Two years ago I wrote a piece about the frequency and duration of Link service interruptions (blockages, accidents, power outages, etc). Given the recent, high-profile service meltdown on June 12, it makes sense to revisit the issue. How often do interruptions occur? How long do they last? Have they become more frequent or longer in duration over the last two years? I parsed several hundred email alerts from Sound Transit to find out!
I measured interruptions starting in December 2015, when I started subscribing to the Sound Transit rider alert emails. From then up to June 28, 2019, I count 144 separate incidents*. That’s roughly one interruption every nine days and is the same rate I estimated in my 2017 piece. Counting the number of interruptions in 30 day intervals confirms that the base rate of these incidents has been quite stable, in a statistical sense.
So, the data do not suggest that interruptions have become more frequent. But have interruptions become more severe in duration? As in 2017, I summarized the length of interruptions in a chart that classifies incidents in four categories (quartile groups) of duration severity.
The level-4 events are usually the memorable ones and tend to involve major accidents or power outages. For instance, the longest duration event I found was due to a 6-hour long power outage somewhere between Othello and Tukwila on August 8, 2017. Our recent bad day shows up in the red on June 12, with a nearly 4-hour disruption. These durations are calculated from when Sound Transit sends out their “service interruption” and “service resumed” emails – so, the data here depend a lot on the quality of ST’s reporting to the public. That said, there have been more level-3+ incidents from January to June in 2019 than in other years. On the other hand, the 2016 count isn’t far behind. And the median duration has remained at about 25 minutes, the same as when I last calculated it in 2017.
Have Link service interruptions really gotten worse? The data I have suggest not so much – at least at a high-level. But that is small comfort when you’re trapped on a stuck train or waiting at the platform during one of these incidents.
[*] Incidents are counted from ST alert emails, so my data could under-count the real number of service interruptions.