Subject: Re: Assessment of predictions (was Earthquake advisory cancelled 11/24/96)
From: gentryd@pipeline.com (Dennis Gentry)
Date: Sun, 08 Dec 1996 16:19:19 -0300
In article <1996120408445918185@seismo.demon.co.uk>,
russ@seismo.demon.co.uk (Russ Evans) wrote:
>Dennis Gentry wrote:
>
>> I'll second that and Pauls' post in conjunction with Alans' post helped me
>> to get a better understanding of the scientific process in determining the
>> validity of a process or identified frequency.
>
>I organised a meeting of those professionally involved in assessment of
>earthquake predictions just last month. It looks like I will be editing
>a volume of proceedings for Geophysical Journal International, but in
>the meantime (if you are not afraid of some serious statistics) you
>might like to cast an eye over the abstracts. Some of them are
>presently to be found on my web site at
> http://www.seismo.demon.co.uk/Nov7th/
Thanks Russ. Actually, as far as serious statistics, I didn't see to
much more than has been discussed on this ng. The abstracts were
very interesting though and enjoyed reading them. At least the ones
that were available.
>We are presently trying to work out how best to deal with the issue of
>archiving such material, and the full suite should become available in
>due course through the RAS site at
> http://www.ras.org.uk/ras/
>
>The meeting was reported in both Nature and New Scientist, but Science
>appear to have ignored us. I believe that the texts are available at
>their respective web sites. A more detailed report will appear in EOS
>(newspaper of the American Geophysical Union) next week, and may appear
>at the AGU web site (http://www.agu.org/, I think). If not, I may seek
>permission to post it here.
>
>The bottom line is that no-one has yet come up with a prediction scheme
>which withstands careful statistical analysis. What's more, there are
>some very strong indications that earthquakes are inherently
>unpredictable, at least in the generally agreed sense of stating when,
>where and how big. Governments should continue to concentrate funding
>on monitoring, hazard assessment and on appropriate mitigation measures.
Couple of things.
One gentleman mentioned that "new mathematical and theoretical tools"
will be needed for earthquake prediction. I wasn't quite clear on what
the author was trying to get across here. Was it in reference to tools
for evaluating earthquake predictions or tools for earthquake predictions?
Also, I missed seeing anything about Alan Jones method for tracking
earthquake predictions (a few abstracts did discuss the null hypothesis
to which I thought Alans' process handled pretty well).
Another thing that was interesting was that several methods at earthquake
prediction were discussed but are being discarded as unreliable due to
one of the big three factors being missing such as location. For those
that were tagged as being unreliable due to location, I didn't see
anything about attempts at triagulation. Didn't see triangulation
discussed at all. Maybe in language that went over my head?
Looks like it was a great meeting though.
Regards,
Dennis
Subject: Re: Assessment of predictions (was Earthquake advisory cancelled 11/24/96)
From: gentryd@pipeline.com (Dennis Gentry)
Date: Sun, 08 Dec 1996 16:19:43 -0300
In article <58772d$7vj@news.Hawaii.Edu>, gerard@hawaii.edu wrote:
>In article <585t1a$5pd@bashir.peak.org>, kellyt@PEAK.ORG (Tim Kelly) writes:
>>Gerard Fryer (gerard@hawaii.edu) wrote:
>>
>>> comment that since the US National Earthquake Hazards Reduction Program
>>> was initiated under President Carter, not one US earthquake has been
>>> predicted.
>>
>>Hey Bob, Dennis: I guess they haven't got our email addresses. The post
>>is somewhat correct. There's been a bunch of them predicted! :)
>
>(note that I have corrected the attribution above)
>
>I was only quoting. The original reference, Midori Ashida's article,
>implied "no potentially damaging earthquake has been predicted."
>Despite Bob and Dennis' efforts, it is true that no earthquake larger
>than magnitude 5 has been predicted in a manner which allows us to
>dismiss the null hypothesis. Farnsworth and Trombley might disagree,
Well, I can't really argue to much due to the way my predictions are/
have been formatted. But I did have hits on the Parkfield 5.0, North-
Ridge 4.5 and 5.0, and the Tahoe 6.0 or whatever it was.
>but their prediction of the 1993 Woodburn earthquake was made under
>dubious circumstances (we'll never know for sure if the event was
>pre-dicted or post-dicted) and it was followed by spectacular
>failure--their late-93 Cascadia event never occurred.
>
>There is no question that Dennis' results are tantalizing, but they are
>tantalizing in precisely the same manner as the VAN predictions: while
>the success rate is high the failure rate is high also. Is Dennis onto
>something? It's hard to say. The same arguments that have been raised
>over VAN apply to Dennis' predictions. See the special issue of
>Geophysical Research Letters (vol. 23, no. 11, 27 May 1996).
When you look at the prediction of all earthquakes, yes the failure
rate is high. When you look at the predictions themselves, the
failure rate isn't that high at all.
Before I've said that since I'm not able to monitor 24 hours a day
I will miss signals/posting predictions. Until that little problem
gets resolved, we won't know if this method can predict all quakes
or only a subset of them.
This gets back to the reliable argument. If a method can only
predict a subset of quakes, and that method (at least for the
posted predictions) displays a better than chance hit rate, would
that method still be considered reliable?
To my line of reasoning, even if a prediction methodology can only
predict a subset of quakes, its a building block in the identification
of earthquake precursors.
Regards,
Dennis
