Slashdot Mirror
A Cheap, Ubiquitous Earthquake Warning System
Tekla Perry writes: Earthquake alert systems that give a 10 or 20 second warning of an impending temblor, enabling automatic systems to shut down and people to take cover, are hugely expensive to build and operate. (One estimate is $38.3 milllion for equipment to span California, and another $16.1 million annually to operate.) But a Palo Alto entrepreneur thinks he's got a way to sense earthquakes and provide alerts far more cheaply and with much greater resolution. And he's got money from the National Science Foundation to begin the first test of his system — covering the Bay Area from Santa Cruz to Napa and the cities of Hollister, Coalinga, and Parkfield. He starts that test next month.
Hmm... let me think... $38.3M one time and $16.1M/year in maintenance does sound like pocket cash to me, given the population at risk here...
Once again, life imitates art.
https://xkcd.com/723/
Learning HOW to think is more important than learning WHAT to think.
The gov't should convince insurance companies to band together and pony up the cash.
"I don't know, therefore Aliens" Wafflebox1
What are the chances that you could get people to sign up for some kind of app, SETI@Home-like, that frequency-analyses the accelerometer in a phone tied with the GPS-location without any extra fancy hardware?
Done en-masse, FFT'ing to a graph of an interesting frequency range, talking back to a cloud server, surely you could spot a pattern even through the noise of every single movement of every phone in order to detect a consistent, regional variation in a certain, shared, frequency range?
Surely, if you just have enough people signed up to the app, you can not only detect an earthquake (whether you can detect it early enough to do anything is still an open question, really - predicting earthquakes is little more than voodoo, and it's only physical movement of the earth itself that we can actually detect and report on!) but you could also use the app to alert those same people as it happens?
And they give more than 10 seconds warning. My Jack Russell cross (with Fox Terrier) had distinctive behavior that announced earthquakes. A minute - a minute and a half before quakes she'd act like it was a bad thunderstorm (try and hide under me). About 20 seconds before hand she'd start barking furiously with a mohawk-type ridge of hair standing up along her spine and try and drag me outside, once outside she'd go back to trying to hide under me. Others have reported the same reaction with Jack Russells
It took a while before we associated the behavior with earthquakes that were often too small or distant for us to notice.
Not all dogs will reliably detect earthquakes but Jack Russells seem to be very sensitive (they can't stand to be near wood fires or in the same room as an audio recording of one either) - possibly because either/or they are a "below ground dog" (love going down burrows); are "ratters" (have the hearing to listen to rodents).
The gov't should convince insurance companies to band together and pony up the cash.
Insurance companies won't be interested. This doesn't save expensive buildings. It gives people a chance to get out of buildings just in time. For an insurance, dead people are cheap.
This funding will complete the purchase of a Doppler Radar system for Southwest Washington and provide for the land and installation costs associated with the system.
The cost of the "expensive" earthquake early warning system is around the cost of 5 Doppler systems. As of 2013 the National Weather service has access to 159 Doppler installations.
Note that the national radar network is being upgraded to high end Doppler for tornado and severe storm detection. So why do those in the Midwest, Gulf Coast and East Coast deserve early warning on tornadoes and California gets peanuts ($5,000,000) for the inevitable large earthquake? Politics.
So they can afford this in Mongolia and it's too much for California? Really?
Why is Snark Required?
(Caution: I read the article.)
Sounds like a pretty good idea, all-told. An engineer does good with his PhD thesis, starting a non-profit company to create inexpensive MEMS-based earthquake sensors that use the cellular network for communication. Makes them cheap enough that he can deploy them all over the place. But who pays for upkeep? Who pays for electricity?
Here, we get to the problem: he depends on the kindness of strangers to bolt these small devices to their wall and plug them in -- permanently -- to an available outlet. Why would sufficiently many people do that? And since the dwelling turnover in California is so high (at least compared to the other cities I've lived in, CA residents seem to switch apartments at a furious pace), what's the plan for transferring ownership / upkeep agreements? WIth tens of thousands of sensors, that sounds like an ongoing, permanent customer service management nightmare.
Don't get me wrong, the idea's a good one. It might be easier to convince people to download an app that looks for tell tale acceleration signatures of a quake. Cell phones already have location information and the owners are already motivated for other reasons to keep them charged and maintained. The potential downside is that the data quality is likely much lower since cell phones aren't rigidly attached to terra firma.
But that, then, suggests perhaps a dual layer system that includes some company-maintained (he's running a business, after-all) sensors, say installed in a less dense mesh on telephone poles or street lights where they have ready access to (a) rigid fixation, and (b) electrical power, and, importantly, (c) won't be screwed with by the dog / kid / furniture mover. Moreover, upright structures with high aspect ratios, like streetlights, likely amplify ground movement, making detection that much easier. Use that streetlight network for coarse sampling, and the voluntarily downloaded apps as lower-grade, spatially denser sampling. And then, as Randall Munroe suggests in XKCD, monitor the twitterverse for earthquake terms. The apps have next to zero running costs, perhaps only sporadic development and a download server somewhere, the mesh network installation costs can be split between local municipalities, the state, and the NSF, with a maintenance contract to the company from the state. Heck, I'm starting to talk myself into a good business plan!
But depending on the kindness of strangers to install and maintain a thing in their house? Not such a good idea.
Put my fist through my alarm clock with its ding-dong death inside my ear. - The Blackjacks.
Depends on the level of mnitoring a day. One ping a day, and inbound alerts on "quake detected"? A PIII on ADSL would probably handle that!
Or, of course, you could give the contract to EDS, and pay $38B.
Sent from my ASR33 using ASCII
The gov't should convince insurance companies to band together and pony up the cash.
This reminds me of an old Monty Python joke. When asked about tax policies, one bowler hat guy quips: "I think we should tax foreigners, living abroad!"
In any democracy, one thing is certain: A bunch of folks think that a bunch of other folks should pay for something all of them need.
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
True, but then, they wouldn't need convincing, would they?
Just another day in Paradise
I should think that something like an earthquake - a regular, powerful, but maybe overall-small contribution, to movement that's visible in frequency data from a range of devices in a geographical area (i.e. averaging out the noise from all devices to leave behind that which is only common to them all) would show up.
I might be vastly wrong here, but even a few thousand devices reporting a set of FFT data of a certain frequency range (which range would take experimentation but I guess existing research would be able to point the way quite quickly), averaged out with nearby neighbours, and then compared geographically should be able to avoid any random noise and provide enough info to know something is up. In the same way that astrophotography often uses image-stacking - take 1000 blurry photographs, center them, overlay them, average them out (so they each only contribute 1000th of the signal for each pixel) and you can get some pin-sharp detail of what's actually there in the images.
A 1000 people running the app in Silicon Valley should be enough to average out "in your pocket bounces", "car vibration", etc. to provide just the background movement that's apparent in all of them.
Tom Geller