Back in the mid 1960's, when I was in my mid 20's, I was an Engineering
Tech with the department of water resources, I had the job of analyzing
crustal strain measurements across the San Andreas Fault.  The data were
annual measurements with a Geodimeter of the distance between mountain
tops, such as Mt. Hamilton to Loma Prieta (30 Km) and down to Palm
Springs. The basic accuracy was better than 1 part/million, or about a
cm on a 20 Km line, plus the onerous temperature error, 1
part/million/deg C.

One of the staff programers wrote me a little program that looked for
anything unusual, based on the expected value and the historical error
on a particular line.  This program, running on an IBM 1620, would print
out "EARTHQUAKE COMING" on the data tabulation for any particular line.

I kept a chart on my wall of predicted earthquakes, and hit on about
three of them.  Remember, we're predicting earthquakes in California
with data taken once a year, so statistics were on my side.

Our group manager read the chart, and perhaps having too much confidence
in a young kid, announced one of the predictions at the AGU meeting.  It
made a good sized article in the San Francisco Chronicle.  As you would
expect, that particular earthquake didn't happen.  Of course you don't
get fired when you work for the government, but the Department of Water
Resources eliminated his position from the budget next year and he went
on to other things.

The Geodimeter program was later taken on by the USGS, and they still do
a lot of Geodimeter work, but they haven't found anything significant.

I personally thought our problem was the long sample (one year typical)
between measurements, so you had to be lucky to catch an earthquake. 
Back then of course, only sooth sayers predicted earthquakes.  It was
some years before the USGS decided that prediction was politically
correct and fell on their face at Parkfield.

Interestingly enough, I did have a pretty solid prediction of an earlier
earthquake near Loma Prieta (the 196? Corralitos Earthquake), where the
mountain actually bumped over about 10 cm from its expected position a
short time before the earthquake.  If you examine the USGS data for Loma
Prieta, prior to the 1989 earthquake, you can see a similar pulse in the
data.  My program would have flagged the 1989 Loma Prieta Earthquake
prior to the event. The USGS used a different prediction criteria and
they reported "no prediction" even though it is obvious in the data.

By the way, we did Parkfield before Parkfield was cool. The data said
that the fault was just ripping along North of Parkfield, and locked up
just South.  We created and measured a large pentagon of Geodimeter
lines in the area of interest before the Parkfield earthquake, and then
came back and measured it again after the Parkfield earthquake, so you
could say we predicted that one too and backed it up with some effort.

It was all written up in a DWR Bulletin, but it was never referenced
much because our leader was so discredited by his failed prediction. I
wrote the report though, and it's good solid information on fault
movement in California for that 10-year period.

-- 
Doug Crice			 http://www.georadar.com
19623 Via Escuela Drive		      phone 408-867-3792
Saratoga, California  95070  USA	fax 408-867-4900
__________________________________________________________

Public Seismic Network Mailing List (PSN-L)