Earthquake Prediction Months In Advance

eegad writes "A UCLA seismologist named Vladimir Keilis-Borok claims earthquakes can be predicted months in advance. In the article at the University of California Newswire, he claims that the "team including experts of pattern recognition, geodynamics, seismology, chaos theory, statistical physics and public safety ... has developed algorithms to detect precursory earthquake patterns." It also says "the team's current predictions have not missed any earthquake, and have had its two most recent ones come to pass." They predict "an earthquake of at least magnitude 6.4 by Sept. 5, 2004, in a region that includes the southeastern portion of the Mojave Desert, and an area south of it." We'll see if they're right."

USGS Earthquake Reference Site

[Anonymous Coward]

USGS Earthquake Reference Site [usgs.gov]

Incidentally, I'm posting this because I want to test the load bearing of this server, one of the ones I take care of here at work. So click away.

(anon to avoid karma-whoring)

PBS

[starvingcodeartist]

For years scientists have known about the signs that the faults give off before an earthquake occurs, but most scientist are skeptical that they'll ever be able to accurately predict them because there are so many environmental factors to consider. Read more on PBS's microsite called Savage Earth, The Restless Planet: Earthquakes [pbs.org]. It talks about prediction and whatnot.

Anyone heard of Kushida in Japan?

[Anonymous Coward]

He made an earthquake prediction in Japan [space.com] based on radio waves, and he actually came pretty close [patriot-paradox.com]. Close enough that his ideas are worth more investigation.

they've been making these predictions 20 years

[peter303]

This Russian group first got attention in the US seismology community when it "predicted" the Loma Prieta (Silicon Valley) quake of 1989. The technique performs spatial-temporal statistical analysis of weaker earthquakes that proceed large quakes. The first President Bush even asked the US Geological Survey to look into this.

The method may work, but it has not yet passed the scientifically required of repoducibility by scientists outside the Russian research group. Several leading US seismologists have tried reproducing this analysis method without success. Either the method is devilishly difficult to reporduce, important details have [perhaps intentionally] not been published, or it really doesn't work. Furthemore, you dont see the US results in press, because people generally dont publish negative results. Hopefully the reproducibility issues will be resolved and there will be a successful prediction method.

(Read my lips: cold fusion)

the most important prediction method

[peter303]

The most important prediction method is to antipicate the maximum horizontal force resulting from an earth quake. A force execeeding 10% the amout of earth's surface gravity, called a "g", at one Hertz can collapse a poorly designed building or overpass. 200% g is observed in the largest quakes. A guide to destruction in terms of "g" is here [slu.edu].

The United States Geological Survey has spent a lot of effort [usgs.gov] predicting maximum forces. this is based on the location of previous large earthquakes and local soil conditions among other factors. This has resulting in relatively low death rates of quakes of similar size. For example last month's central California quake and Iranian quakes were about the same size with death tolls of 3 and 30,000. Ditto 1994 Northridge and 1995 Kobe Japan with tolls of 55 and 6,000.

Re:So that means...

[Anonymous Coward]

You don't have to be an expert to predict earthquakes in Iran:

Dec. 26, 2003: Southeastern Iran, magnitude 6.5; at least 20,000 killed.
June 22, 2002: Northwestern Iran, magnitude 6; at least 500 killed.
May 10, 1997: Northern Iran, magnitude 7.1; 1,500 killed.
June 21, 1990: Northwest Iran, magnitude 7.3-7.7; 50,000 killed.
Sept. 16, 1978: Northeast Iran, magnitude 7.7; 25,000 killed.

need to be like bad weather predictions

[peter303]

Meterologists have found that people dont pay attention to tornado or hurricane predictions unless they are better than 30% accurate over a city-size area and couple hour time window (one day for hurricane). Too many false alarms are ignored.

An earthquake prediction is considered successful in the scientific sense if it beats background chance. (Backround chance is computed by counting space-time windows through seismic catalogs). Earthquakes are so rare, e.g. large ones in tens of thousnds of days in California, that large space-time window can beat chance. However, no one has published a reproducable methods for general earthquake prediction (ecuding aftershocks, maximum force, etc) that has eat chance.

Re:Where is said prediction? / Why it can't work

[Anonymous Coward]

Here is the prediction made by Vladimir.

Keilis-Borok's team now predicts an earthquake of at least magnitude 6.4 by Sept. 5, 2004, in a region that includes the southeastern portion of the Mojave Desert, and an area south of it.

Re:PBS

[YU Nicks NE Way]

In fact, von Neumann was provably wrong. Ed Lorentz' work on chaotic attractors in the Navier-Stokes system was so controversial presiely because it showed that long-term weather prediction over a period of more than about 23 days is impossible -- at least, if quantum mechanics is a valid theory.

Re:USGS Earthquake Reference Site

[WuphonsReach]

You know, if you *really* want to test your own systems, go grab a copy of OpenSTA [sourceforge.net].

Reasonably flexible and GPL'd.

Re:There's been other studies

[dejamatt]

Only study I could find seemed to offer no evidence of this: http://www.johnmartin.com/earthquakes/eqpapers/000 00072.htm [johnmartin.com]

CONCLUSION This study shows that a significant positive correlation does not exist between the behavior of pets in the San Jose area and the occurrence of earthquakes within the same area over the three year period from January 1983 through December 1985. Based on this random disappearance of pets with respect to earthquakes,no scheme seems possible to predict earthquakes using newspaper reports of missing pets.

Earthquake alarm systems

[peter303]

Depending on the geographic situation, there can be seconds to minutes for the most descruction seismic waves to hit you (surface waves travel about 3 miles a second). That might give you enough time to shut down computers, natural gas feeds, subways, etc. A conference last month reviewed [agu.org] progress in this area. Mexico probably has the best situation because its west coast quakes take about six minutes to reach Mexico City which has been mostly constructed on "mud". Southern California is less lucky, because it can be right over the quake. Japan and Taiwan are inbetween with cities about a minute from major faults. The Mexican system even puts text warn on TV like tornado reports, according to the abstract.

The traditional alarm methods listen to several stations in order to block out non-earthquake events and triangulate the location. But this takes 2-5 minutes waiting for enough information. Some research is going towards single-station, first couple second analysis, which may be useful for Los Angeles.

Re:USGS Earthquake Reference Site

[Reziac]

As of this instant, your server is downright swift. [wanders around] Oooh, this works nice, and no evil javascript required.

My university had the complete USGS survey book, the big thick one with maps of everything anyone ever tracked, from climate to weeds. (Wonderful resource, that book.) I remember the earthquake details as compiled up thru 1958, and that if you want a quake-free location, the closest to that is North Dakota (only 3.n magnitude or less on the map). And it's amazing how many major metro areas are planted directly atop historical large-quake clusters.

Re:Richter scale...

[GeoGreg]

Actually, the Richter scale is no longer used to describe earthquakes. What is reported in the media as a value on the "Richter scale" is usually a moment magnitude, a much better estimate of the released energy. I think the USGS has been trying to educate the reports not to use the term "Richter", and it seems to be working, as one usually now hears about "magnitude 7.3" earthquakes.

Using the Mercalli scale is much more difficult, as it is not quantitative. Mercalli intensity is a qualitative description of the amount of shaking felt and the amount of property destruction. Plus, Mercalli intensity is not a single value, but rather may be different at every location. Nevertheless, the USGS has been working on a product called ShakeMap that can estimate Mercalli intensity within a few minutes of a quake. However, constructing these maps requires extensive local seismic networks. For an example of a ShakeMap, see this link [usgs.gov].

Predicting the shaking from a given quake (e.g. mag. 7 and 15 km depth in a particular location) before the fact for planning purposes is also done. Small variations in the earthquake parameters (location, direction of slip, depth, etc.) may significantly affect the shaking felt at a given location. Local geology also has a big effect on the amount of shaking experienced. So, it's a tough problem that requires lots of data.