NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/chile/chilel/1499

Internet Message Format | 1993-01-27 | 6.8 KB

Path: sparky!uchdcc!@uchdcc.dcc.uchile.cl
Return-Path: <@cecux1.cec.uchile.cl,@uchcecvm.cec.uchile.cl:[e-mail]>
Message-ID: <[e-mail]>
Sender: Discussion regarding Chile <[e-mail]>
From: "Juan Carlos Barroux - Technical Consultant - Latin America"
              <[e-mail]>
Subject: Re: Earthquake Alarm?
Original-To: Multiple recipients of list CHILE-L <[e-mail]>
Newsgroups: chile.chile-l
Distribution: chile
Sender: [e-mail]
Approved: [e-mail]
Date: Wed, 27 Jan 1993 14:07:16 CST

  Estimados CHILE-Len{o,a}s,

  Adjunto encontraran la descripcion teorica de un sistema de alarmas
para terremotos que tendria una excelente aplicacion practica en Chile.

  A mas de alguien le podria interesar esto en Chile, digo yo....

  Muchos saludos a todit{o,a}s.

  Frase del Dia Patron:

  "I submit to you that if a man hasn't discovered somthing he will die
  for, he isn't fit to live."

  Martin Luther King 1929-1968 - Speech in Detroit, 23 June 1963

----- Begin Included Message -----

>From barroux Tue Jan 19 14:06:10 1993
From: barroux (Juan Carlos Barroux - Technical Consultant - Latin America)
To: me
Subject: Re: Earthquake Alarm?
Content-Length: 5651


In article [e-mail], [e-mail] (Richard Stead) writes:
>In article <[e-mail]>, [e-mail] (Jeff
 Bandle) writes:
>> quake hit.  The seismologist said that he was walking into a safe area
 because
>> their earthquake alarm had gone off a few seconds before the shaking.
>>
>> Does anybody have any details on what this earthquake alarm is and is it
>> possible to get one for a home?
>
>Short answer is no.  But I'll explain how such a system operates, and then
>explain how I would run a real early warning system.
>
>It is probably like to Caltech's (which has been running for years).
 Basically,
>it is an alarm connected to the computer responsible for automated detection.
>The way Caltech's works is that more than 200 stations throughout S CA are
>telemetered realtime to the seismo lab there.  They are then digitized and
>streamed into a computer.  An algorithm called a "detector" is then run
>on the data.  There are a wide variety of detectors, the most common is
>"STA/LTA" (short-term average to long-term average ratio).  In this algorithm,
>the signal for each station is run through several bandpass filters and
>then each bandpass has STA and LTA computed continuously, when the ratio
>exceeds a certain minimum, a detection is registered.  There is some post
>processing to verify the detection, compute onset time, etc.  At that point
>the station has a detection at a certain time with a known amplitude.
>The Caltech method then waits for a certain minimum number of stations
>in a "subnet" group of stations to "trigger" within a limited time window.
>This is then called a subnet trigger.  A crude location of the event is
>also made available, so the magnitude becomes known.  To trigger the alarm,
>certain pre-configured subnets must all trigger at their minimum magnitudes.
>The alarm is then triggered.  At Caltech, the alarm also dials the beepers
>of key lab personel (to summon them to the lab for work), and also rings
>at Caltech security (where they have a checklist they follow when this alarm
>rings).  It is very hard for this system to get a false alarm, but it often
>misses big events.  The alarm itself was not designed to be ideal - it was
>more or less attached to the system as an afterthought (the system is
>required to perform well - it is responsible for the initial solutions and
>setting aside the data the analysts will use to get the final locations
>and magnitudes).
>
>From time to time, the "early warning" question appears in this group.
>I have long claimed this is a solved problem, but somewhat expensive to
>implement.  I think to get it implemented, the state really only has to
>contract with a team of companies that would include some seismological
>expertise, but mostly communications and instrumentation.  It would
>require a seismologist in charge who recognixes that this would not be
>about collecting interesting data or about research.  Most seismologists
>would have an interest in using such a project to do seismology, and it
>is not about that.  What needs to be done is to design a very simple,
>very inexpensive instrument package that could provide simply time,
>amplitude and period each time its detection requirements are met.
>Some means of sending this detection information to a central site
>would be required.  It would not be much information - could be up
>to 10 Kbytes a day.  But it would have to be realtime.  This is why
>communications may be the most important aspect.  In this design, there
>would be 1000's of these detectors carpeting CA.  Once at the central
>site, a single workstation would be sufficient to recognize events in
>realtime, get a "good enough" location and a pretty good magnitude,
>then send it to the broadcast system.  The broadcast system would be
>responsible for sending the notice of location and magnitude to all of
>CA.  Receivers for this signal would be design to be inexpensive,
>and would have a simple capability to have their location input (or know
>it from GPS), and compute the distance between the announced location
>and the receiver location, then compute the expected intensity at the
>receiver and sound an alarm if this exceeds some minimum.  Fancier ones
>could provide closed circuits at a variety of minimum intensities for
>automated shutdowns, switchovers to generators, etc.  I would estimate
>that 3 seconds of signal would be required at each of the three nearest
>stations, 2 seconds of propagation time (for the quake to reach each of
>3 nearest stations), 2 seconds of instrument and communication overhead,
>4 seconds of processing at the central site, 2 seconds communication
>overhead on the broadcast and 1 second processing by the receiver.
>That's 14 seconds from the start of the event until you hear the alarm.
>The damaging shaking will have only gone about 5 km in that time.
>Of course, these are optimistic numbers, I suspect that damaging shaking
>might get 2 or 3 times that far before someone could receive and react to
>the alarm.  Nevertheless, in a big quake, the fault would still be
>breaking at this time.  The system would issue updates as the earthquake
>grows in size, triggering more receivers.  It would work, it would likely
>have few false alarms (and if a receiver waited for confirmation from
>later updates, the false alarm rate could be adjusted at the receiver),
>and would get every real event.  The communication would cost a fortune,
>but could probably be offset by a "subscription" cost for the broadcast
>service.
>
>
>--
>Richard Stead
>Center for Seismic Studies
>Arlington, VA
>[e-mail]





----- End Included Message -----