train schedule. The Emeryville-based company is the lynchpin for not only Metro’s countdown displays, but also Google Maps and the GoMetro app, among others. If you’re relying on a real-time transit planner, NextBus is probably involved behind the scenes.
Unluckily for me, Metro allows a 1 to 2 minute window around a predicted time to still consider it accurate, so when I arrive at the platform with one minute left on the countdown, I see my train already leaving the platform.
The system’s biggest flaw is that it only knows where a train is if that train is moving.
Al Martinez, senior IT director for Metro, says the major cause of inaccurate predictions is service disruptions, anything from malfunctioning doors at MacArthur Park to police activity at Hollywood/Highland to a traffic accident in South Pasadena. If a train car is sitting still on a section of track without the embedded location-tracking circuits, it effectively becomes a ghost train. There's no way to see its exact location. Depending on how drastic the disruption is, Metro will choose whether or not to disable the ETA predictions entirely.
The embedded circuit method was implemented mainly to keep cars far enough apart so they won’t collide. The same method is found in many other U.S. rail transit systems, including New York City’s, that were built prior to GPS or autonomous technology. They simply weren’t designed “particularly for accurate and precise tracking of train position,” says Martinez.
Another technological hurdle is the speed of data transfer. The frequency of updated location info, and how fast Metro and NextBus can crunch their numbers—not too mention how fast Metro staff can communicate service disruption notices—can fudge estimates.
Sometimes, coming home after a long day, I’d grimace at a cryptic “804” on the countdown display. Martinez says that error message typically requires a system reboot to resolve.
If the system gets overwhelmed for any reason, it can stop communicating entirely. “Once identified, the system generally needs to be reinitialized to clear the error and return to normal working order,” he says. It’s basically a transit version of an old school TV test pattern.
To Martinez, the ideal solution would be to install a GPS-based tracking system for Metro's train cars that would update every five seconds (similar to the system already in use on buses), but even with Measure M’s windfall, no money has been set aside for that.
The ideal method for predicting arrivals seems to lie in driverless train systems.
In Copenhagen, where departures were on time 99.2 percent of the time last year, the rail system is automated. It bundles both the speed controls and location-tracking technologies under one hood. This ensures that the trains stay a safe distance from one another and keep a consistent speed.
While I now take Boston’s T more often than Metro rail, as a commuter in any city, I’ll just have to accept that there will always be service disruptions. But I hope that for LA commuters, knowing how the ETA-sausage is made will (at the very least) help plan their trip with a grain of salt.