# Tokyo earthquake predicted by astronomer?

Staff Emeritus
Gold Member

## Answers and Replies

Related Earth Sciences News on Phys.org
Tsu
Gold Member
Can any of the geologists here answer my question that I posted in Ivan's thread?

Staff Emeritus
Gold Member
Interesting: A striking [earth shaking] coincidence?
TOKYO, Sept. 26 — Three powerful earthquakes, one of them of potentially historic magnitude, struck Hokkaido in northern Japan early Friday morning, causing major structural damage, NBC News reported. The quakes injured more than 240 people and generated a 7-foot-high tsunami off the coast of Hokkaido. Tsunami advisories were issued for much of the Pacific region, including Japan, Russia and the Philippines.
http://www.msnbc.com/news/971921.asp?pne=msntv

This second - estimated as an 8.0 - event was off by about 660 miles and one week from the prediction. The chance of coincidental success appears to be about 1:600 in this case.

The first event was about three days late, and it was a 5.5 instead of a 7.0.

Last edited:
russ_watters
Mentor
Originally posted by Ivan Seeking
The chance of coincidental success appears to be about 1:600 in this case.
Based on what?

Predicting earthquakes in Japan is like predicting snow in Utah or fog in England: not very difficult if you don't have to be too specific about it.

From the first article:
The earthquake research establishment has largely ignored the warning.

Forecasting quakes is generally considered to be impossible with current technology, and Kushida's method of using anomalies in the VHF range of radio waves to predict the timing and intensity of tremors has not gained many believers in the scientific community.

Last edited:
Staff Emeritus
Gold Member
Originally posted by russ_watters
Based on what?

Predicting earthquakes in Japan is like predicting snow in Utah or fog in England: not very difficult if you don't have to be too specific about it.
This estimate is based on data from the USGS. There have been 9 quakes in excess of 7.0 in Japan since 1891. Please see the linked thread in the S&D forum.

Staff Emeritus
Gold Member
The earthquake research establishment has largely ignored the warning.

Forecasting quakes is generally considered to be impossible with current technology, and Kushida's method of using anomalies in the VHF range of radio waves to predict the timing and intensity of tremors has not gained many believers in the scientific community.
Yes, so? The article was printed before the quakes.

russ_watters
Mentor
Originally posted by Ivan Seeking
This estimate is based on data from the USGS. There have been 9 quakes in excess of 7.0 in Japan since 1891. Please see the linked thread in the S&D forum.

From your post it still isn't clear how you calculated that probability. Its really not that hard. From your numbers, the magnitude 5.5 earthquake was off by 200x from the predicted magnitude 7.0. Then multiply that by the frequency of magnitude 5.5 earthquakes in Tokyo (if I had to guess, I'd guess one or two a year). That would make the odds at best 1:200,000 of being that close on picking a day and a magnitude.

Staff Emeritus
Gold Member
Originally posted by russ_watters
From your post it still isn't clear how you calculated that probability. Its really not that hard. From your numbers, the magnitude 5.5 earthquake was off by 200x from the predicted magnitude 7.0. Then multiply that by the frequency of magnitude 5.5 earthquakes in Tokyo (if I had to guess, I'd guess one or two a year). That would make the odds at best 1:200,000 of being that close on picking a day and a magnitude.
Where did you get this?

On the the average, one quake over magnitude seven every 600 weeks for all of Japan. There was a 5.5 in Tokyo, and then a 7.0 and an 8.0 a week later; about 660 miles north. The big quakes are what most interest me. These are fairly rare events as indicated by the USGS data. To miss a 7.0 anywhere in the neighborhood by only a week ain't bad. I am sure that you would normally argue strongly against an explanation that only has odds of 1:600 of being true.

Also, it is not clear how to calculate the accuracy of the original quake prediction. His exact method is not given; it may ultimately be based on energy, magnitude, or even as a linear function of the magnitude value. The fact is, this astronomer went public with a risky prediction. A week later, it appears that against odds of at about 1:600, he may have gotten it right. A lot of the interpretation of success seems to depend on his method; but even if we understood his method, we may have no theory to explain the observations.

Edit: Amateur Astronomer

Last edited:
FZ+
Do we have any details on the theoretical background of his predictions?

Staff Emeritus
Gold Member
Originally posted by FZ+
Do we have any details on the theoretical background of his predictions?
I couldn't find his website the first time I looked. It popped right up this time.

Astronomers have been observing meteors using the back scattering (*1) of VHF radio waves beyond the range of sight. It was noticed that such observation detects specific temporal changes (preseismic anomalies) before the occurrence of an earthquake. The level of the temporal changes is proportional to the scale of the expected earthquake, and the time lag preceding the seismic activity varies depending on the type of earthquake. Our systematic observation continued since 1995 revealed the followings: (1) the region covered by radio monitoring is defined by the location and output power of each VHF transmission station; and (2) the anomalies detected by radio monitoring provide the basis for estimating the location, scale, and time of future earthquakes despite a certain degree of uncertainty. This document is based on the report "On a Possibility of Earthquake Forecast by Radio Observation in the VHF Band, 1998, Y. KUSHIDA and R. KUSHIDA, RIKEN Review, No. 19, p. 152-160," supplemented with new information from recent observation work.
http://www.jpinfo.ne.jp/yatsu/yochi/egaku/ [Broken]

Last edited by a moderator:
Staff Emeritus
Gold Member
Here are the initial data and comments

Magnitude 8.3 HOKKAIDO, JAPAN REGION
2003 September 25 19:50:06 UTC
Preliminary Earthquake Report
U.S. Geological Survey, National Earthquake Information Center
World Data Center for Seismology, Denver

Magnitude 8.3
Date-Time Thursday, September 25, 2003 at 19:50:06 (UTC) - Coordinated Universal Time
Friday, September 26, 2003 at 04:50:06 AM local time at epicenter
Time of Earthquake in other Time Zones
Location 41.83N 143.83E
Depth 33.0 kilometers
Region HOKKAIDO, JAPAN REGION
Reference 135 km (85 miles) SSW of Kushiro, Hokkaido, Japan
245 km (150 miles) NE of Hachinohe, Honshu, Japan
245 km (155 miles) SSE of Asahikawa, Hokkaido, Japan
770 km (480 miles) NNE of TOKYO, Japan

Location Quality Error estimate: horizontal +/- 5.7 km; depth fixed by location program
Location Quality
Parameters Nst=199, Nph=199, Dmin=575.6 km, Rmss=0.84 sec, Erho=5.7 km, Erzz=0 km, Gp=34.2 degrees
Source USGS NEIC (WDCS-D)
Remarks At least 400 people injured, extensive damage, landslides and power outages occurred and many roads damaged in southeastern Hokkaido. A tsunami generated with an estimated wave height of 1.0 meter along the southeastern coast of Hokkaido. Felt strongly in much of Hokkaido. Also felt in northern and much of central Honshu as far south as Tokyo. Recorded (6L JMA) in southern Hokkaido; (5U JMA) in parts of eastern Hokkaido; (5L JMA) in central Hokkaido; (4JMA) in parts of Hokkaido and northern Honshu; (3 JMA) in southern Iwate, Akita, Miyagi, Yamagata and northern Fukushima; (2 JMA) in southern Fukushima, northern Niigata, eastern Tochigi, Ibaraki, Saitama, Tokyo, Chiba, Kanagawa, eastern Shizuoka and eastern Yamanashi; (1 JMA) in eastern Gumma and parts of Nagano Prefectures. Also recorded (1 JMA) on Miyake-jima and Hachijo-jima.

Tectonic Summary
The preliminary location and focal-mechanism of this earthquake imply that it occurred as the result of thrust-faulting on the plate interface between the overriding North American plate (which extends into the northeast corner of the Eurasian landmass) and the subducting Pacific plate. The Pacific plate is moving west-northwest at a rate of about 8.2 cm per year relative to the North American plate. In addition to experiencing great thrust earthquakes that originate on the interface between the plates, eastern Hokkaido experiences great earthquakes that originate from the interior of subducted Pacific plate. The earthquakes of March 4, 1952, and May 16, 1968 (cited below) were interface-thrust earthquakes, whereas the earthquake of January 15, 1993 (cited below) occurred within the interior of the subducted Pacific plate. The recent earthquake appears to have involved rupture of the same section of the plate interface that ruptured in 1952.

Magnitude 8 and greater earthquakes are capable of devastating large areas. The shallow September 25 Hokkaido earthquake occurred about 60 km offshore. If the earthquake had occurred directly beneath a populated region, damage would have been more severe.

Previous Deadly Earthquakes in this Region
Date UTC Magnitude Fatalities Damage
1952 March 4 8.6 31 31 killed, 72 injured; 713 houses destroyed, 5,980 damaged. 28 killed and warehouses destroyed at Kushiro. 3 killed and 309 houses destroyed at Kiratapu. 1,000 houses destroyed or damaged at Shiranuka and 400 schools collapsed at Sapporo. 10-foot tsunami.
1968 May 16 7.9 48 Damage estimate at 25 million USD.
1993 January 15 7.6 2 614 injured and substantial damage (VI JMA) at Kushiro, Hokkaido and Hachinohe, Honshu. Felt (V JMA) at Hiroo, Nemuro, Obihiro, Otaru and Urakawa; (IV JMA) at Hakodate and Tomakomai; (III JMA) at Sapporo, Hokkaido. Felt (IV JMA) at Aomori and Morioka; (III JMA) at Akita, Fukushima, Sendai, Tokyo and Yokohama, Honshu. Also felt (VII) on Shikotan and (VI) at Kurilsk, Kuril Islands. Landslides and subsidence occurred in the epicentral area.

The last great earthquake (magnitude 8 or greater) in the world was a magnitude 8.4 that occurred on June 23, 2001, near the coast of Peru. This earthquake killed at least 75, including 26 killed by the associated tsunami.

http://neic.usgs.gov/neis/bulletin/neic_zdap.html

Last edited by a moderator:
Nereid
Staff Emeritus
Gold Member
Where can one find a complete set of Kushida's public predictions (including date, time, and 'publication'), and a complete set of earthquakes in the areas monitored over the time he published his predictions (and go back a week before the start)?

Staff Emeritus
Gold Member
This is the information on the quake near Tokyo that was first associated with his prediction.

Magnitude 5.4 WESTERN HONSHU, JAPAN
2003 September 24 20:56:54 UTC
Preliminary Earthquake Report
U.S. Geological Survey, National Earthquake Information Center
World Data Center for Seismology, Denver

Magnitude 5.4
Date-Time Wednesday, September 24, 2003 at 20:56:54 (UTC) - Coordinated Universal Time
Thursday, September 25, 2003 at 05:56:54 AM local time at epicenter
Time of Earthquake in other Time Zones
Location 35.37N 135.21E
Depth 370.2 kilometers
Region WESTERN HONSHU, JAPAN
Reference 80 km (50 miles) NNW of Osaka, Honshu, Japan
95 km (60 miles) E of Tottori, Honshu, Japan
120 km (75 miles) SW of Fukui, Honshu, Japan
415 km (255 miles) W of TOKYO, Japan

Location Quality Error estimate: horizontal +/- 6.9 km; depth fixed by location program
Location Quality
Parameters Nst=125, Nph=131, Dmin=300.0 km, Rmss=0.82 sec, Erho=6.9 km, Erzz=0 km, Gp=71.4 degrees
Source USGS NEIC (WDCS-D)
http://neic.usgs.gov/neis/bulletin/neic_zcat.html

Last edited by a moderator:
Staff Emeritus
Gold Member
Many comments on earthquake prediction:

IUGG 2003 Scientific Program
example:
TIME [ 1400-035 ] [ JWS01/02P/D-001 ] [ Poster ]
AN INDEPENDENT EVALUATION OF THE WEST PACIFIC SHORT-TERM EARTHQUAKE FORECAST
Vladimir G. KOSSOBOKOV(IIEPT&MG, Russian Acad. Sci.; IPG de Paris)

Using the literal measure of territory in square km one may overestimate statistical significance of the obtained results by equalizing the areas of high and low seismic activity, at the extreme, areas where earthquake happen and do not happen. We suggest a natural null-hypothesis, i.e.,seismic roulette, which accounts for heterogeneity in spatial distribution of earthquakes for an assessment of the efficiency of an earthquake prediction algorithm and demonstrate how to apply it. Specifically, given the trek-record of the real-time earthquake probabilistic forecasts (http://scec.ess.ucla.edu/~ykagan/predictions_index.html; [Broken] Kagan and Jackson, 2000. Probabilistic forecasting of earthquakes, Geophys. J. Int., 143, 438-453), we analyze the predictions arising from setting a threshold probability or a threshold probability ratio. In either case, the effectiveness of prediction came out to be hardly better than random guessing. The conclusion holds even when evident aftershocks are included into statistic of successes.
http://www.jamstec.go.jp/jamstec-e/iugg/htm/abstract/abst/jws01_p/p2.html [Broken]

Other related info:
Earthquake
All the current Earthquake news

http://earthquake.newstrove.com/

Last edited by a moderator:
FZ+
Originally posted by Ivan Seeking
I couldn't find his website the first time I looked. It popped right up this time.

http://www.jpinfo.ne.jp/yatsu/yochi/egaku/ [Broken]
So in effect, it appears to be an empirical observation, based on statistical correlations? I think I understand the doubt that was placed on his predictions... Hmm, it very interesting, but I think we should wait for more results before we can decide there is something in his method.

As for his hypothesis for the mechanism...
During the earthquake preparatory process, a change in mechanical stress caused by microcracks may alter electric charges on and near the earth's surface.
This appears to be the critical part of it. Would it be possible to cut out the meteor/radio bit and just directly measure for this to validate his hypothesis? And if this was the case in this earthquake, can this form part of a general method for all earthquakes? (Some rocks may respond like this, but could this just be a lucky break?)

Last edited by a moderator:
russ_watters
Mentor
Originally posted by Ivan Seeking
This is the information on the quake near Tokyo that was first associated with his prediction.
Ah, but the question was what about his other predictions? Regardless of odds (since odds are tough, I'll let that line go) if he's made thousands of predictions and this is the only one that was even close, that would call his theory into question.

Staff Emeritus
Gold Member
Originally posted by russ_watters
Ah, but the question was what about his other predictions? Regardless of odds (since odds are tough, I'll let that line go) if he's made thousands of predictions and this is the only one that was even close, that would call his theory into question.
True. I assumed [bad thing to do] that since his prediction caused such an uproar, he has not made too many false predictions in the past.

russ_watters
Mentor
Originally posted by Ivan Seeking
True. I assumed [bad thing to do] that since his prediction caused such an uproar, he has not made too many false predictions in the past.
Or maybe none. People tend to jump on the first prediction and if thats wrong they forget the person even existed. Pretty soon, 600 people have made predictions and lo and behold (with 1:600 odds), one gets a hit. EVERYONE predicts earthquakes. Hell, I have a bunch I'm predicting. I won't tell you when or where though - you have to buy my book.

Staff Emeritus
Gold Member
Originally posted by russ_watters
Or maybe none. People tend to jump on the first prediction and if thats wrong they forget the person even existed. Pretty soon, 600 people have made predictions and lo and behold (with 1:600 odds), one gets a hit. EVERYONE predicts earthquakes. Hell, I have a bunch I'm predicting. I won't tell you when or where though - you have to buy my book.
You are of course absolutely correct; mark that down I won't say it often. However, , the potential differences here are that first, he claims to be following empirical evidence that can be subjected to scrutiny. Next, unless you can find a significant number of near term predictions for a quake >= 7.0 in Japan, I don't think these predictions exist. Finally, even if these predictions did exist, he still only had a 1:600 chance of being correct.

My main point is not that he was right but that he deserves to make his case. Unfortunately, he may have no idea how to interpret this event himself. Clearly he does not have the location and timing nailed down.

If you were a geologist working on earthquake prediction, would you ignore this event?

Nereid
Staff Emeritus
Gold Member
Surely the first (one of the first?) things to do is look at the track record and what the prediction(s) actually are.

To repeat what I said elsewhere, Did this guy make public, ahead of time, ALL his predictions? If so, we can assess his track record, by making sure we match (or not) his predictions against actual earthquakes, and vice versa.

It would also be a good idea to write down, before looking at all his predictions, how you will assess the predictions, against actual earthquakes (and vice versa). This will help to avoid (unconscious) bias in the assessments themselves; it's always harder to do this after you've looked at the data.

One area where the assessment protocol needs to be particularly careful is the magnitude. As you point out, the Richter scale is logarithmic, and IIRC earthquake magnitudes follow a power law quite closely. This would suggest, for example, that a prediction of a magnitude 6 earthquake and an observed one of 5 should carry a different weight in any assessment from a prediction of 5 and an observed magnitude 6 quake.

Staff Emeritus
Gold Member
Originally posted by Nereid
To repeat what I said elsewhere, Did this guy make public, ahead of time, ALL his predictions? If so, we can assess his track record, by making sure we match (or not) his predictions against actual earthquakes, and vice versa.
In the link in the first post of this thread you will find a newspaper article that came out a week before the quake. He had causes quite an uproar by predicting a >= 7.0 in Tokyo. As far as his other "predictions", I have posted everything that I could find so far. You can email him and ask - his contact info is given at his website.

Ivan,

You are giving the man credit for having gotten the magnitude right, but the time and location off somewhat. From the article you linked in the other thread the man himself is claiming the reverse, that he got the time and location right (The tokyo quake) but the magnitude off:

"But on Saturday, the city shook as a tremor measuring just 5.5 struck. The only injuries were caused when a wall at a Buddhist temple collapsed; seven people received minor bruises.
This week, Mr Kushida was claiming he had got it right. The earthquake has already happened as I predicted,' he said from his observatory, arguing that his prediction was within the acceptable margin for error."

He is claiming that a 5.5 when he predicted >7 is "within the acceptable margin for error". As you demonstrated in the other thread however you chose to look at it, the difference between a 5.5 and a 7 is either large, or enormous.

So it bothers me that he is so cavalier about the aspect of his prediction that was so far off as to be outside any really accepable margin for error in my view. As far as the magnitude goes he was plainly wrong.

I think it's important to reiterate that he, himself, is not claiming success by adopting the Hokkaido quakes.

-Zooby

Last edited:
Staff Emeritus
Gold Member
Originally posted by zoobyshoe
Ivan,

You are giving the man credit for having gotten the magnitude right, but the time and location off somewhat. From the article you linked in the other thread the man himself is claiming the reverse, that he got the time and location right (The tokyo quake) but the magnitude off:

"But on Saturday, the city shook as a tremor measuring just 5.5 struck. The only injuries were caused when a wall at a Buddhist temple collapsed; seven people received minor bruises.
This week, Mr Kushida was claiming he had got it right. The earthquake has already happened as I predicted,' he said from his observatory, arguing that his prediction was within the acceptable margin for error."

He is claiming that a 5.5 when he predicted >7 is "within the acceptable margin for error". As you demonstrated in the other thread however you chose to look at it, the difference between a 5.5 and a 7 is either large, or enormous.

So it bothers me that he is so cavalier about the aspect of his prediction that was so far off as to be outside any really accepable margin for error in my view. As far as the magnitude goes he was plainly wrong.

I think it's important to reiterate that he, himself, is not claiming success by adopting the Hokkaido quakes.

-Zooby
I realize this. Like I said, the mag 7 and 8 are what got my attention. The odds are still the odds. I could not be sure of the odds of success in the case of the smaller quake. Hopefully his method will get the attention needed to sort this out. I just wonder if he will continue with public proclamations after missing this one.

Originally posted by Ivan Seeking
Hopefully his method will get the attention needed to sort this out.
I agree. The prospect of knowing when a quake is going to occur because of definite, measurable signals it sends out shortly ahead of time would be priceless. I would definitely like to see more people around the world trying it, both to see hw much there is to it, and if there is anything, to refine it.

Nereid
Staff Emeritus
Gold Member
zooby: He is claiming that a 5.5 when he predicted >7 is "within the acceptable margin for error" . As you demonstrated in the other thread however you chose to look at it, the difference between a 5.5 and a 7 is either large, or enormous.
Ivan: Like I said, the mag 7 and 8 are what got my attention. The odds are still the odds.
Now let's see ... if he predicted 5.5, and the observed was >7, and this is "within the acceptable margin for error", does it follow that he could claim a <4 magnitude quake validated his work? And how frequently do earthquakes of mag <4 to >7 occur in that part of Japan?

I think the odds just got rather shorter.