Slashdot Mirror
NASA Quakesim Predicts 15 Out of 16 CA Quakes
Saint Aardvark writes "NASA's QuakeSim project has successfully predicted15 out of 16 of California's earthquakes with magnitude > 5, including 11 since the map was published in 2002. "So far, the technique has only missed one earthquake, a magnitude of 5.2, on June 15, 2004, under the ocean near San Clemente Island.""
I think it's a fair to ask how far in advance the quakes were predicted.
So, have they predicted when California is going to fall into the ocean?
How many did it predict that never happened?
When they first started to try to predict tornados (right before they happened) a lot of people thought it was impossible, but they managed to do it. And this was before they had Doppler 8 million, just with changes in the pressure and such. A lot of people have tried to say predicting earthquakes will be impossible, but I think with enough research and data it will be as easy and predictable as tornados.
That means it won't be able to predict Quake IV :(
Do they have a multiplayer version of this? Will it run on Linux? ;)
fact is, though, this is a big step in the right direction. To successfully predict a quake, you need location, magnitude and time. They've got location down and time vaguely constrained (article doesn't state time effect, other than a quake in this location within 10 years) and a lower limit on magnitude (at least 5). That's better than nothing. and certainly something to use as a starting point for future modelling efforts.
Check out the archived version here
From the blurb:
Following recent seismic activity in California and the threatened eruption of Mount St. Helens in Washington State. Forum takes a look at scientific advancements in the study of earthquakes and volcanoes.
Host: Michael Krasny
Guests:
It's easy to "predict" something that you cause yourself.
/me puts on tinfoil hat.
I predit that I will put my tinfoil hat on!
Ooh, I must be psychic or something!
They're just putting their earthquake machines through the paces before they use it to ensure W's re-election, and to make a tidy profit for Haliburton, which owns a lot of soon-to-be-coastal property along the California-Nevada border.
The good and new comes from no quarter where it is looked for, and is always something different from what is expected.
The article is vague. No mention of whether there were any false positives. People will get fed up really quickly if told to evacuate and no quake comes.
Also, it's not clear to me that what their predicting ("hotspots") is the same thing as predicting when an earthquake will happen. How long do individual "hotspots" exist in one place?
Still, earthquake and other Earth-sci prediction simulations can be useful. Just probably better for long-term planning than individual predictions.
I wouldn't exactly say they predicted these quakes. From what the article says, it seems like they just have a list of places where they say a >5 quake will occur between 2000 and 2010. Whoop-dee-fucking-doo. Anything under a 6 is less annoying to a Californian than a fly buzzing around your house. For anything over a 6.5 or so, being this accurate about WHERE the quake will occur is next to useless if you can't be more accurate about WHEN it will occur.
This is not meant to diminish the accomplishments of these researchers. I'm sure this is a very important stepping stone to greater advances. However, this won't be useful to most people until they can predict with much greater accuracy the magnitude of the quake and the timeframe in which it will occur.
This sounds like a really cool endeavor, but why is this funded by NASA? NASA takes a lot of flak for using up lots of money and being a big administration, and its not surprising to see why with projects like this under its wings.
In the end, shouldn't such earthquake research be under a different organization?
The space unintentionally left unblank.
#include <stdio.h>
int main()
{
int lat, lon;
for (lat = -90; lat <= 90; lat++)
for (lon = -180; lon <= 179; lon++)
printf("prediction: lat=%d +/- 0.5, lon=%d +/- 0.5\n", lat, lon);
}
The articles cites don't speak much on methodology beyond using historical data (since 1932) to predict areas where magnitude 5+ earthquakes are likely. It would be interesting to read about the methods and determine if they can predict larger (6 - 6.5+) damaging earthquakes of a larger magnitude. Or more significantly, earthquakes whose damage potential (as measured using the Mercalli scale) is high.
Geologists have been using historical data to predict future potential for many years. Stream flow data are used to predict flood potential; historic hurricane landfall data are used to predict beach erosion potential; lahar ages are used to predict volcanic mud flow damage potential.
What is more useful that merely predicting areas where magnitude 5+ earthquakes are likely to occur is predicting the frequency and areal extent of damage potential -- Mercalli intensity VIII +, roughly correlating to Richter magnitude 6.5+.
Then again, we Americans continue to rebuild on 100 year flood plains, hurricane-savaged barrier islands and earthquake-prone areas. The engineering geology I learned was to avoid areas where Mother Nature is going to win in the long run.
Mind you I didn't read the article.
...But then again, who knows, maybe they did hype it up some time ago... however, this is the first I have heard of this effort.
But my point has nothing and everything to do with that.
As far as the field of "Earthquake Prediction" goes or any metric forecast for the fact of the matter, this is the first time I have heard of results before an announcement of intent.
And I wholeheartedly applaud that.
It's great to see an endeavor that "just does" without yapping it up to create the hype and controversy. Mind you, I understand it's NASA and thus publicly funded, but still, they rely upon Congressional approval for funding. I simply find it amazing to see this research performed under the radar until proven results were found. Such a course of conduct is quite admirable and after seeing hype after hype about other disperately related projects that talked up computational predictions of natural phenomena only to produce such an obscene amount of false positives as to be utterly useless.
As a geologist, I do find the prospect of earthquake prediction quite exciting. I even worked a few doors down from Dr. Keilis-Borok (predicted Japan and San Simeon, failed prediction in Southern California) this summer at UCLA, doing some earthquake research.
0 5071107.htm
Dr. K-B's approach used statistical analysis and was quite an interesting idea. His paper even correlated some previous earthquakes (such as Landers and Northridge) using his "tail-wag-the-dog" method to try and verify his results.
Anyway, regarding these latest predictions by John Rundle and his team, I decided to read the paper. You can actually find it here.
I'm not understanding how they succesfully predicted certain things or how useful his theories are. They are saying they predicted three of the earthquakes that happened in Big Bear.
From what I am understanding, the way their method works is that it shows potential "hot spots" for earthquakes for the next 10 years. That means the whole Big Bear/San Bernardino Mountains area should show up as a hotspot on their map. This doesn't mean they have succesfully predicted all 3 earthquakes though if I understand this right. They predicted the potential for one M5.0 or greater there withing the next 10 years. The fact that there were three of them is just icing on the cake I suppose?
I also can't find any information that shows how many false-positives they nailed as well. This might be kind of hard since they won't know about false-positives until after their prediction period is up in 2010. Without that data though, we can't really be sure of how good this method works. And even if it misses some, it only reduces the chance of an earthquake happening in the next X years to some percentage (which we already have certain data for from the USGS. 67% chance of a M6.7 or greater striking the Bay Area before 2030 and an 80% chance of a M7.0 or greater striking Southern California before 2030).
Admitedly, if this method is promising, it might put better constraints on the data though, so we could say something like, "97% chance of an M7.0 striking within 10 years." However, this still won't help all that much in the scheme of things.
Additional information:
http://www.sciencedaily.com/releases/2004/10/0410
John Rundle's Paper