Earlier this week a hurricane-like storm system hit the Pacific Northwest, bringing severe winds and precipitation across the region. The first and strongest round of the windstorm hit on Tuesday evening and a second, weaker bout is expected tonight.
With fallen trees comes downed wires and blocked roads, and the storm this week has left hundreds of thousands of households without power for days. In addition to damaging overhead wires, falling trees pose a risk to anything in their path, including people, homes, buses, trains, and cars.
Headlines:
- 2 killed, hundreds of thousands without power in Western Washington amid pacific bomb cyclone winds (KUOW)
- What the Western WA windstorm means for buses, light rail and ferries (Seattle Times; $)
- Tree falls on a King County Metro bus in North Seattle (Seattle Times; $)
- Amtrak engineer nearly impaled by tree hitting train (Seattle Times; $)
- Powerless EV drivers storm Seattle-area grocery store to recharge during outages (KUOW)
- Photos: Seattle area continues to recover from widespread power outages and property damage (Seattle Times; $)
Community Resources
If you know anyone in need during this severe weather event, please share these resources for finding support and warmth on these cold, dark nights:
How to find warmth, electricity, and other storm resources in the Seattle area (KUOW)
Seattle City Light announced this afternoon the City of Seattle and the Red Cross will open an overnight warming shelter at the Garfield Community Center in Seattle.
- The Garfield Community Center is located at 2323 E Cherry St, Seattle, WA 98122.
- The warming shelter opens at 7:00 p.m.
Outage Maps:
Service Alerts:
If you see a downed wire or fallen tree posing a danger to the public, stay clear of the hazard and call 911.
This is a semi-open thread; acceptable comment topics include windstorm-related observations, transit’s role in emergency response, and similar topics.
I road the bus yesterday from Kirkland to Crossroads and it took about an hour. Much of it was a long backup stemming from the signal being out at Northup Way. But, it was exacerbated by the “don’t block the box” rule. Essentially, whenever there’s enough space after an intersection to fit a car, but not a bus, the bus has to stay out while cars from the other lane keep cutting in front of it. The result is the bus being stuck at the intersection for much longer than a car would be in the same traffic.
The experience has me wondering if transit buses (at least those in active service) should be outright exempt from the “don’t block the box” rules so that the bus is able to have some semblance of motion. Especially since any delay impacts not only people on the bus, but also anyone who will board the bus at any later stop along the entire route.
The experience also has me wondering why in 2024, traffic signals are not equipped with battery backup, at least those at busy intersections. How big would a battery actually need to be to power a traffic light for a few days?
Especially now that they are all LED. Excellent point asdf2.
It’s absolutely ridiculous traffic signals don’t have battery backup. Railroad signals have had these for well over 100 years (and I have photos of 80 year old Edison glass encased batteries still being used to power a signal control
center in the 1990s).
Looking online the typical backup batteries can last 8 hours if just flashing red or only 2 hours of normal operation.
https://svcleanenergy.org/news/backup-batteries-at-traffic-lights-increase-safety-and-energy-resilience-in-saratoga/
It looks like the cost of the battery itself isn’t too much but the installation can be kinda expensive. Checking the psrc sheet they quoted a backup one as 8k. Seattle construction crews though is probably what’s expensive. Idk I’d assume at least 40k in labor for closing down the intersection, shutting off the power etc…? (If anyone knows better feel free to correct here)
So ~1,000 intersections for around ~50k costs around 50 million dollars
Honestly a bit cheaper than I expected. This does have the caveat that there needs to be enough room in the existing traffic control box, if not it’ll probably cost a lot more.
Saratoga paid $182k for 14 intersections ($13k per intersection) and it looks like all they did was bolt a bunch of marine lead-acid batteries to the top of the existing control boxes. I would expect something like $15-20k per intersection to do a similar job in Seattle, with higher costs associated with larger intersections featuring more signals or more complicated control boxes. For our 1,100 signalized intersections, something like $22M? Maybe fluff that to $25M, throw a 20% contingency on there, call it $30M? That’ll be closer to $50M by the time the next levy rolls around, but seems like a worthwhile project. Or, have it small addition onto an intelligent signals upgrade/maintenance program.
And when the power is back on at traffic lights… kind of amazed with where the technology is today (but hopefully in the future with AI) that traffic lights can’t better detect the bus that runs every 20 mins approaching and give it the light or in very suburban locations detect the rare pedestrian or bike and give it the light. There’s supposed to be all this advanced signal technology yet still does the long conventional cycles and analog timing (i.e. for green wave) like its done for generations and doesnt detect the different street users. It would be nice to see AI enable traffic lights to function like an automated traffic cop that can use their brain and read the traffic at the intersection and hold lights if need be, let an occasional ped cross without having to wait through 8 cycle phases, give a bus priority, etc.
@ Poncho:
I don’t think AI can predict pedestrian behavior. The biggest constraint is not so much what cars and buses are doing — but what the pedestrians are doing. For example, a wide street like MLK or Aurora takes a pedestrian countdown about 45 seconds.
AI can’t guess if someone on a street corner wants to cross until they punch a crossing button — and in many cases the best thing a signal can assume is just to expect that pedestrians want to cross every few minutes or at every signal cycle.
Even if AI can guess is an approaching pedestrian is wanting or use a crosswalk versus merely just walk along the sidewalk, it’s still a guess. Plus pedestrians may choose to cross at a different corner because of things like avoiding a panhandler or an e-scooter in the way or a debris-filled or flooded sidewalk blocking the preferred path.
AI logic is powerful. It has power when it looks at moving vehicles only — which is how many signal systems work including those that anticipate moving buses. If there are no pedestrians, signals can give priority to buses with just a few seconds advance notice. However, if we want to encourage more people on foot, we can’t look to AI to be as anticipatory.
Some suggest grade-separated pedestrian crossings, however that adds a vertical level change effort that many dislike — along with safety concerns when using a pedestrian tunnel or noise concerns when using a pedestrian bridge.
What to me seems to work better is to create less-wide one-way streets a block apart — rather than a really wide street that both encourages higher speed movement and requires more time to walk across. Signals on one-way streets also have fewer turn phases so pedestrians have less time to wait and can get across faster.
It works Downtown and has for decades. Think about how awful walking Downtown would be if there was a 5 or 7 lane two-way street running down the middle. It’s what happens on Alaskan Way on the waterfront or to a lesser extent on Jackson — and the result is more pedestrian frustration and crossing concerns.
Certainly it’s really costly and disruptive to retrofit a street like Aurora or MLK into two one-way streets with a block in between the directions. So we are left with many miles of pedestrian-hostile overly-wide streets and highways. And long waits at crosswalk corners.
Bus signal-priority technology already exists — no AI needed. The bus just sends a ping to a transponder to show it’s there, and the scheduling system can give it the appropriate weight. It’s just a matter of Bellevue approving it and installing it. Approval is the biggest issue: an SOV movement may arise in opposition.
Re predicting pedestrians, perhaps the technology can be adapted from the Happy Vertical People Transporters in the Hitchhiker’s Guide to the Galaxy. They use precognition to tell when somebody will want to go up or down in the next minute. The main issue would be that elevators are grade-separated so the only factors are prospective passengers and current passengers. With a bus in mixed traffic at an intersection a lot of other factors come into play, so a bus may not be able to be as responsive as an elevator.
https://hitchhikers.fandom.com/wiki/Happy_Vertical_People_Transporters
https://www.goodreads.com/quotes/685736-modern-elevators-are-strange-and-complex-entities-the-ancient-electric (dismiss the login dialog)
https://hitchhikersguidetoearth.fandom.com/wiki/Sirius_Cybernetics_Corporation
@poncho
Intelligent traffic signals are already available and fairly widely deployed. Many of the signals in the city still need to be replaced/upgraded. The transportation levy included a bunch of money for that, and most major projects include signal upgrades (see: 48 TPMC or RapidRide J)
Bellevue is currently testing out pedestrian detection (links below). Not sure what their goal is.
https://bellevuewa.gov/sites/default/files/media/pdf_document/2024/Bellevue%20RTSSI_SMART%20%20FY23.pdf
https://bellevuewa.gov/sites/default/files/media/pdf_document/2024/Passive%20Pedestrian%20Detection%20Real-Time%20Safety%20Application%20Phase%20Extension%20Pilot.pdf
@Al, I do think that AI and video technology will change traffic light controls. Yes, it won’t get it always right, but for the most part AI can detect intend to cross a street, such as in https://www.autoevolution.com/news/smart-pedestrian-crossing-uses-cameras-and-ai-to-control-traffic-lights-160006.html
People in a bike lane are even easier to detect. Greg Spotts (SDOT) told me that the city is switching from RF bike loops to IR cameras to detect approaching bikes. I very much appreciate this as the loops only work for metal bike frames, not carbon frames.
Will just leave this here for those interested in AI. Pedestrian behavior predictions included.
https://en.derq.com/derq-sense
You don’t actually need ai. Some European countries have had pedestrian detection cameras for a decade now. Granted they are used everywhere typically more for downtown areas
They are not* used everywhere
AI should be able detect whether a pedestrian is waiting at a location where they likely want to cross, versus just walking along. But I wonder how well it can “learn” what is “best” to maintain the overall traffic flow. There is not going to be enough training data where pedestrians were actually given various amounts of priority (eg wait time before the signal is preempted), because cars are prioritized a priori as a matter of policy. Maybe you have a dedicated experimental training period during which you expect some traffic disruptions?
My point is with technology it can customize the signals for the unique current situation. So often lights make the occassional pedestrian wait needlessly, the motor vehicle traffic is going to have to be stopped to let the pedestrian cross one way or another, why not give the pedestrian the light almost instantaneously? This is especially the case where pedestrians are relatively rare, say your stroad, like Intl Blvd by the airport. It could even actually provide a red light for turning vehicles when that rare ped has the ROW.
I know people say there’s bus priority at lights but I’m very skeptical that there really is this bus priority at signals even on lines like the G. Seems more like timed lights timed to favor the bus, the G always stops at Boren and many other stops. What I am proposing would see the bus come and hold or give the bus the light as it approaches by seeing the bus or learning its typical dwell time at a nearby stop (and even likelihood of stopping at the stop).
What I’m getting at is AI can learn the patterns of buses, bikes and pedestrians which are typically not as frequent as car traffic. There’s always a neverending flood of cars no matter what. This technology would allow other non-motor vehicle modes that are much less frequent on the street to navigate the constant flood of cars with minimal impact. If the bus comes every 5 minutes let the car traffic go uninterrupted and then when that bus comes it gets priority. Let the car traffic have priority then when that rare cyclist or pedestrian comes give them a little more priority and safety with signals.
“For our 1,100 signalized intersections, something like $22M? Maybe fluff that to $25M, throw a 20% contingency on there, call it $30M?”
• If you calculated out what the long term costs are for delays and injuries from dead traffic lights, it’d be interesting to know if there would wind up being a net financial gain or expense.
• 1,100 traffic signals ≈70 watts each for lights + the controller x 365 days a year ≈ 674,500 kwh per year, or ≈ $81,000 per year in power costs. Seattle City Light charges *half* the peak rate for non-peak consumption, so they could probably save some $40,000 per year by going with a system that runs off the batteries during peak consumption times, except when inclement weather is forecast.
@Poncho, Siemens etc is using AI to adjust traffic light phases based on current and predicted traffic load.
https://blog.siemens.com/2021/11/ai-makes-cities-livable-by-forecasting-traffic-light-phases/
The ai part is for better pedestrian recognition for the camera.
> But I wonder how well it can “learn” what is “best” to maintain the overall traffic flow
The actual traffic modeling has never needed ai. ai is for massive amounts of data — for traffic signals it’s relatively simple just counting the number of cars and how many pedestrians. Regular statistics has always been fine. They’ve been able to simulate it decades ago.
> AI to adjust traffic light phases based on current and predicted traffic load.
Siemens and other companies just slap the ai title on for marketing purposes. Many la cities and Bellevue already do this.
> So often lights make the occassional pedestrian wait needlessly
This has been done in Europe a decade ago already.
The hardest part is just actually getting all the signals wired up to some central server. Otherwise they are all working independently.
Throughout king county they are already working initiatives to tie them all together.
If you guys are curious here’s the most recent update
https://www.psrc.org/media/7492
There’s also a separate map that shows all the traffic signals by type of “smart feature enabled” that’ll I’ll try to find.
Generally the Eastside has “smarter “ signals but they focus heavily on cars. Aka if all the cars from north/south direction already went it’ll automatically switch early to east/west.
Seattle on the other hand while some signals have switched to the smarter system a lot of them are on the older system. For instance 2nd and 4th use the older timed system of scheduled north/south “waves”
In any case AI outside of the camera recognition has nothing to do with the traffic modeling. It’s really whether or not they’ve centralized the signalling rather than at each intersections box
Guys no amount of detection is going to erase the human fact that pedestrians need time to cross. The City has to allow enough time for a pedestrian to cross the street. Because these streets are so wide, it’s likely going to still take 50 seconds by the time you add all the seconds involved when having a crossing allowed.
I guess AI could detect that someone has gotten across the street and shave several seconds off the allocated walk time. I don’t see that shaving more than 10-15 seconds unless it’s a jogger crossing the street. And as long as there is a countdown signal I don’t see any way to suddenly jump from 15 seconds to 0 without risking a lawsuit from any pedestrian injury that happens.
@Al
> Guys no amount of detection is going to erase the human fact that pedestrians need time to cross.
That is not what we are talking about.
How it works is the system automatically detects using cameras whether a pedestrian is there or not. For a downtown area where there is always a pedestrian crossing the street, the pedestrian detection system is effectively useless as the road will always need to trigger to let the pedestrian cross and need to wait the full amount.
It’s more useful for say Bellevue or Renton or even Seattle at low traffic intersections. It can automatically detect when a pedestrian is coming close to the intersection and then add the pedestrian cycle into the intersection without them having to press the button.
For more “auto-oriented” cities they can continue to skip adding the pedestrian phase if not pressed. For Bellevue they are targeting Bellevue Way and NE 8th St.
https://bellevuewa.gov/sites/default/files/media/pdf_document/2024/Passive%20Pedestrian%20Detection%20Real-Time%20Safety%20Application%20Phase%20Extension%20Pilot.pdf
When a pedestrian has finished crossing the crosswalk, the signal light will wait longer before turning the green light on for cars.
for the traffic signaling bellevue is using “The city uses an adaptive signal system called Sydney Coordinated Adaptive Traffic System (SCATS). Within the SCATS programming, custom coding was added for passive pedestrian detection”
> The City has to allow enough time for a pedestrian to cross the street.
In many eastside cities, by default if one does not press the pedestrian button, the light cycle does not give one the pedestrian signal. Aka they prioritize the cars and the light cycle might not be long enough for one to cross the road.
One advantage of the automatic pedestrian detection is it can be activated even before one reaches the road. While not really the best reason to install it, it does still still lower the average amount of time for a pedestrian to cross the road.
But as mentioned before there are other reasons to install it, such as lengthening the pedestrian light cycle. Or in other smarter cases can even lengthen it depending on the number of people
That leaves people not knowing whether they have to press the button, so they go out of their way to press it anyway even if they didn’t have to. Or if their hands are full with heavy groceries and they’re too far away from the button, they don’t. Then they don’t know whether the light will turn green or not. Then it does for cars but the walk sign stays red and they don’t know whether the sign is broken or whether it will turn green later, and they may be so frustrated or tired they don’t want to wait another four-way cycle to find out. So they walk across because the light is green, and then either everything is fine (only right-turning cars turn in front of them), or a left-turning car turns in front of them because it’s one of those new signal
arrangements with separate pedestrian, left-turn, and regular phases.
–Life of a pedestrian in a car-oriented world
@Mike, with camera based (and AI supported) pedestrian and bike detection, you don’t need buttons anymore. You don’t really need countdowns anymore either, as the camera can detect that a pedestrian has crossed the road and change the signal. You may need to show a yellow signal so that other pedestrians don’t enter the crossing anymore.
Maybe a solar panel could be installed to extend the life of those batteries if the outage is longer. Maybe overkill, since they would only be needed about once a year. But just a thought.
Yes and yes again. Buses should have the advantage in *any* situation like this, as well as lots more bus-only lanes and traffic light overrides. Ain’t no better way to get people on the bis than by seeing buses roll by while cars are stuck in traffic.
I absolutely agree, I’ve been on a 44 waiting to turn left onto 45th for 30 minutes because the driver was too timid to pull into the intersection despite SOV drivers having no reluctance to do so. Eventually he got on the PA and put it too a vote, answer was a resounding YES PLEASE.
My only concern would be blocking crosswalks, but I think bus drivers in general wouldn’t take much convincing that they should do their best to keep the bus moving but be safe for pedestrians.
Hoping everyone is safe and warm, and those powerless, can recover…
In a previous article, it says RR D Line was the 4th busiest KCM route, behind the E Line and Route 7…
But I just finished reading an article from the Urbanist, it shows the D Lin is actually the third busiest KCM route of the system, not the 4th, just a revision of data (unless you actually know which one is the third busiest… Maybe A, C or H Lines)…
It changes month to month. But as of last month, according to the system dashboard:
E Line: 13,500
7: 11,300
A Line: 9,600
D Line: 9,300 (8,800 in August)
H Line: 8,600
40: 8,500
C Line: 7,500
The D Line is the third in terms of RapidRide, but the fourth in terms of Metro buses (in October).
Wonder what will happen with this Talgo, since they’re planning to retire them for Siemens trainsets anyways. Is there enough rolling stock without it / can they borrow one until the normal ones arrive or will they have to send it in for repairs?
Time for a raiding party back east to
sneak another Horizon set out of the Midwest.
Can’t read the article. Did the trainset derail? And why was a Talgo running? I thought the FRA embargoed them after the Dupont wreck. I did see one a couple of months ago and wondered then.
Here’s the story on KUOW: https://www.kuow.org/stories/amtrak-engineer-on-seattle-bound-train-nearly-impaled-by-tree
Some more details in a Trains Magazine article: https://www.trains.com/trn/news-reviews/news-wire/engineer-injured-equipment-suffers-significant-damage-after-cascades-train-hits-tree/
“The train involved is one of two Talgo Series 8 trainsets owned by the state of Oregon. Older Talgo Series VI trains owned by Amtrak and Washington state were removed from service in 2019 in reaction to findings regarding a fatal 2017 derailment involving one of the trains [see “WSDOT prepares to remove Talgo Series 6 …,” Trains News Wire, May 24, 2019].”
It’ll buff out.
Just run it through the train washer!
Nathan, thanks fir the info and the links I didn’t know that the embargo was just for some Talgos.
Latest rumor from a known foamer, the damaged trainset is being sent back to Wisconsin for repair and the northwest is being sent (at least) a trainset’s worth of Horizon equipment.
Watch for Evergreen, Castilian Copper, and Double Latte (and a splash of Nugget) heading east.
I hope the PNW gets the loaner Horizon equipment post haste!
Don’t want to cancel trains during the busiest day[s] of the year!
It was ridiculous to retire the original Talgo equipment, they weren’t that old… Most train cars aren’t designed to fly off bridges. Now the Cascades are stuck with older, rattier Horizon cars and struggling to get the few spares available in the national fleet.
Jim, probably from Beech Grove just outside Indianapolis. That’s Amtrak’s national repair facility.
The other side to the Talgo rumor is they just send a cab car west.
RENFRE has the manufacturing/maintenance facility in Wisconsin for the trainset the state was going to purchase until Rick Scott canceled the contract, and Nigeria was supposed to get them.
I have no idea what the status of that facility is.
Wrong Scott. Rick Scott is the Medicare-swindling (ajudicated) Senator from Florida. Scott Walker was the governor of Wisconsin.
And Spain’s rail ststem is “RENFE”
My apologies.
Thanks for the corrections, Tom.!
There really needs to be a west coast car repair facility somewhere. The Amtrak facilities (Eg Seattle and California) are ok for maintenance but aren’t really equipped for complicated repairs. The Siemens plant in Sacramento is production and not particularly suited for heavy repairs either.
Alstom’s facility in the Bay Area can do some of this type of work, but mostly that shop is running repairs. More complex sheet metal and wiring work can be a real undertaking for them as they just don’t have enough staff and equipment.
There used to be a couple such places in the west. Eg: Morrison-Knudsen rebuilt passenger cars for the Alaska Railroad. WabTec converted it to locomotive only manufacturing before closing it.