Can earthquakes be predicted?

Total Pageviews

Search This Blog

Friday, August 09, 2013

Earthquake prediction Evaluation- a physical check

hi
  I have been trying to figure out , the results of my prediction, statistically Roger is kind enough to help me in the process by ,special programs devised by him. However, it is some time becomes little confusing as to what the results suggests
  hence ,I have started to find it out ,with a hard ,laborious, physical verification.  It is really tough ,if one tries to figure it out ,by actual verification
  As a simplest and most crude method,  I took only two criteria ( ie part B  of trigger)
 The only two triggers i took were
1) Moon closest
2) Moon at maximum declination ( both north and south)

 Following criteria taken
 a) period 01.01.2005  to 31.12.2012
b) Date on which Moon Is closest taken
c) + or - one day taken for Maximum Moon declination

 No other criteria taken ,to make it simple for physical verification

 1) total days---------2880
 2) total quakes of 6.5+------389 ( actual quakes occurred)
3)  windows selected as per criteria-----532

hits
6.5+--------------108
7+........................41
7.5+.....................15
8+........................3


Actual quakes during selected period

6.5+........................389
7+...........................128
7.5+.......................44
8+..........................10

 It is up to readers to find out ,as to whether it is below or above average

please note
 This is very crude method , as i have not taken all criteria which I consider while giving dates. Roger have considered more criteria ( only left out was fine tuning and some aspects)
  Also, this is based on actual calculations, it is lengthy and laborious, There are chances that, one or two are missed or wrongly taken. But by and large it is fairly accurates

 Readers comments are expected
 Amit

56 comments:

Roger Hunter said...

Amit;

I've wondered what happened to you but if you did all that by hand I can understand your absence.

Where are you getting your quakes? I use the NEIC catalog. I only find 349 mag 6.5+ during that period.

My method of finding windows is to look for cases when minimum distance is with 3 days of min or max declination and there are only 31 instances of that and only 7 of them have a quake within.

If you can explain what you're doing I will replicate it.

Roger

Roger Hunter said...

Amit;

If you count each instance of max dec, min dec and min dist as predictors there are still only 319 of them.

Chance is ,312 for a 3 day window so over 100 hits would be expected.

Where did you get 532 predictions?

Roger

AMIT said...

Roger
I use European seismological data

http://www.emsc-csem.org/Earthquake/?filter=yes

It shows 389 quakes of 6.5+ for 1.12005 to 31.12.2012. I do not know why both differs.
If we count maximum moon declination for these years it is like this
2005...............26
2006...............26
2007...............26
2008...............28
2009...............27
2010...............26
2011...............27
2012...............27
Thus total would be 213 ( three day windows)
532 was wrongly considered ( including perigee of Moon) which comes out 105 for the period
Amit

Roger Hunter said...

Amit;

I count 214 declinations but only 73 hits.

Just chance.

Roger

AMIT said...

Roger
one more test for the same period

1)same period 1.1.2005 to 31.12.2012
2)Jupiter and Venus closest dates taken
3) 15 days on either side of closest dates as period to be considered
4) Only day of Moon maximum declination taken
results are
1) Actual quakes ....44
2)35 single day window
3)3 date hits for 7.5+

odds are 44/2880 days ie 0.0152 ie 1.5%
35*0.0152=0.532 less than one 7.5+ quake
actual quakes 3 ie 6 times odds
Amit

Roger Hunter said...

Amit;

I get 41 quakes, 13 predictions and 1 hit.

That's better than chance but not enough better to pass the test.

How did you get 35 predictions?

I looked for Venus and Jupiter distance minimums then checked 15 days either way for a moon trigger.

Did you take more than one moon trigger per planet minimum?

Roger

AMIT said...

Roger
yes.At least two moon maximm decination positions.sometimes three.
hence two or three triggers per planet closest position
Every 14/15 day mon at max declination.hence atleat two
Amit

Roger Hunter said...

Amit;

This one was harder than I thought.

There were 41 mag 7.5+ quakes. The rules created 29 predictions but only 1 hit.

That's slightly better than chance.

Roger

AMIT said...

Roger
7.5+ means 7.5 and above
There are three quakes
15.06.2006..........7.5
27.01.2006........7.6
8.8.2007............7.5
Amit

Roger Hunter said...

Amit;

There was a quake on 2006/1/27 but the moon was on 1/26; miss

There was no quake on 2006/6/15; miss

The quake on 2007/8/8 was a hit, the only one.

Roger

AMIT said...

Roger
Here are the exact (hit)figures (other than the hit you have suggested)

1)27th January 2006
quake of 7.6 at Banda sea at 16.58 UTC AND Moon at maximum declinations at 00.14 UTC (- 28.492)

2)16Th May 2006 ( wrongly mentioned as 15.06.2006)
Quake of 7.5 at Kermadic Island at 10.39 UTC
and Moon at Maximum declination at 03.52 UTC (-28.539)
Thua there are three hits (including 8.8.2007)
So,this is more than five times the average odds
Amit

Roger Hunter said...

Amit;

You say perigee at 2006/1/27, I said 1/26. NASA says 1/30 so we're both wrong.

NEIC shows no quake on 5/16.

But suppose you were right. What good is it top make 29 predictions and only get 2 or 3 hits?

Who would believe you?

Roger

AMIT said...

Roger
It is max declination and not perigee.
I am not interested in making people believe.
The whole issue is the statment " My predictions. are below averages" isincorrect.

In fact they are five time more accurate than averages.


Now I am convincwd that the hits are several times more thanaverages.

Amit

AMIT said...

Rogee
In fact from my postigs over years , one can find
several date hits.
Amit

Roger Hunter said...

Amit;

That's the wrong way to look at it.

The fact is that if you repeated this 100 times, there'd be a couple of cases that were as good or better.

It's just chance. And too small a sample anyway. With small samples, one or two hits can make a big difference.

Roger

AMIT said...

Roger
This is 8 year study.
Now the questions are
1) How much big is big data
2)How many times more accurate,(than averages), predictions. shoul be

Amit

Roger Hunter said...

Amit;

I can't give you a simple answer. It depends on the probability of success.

I generally like to see as many samples as possible. Generally 100 would be a minimum; more is better.

Roger

Roger Hunter said...

Amit;

Another thing to consider is the usefulness.

Suppose you found something that was always true for mag 7+ quakes but made 100 predictions for 5 hits.

You'd be wrong 95% of the time. Who would listen?

You don't know where, so what good is it anyway?

Roger

AMIT said...

Roger
As per plate tectonics annd sesmology,we will never be able to predict cent percent accuracy.
Why are we wasting tax players huge money over research
which predictions are cent percent?
whether report can not say precisely how many cms of rain wiil come on a day at specified location.
Amit

Roger Hunter said...

Amit;

If the weatherman was wrong 95% of the time, no one would listen to him either.

Who would evacuate a city if there was a 95% chance of failure?

A quake prediction needs to be 100% correct. Lives are at stake and lives can be lost just in evacuating so you can't afford to be wrong.

Roger

Anonymous said...

Other than general interest and self gratification, no government is going to ever announce earthquakes.....
As said before, life has to go on regardless, money has to be made....


The governments have the best on the case, they know and a lot more
then we will ever know.

They spend the money because they can, nothing you or anyone says will change that unless they want to.

Be safe

Jazzman :)

Roger Hunter said...

Amit;

Thanks to you, I have improved my Julian Day Number calculations so that days agree with NASA and JPL within minutes (and I'm working on that)

So now I can say with confidence that your last method gets 29 predictions with only one being a hit.

Below chance.

Roger

Roger Hunter said...

Amit;

I expanded the search to 1993-2012 and found 71 predictions but only 2 hits, pretty close to chance.

Any new ideas?

Roger

AMIT said...

Roger
Three hits have already been shown above ,between 2005 to 2012.
I have not worked on data since 1993, but can be done ,with some efforts
Amit

Roger Hunter said...

Amit;

You're claiming a hit for a non-existent quake.

Roger

Roger Hunter said...

Amit;

You may get the feeling that I am persecuting you with these tests.

Nothing could be further from the truth.

I would love to see you succeed. My tests are intended to keep you from wasting your time on a false method.

There's no need for you to spend weeks testing ideas "by hand" when I can do it in hours.

Roger

AMIT said...

Roger
pl see the link

http://www.emsc-csem.org/Earthquake/?filter=yes

pl note the quake of 7.5 on 16.05.2006

and the link for Moon max declination

http://astropixels.com/ephemeris/moon/lunarstandstill2001.html

I am the last person to claim wrong hits. If I am convinced that, my prediction is not at par with odds ,I will quit myself
Now, what I can doe for two catalogs not matching.,ie to USGS and European Seismological center

I suppose, Petra was talking of this catalog errors, all those years.
How can science and seismology, ask for cent percent accuracy in prediction ,when two of them do not match ,yaers after a quake ,as big as 7.5 ,have already occurred ?
Does it not indicate that, Plate tectonic theory is being safeguarded by science?
Amit

AMIT said...

Roger
Two more off hand are here
1)26.12.2004...9.1 quake on Max Declination day

2) 27.02.22010.....8.8 on perigee day

Amit

Roger Hunter said...

Amit

> I am the last person to claim wrong hits. If I am convinced that, my prediction is not at par with odds ,I will quit myself

That's the proper attitude.

> Now, what I can doe for two catalogs not matching.,ie to USGS and European Seismological center

Good question. Mag is an average and depends on the readings available. The best one can do is to be consistant. I use the NEIC catalog which listed that quake as 7.4

> I suppose, Petra was talking of this catalog errors, all those years.

Yes. But consider, if everything hinges on one or two quakes, you're too close to the edge anyway.

> How can science and seismology, ask for cent percent accuracy in prediction ,when two of them do not match ,yaers after a quake ,as big as 7.5 ,have already occurred ?

Nothing is perfect but some things are more important than others.

> Does it not indicate that, Plate tectonic theory is being safeguarded by science?

No it doesn't.

> Two more off hand are here
1)26.12.2004...9.1 quake on Max Declination day

But no Venus or Jupiter around to start the process.

> 2) 27.02.22010.....8.8 on perigee day

Same thing. No prediction declared.

If you take every moon indicator the odds will overwhelm you

Roger

Roger Hunter said...

Amit;

This is interesting. In checking your source against mine I ind a one day difference. Both sources are reliable so I don't know how to account for the difference.

I do not understand the "standstill" concept well enough to say if that's the problem but the declination is larger on 2004/12/27 than on 12/26 according to ICE (Interactive Computer Ephemeris) that I use.

Roger

Roger Hunter said...

Amit;

I found the error.

My program reads files of planetary positions at the beginning of each day. It finds peaks and troughs in declination and distance by comparing 3 successive values. If the 2nd is larger, it's a peak and vice-versa.

This is fine for everything but the moon. It moves so fast that a peak or trough can occur during a single day with the next day being larger than the beginning of the second day, causing the peak to appear to be in the 3rd day.

I should have known better and I apologize for it. I'll use the tables you use, since hourly values would make the file HUGE!

More later.

Roger

AMIT said...

Roger
There were three criteria for software cconsideration Jupiter, venus and Sun at closest position On 26.12.2004 Sun was at 0.983 AU
In fact every year Sun is closest in Dec last week to Jan first week
If we consider. 15 day either side and Moon triger , we may find more major quakws during this period( This is my openion , I have not physicaly checked.
Amit

Roger Hunter said...

Amit;

Ok, I'll put the sun back in after I get the moon straightened out.

I can't use the tables you did, they don't go back far enough.

This may take some time...

Roger

Roger Hunter said...

Amit;

I couldn't find any older declination tables so I computed hourly values and extracted the maxima from them.

It didn't help. I did 1993-2012 and found only 2 hits.

Next I'll add the sun distance and see what that does.

Roger

Roger Hunter said...

Amit;

here are the results with sun distance added.

There were 115 quakes of the requested size. (7.5+)
There were 117 predictions ( 117 days or 3.2 % )
There were 2 hits out of 114 quakes ( 1.8 %)
Out of 117 predictions, there were 2 hits which is 1.7 %.
The odds on a 7.5 + quake in a 1 day window are 1.6 %

It made it worse; more predictions, no more hits.

Roger

AMIT said...

Roger
For which duration these results are?
Amit

Roger Hunter said...

Amit;

1993-2012, mag 7.5+, 1 day windows.

Roger

Roger Hunter said...

Amit;

The problem you're facing is that astronomical factors are regular. The lunar cycle is 28 days, the day is 24 hours, etc.

But quakes are random. Sometimes the two will correlate but that's all it is. Not causation, just occasional correlation.

Roger

AMIT said...

Roger
A few I can give here

dd-mm-yyyy mag
1)16.05.2006.........7.5
2)27.01.2006.........7.6
3)08.08.2007.........7.5
4)26.12.2004.........9.5
5)28.12.1994........7.8
6)04.01.1998........7.5

If I do physical check ,I may get few more

Amit

Roger Hunter said...

Amit;

Are you claiming these as hits?

Roger

Roger Hunter said...

Amit;

1994/12/28 Moon was 12/31
1998/1/4 Sun was 1/5
2004/12/26 Hit #1
2006/1/27 Moon was 1/26
2006/5/16 Quake was 7.4
2007/8/8 Hit #2

We're not agreeing with what was where when and we're both quoting authoritative sources.

I don't know how to resolve this.

Roger

AMIT said...

Roger
one more issue is that, the quakes merely apart by 3 hrs from perigee hour of Moon are some times counted in different dates. I will give you a classic example

Quake....on 3rd November 2002, at 22.12 UTC..OF MAG 7.8 AT CENTRAL ALASKA

Now Venus is at 0.272 AU ie ie well with in 15 day margin of minimum distance

Moon at Perigee on 4th November at 00.50 UTC
Now the diff between the quake and perigee is hardly 3-4 hrs,but this is not a hit ( as per your software ).
Actually ,software should calculate the hts difference of + or - 12 hrs ,for a day change
How is the software going to deal with this.
This is only one sample I have shown you, there could be others also
Amit

AMIT said...

Roger
Now take date 4th Jan 1998...7.5 quake
sun closest ..
Venus closest ( both are closest in 15 day period )
Moon at 57.0 ER ie closest
what else is required to consider it as a hit
Roger, I have such examples ,which prompt me to believe, the software can some time miss judge the hits

Amit

Roger Hunter said...

Amit;

> one more issue is that, the quakes merely apart by 3 hrs from perigee hour of Moon are some times counted in different dates. I will give you a classic example

Sure. Could be minutes apart if both near midnight.

> Quake....on 3rd November 2002, at 22.12 UTC..OF MAG 7.8 AT CENTRAL ALASKA

> Now Venus is at 0.272 AU ie ie well with in 15 day margin of minimum distance

> Moon at Perigee on 4th November at 00.50 UTC
Now the diff between the quake and perigee is hardly 3-4 hrs,but this is not a hit ( as per your software ).

Correct.

> Actually ,software should calculate the hts difference of + or - 12 hrs ,for a day change

> How is the software going to deal with this.

It can be done but it's getting unreasonable to be that fussy over time. You could use a bigger window but the odds go up accordingly.

> This is only one sample I have shown you, there could be others also.

Always possible. Calculations are done to the DAY things happen. Taking it to the hour multiplies tables by 24x.

As to the next post:

The quake was on Jan 4

Sun was closest on Jan 5
Venus closest on Jan 16

Moon dec lowest 1997 Dec 29
Moon dec highest 1998 Jan 11

One day predictions on moon days only.

Not a hit

Roger

Roger Hunter said...

Amit;

I think I can do this.

I have moon to the nearest hour and quake to the nearest minute so I can compare the two and call a hit within +/- 12 hours.

The planets should be ok to the nearest day since they don't change much.

Roger

AMIT said...

Roger
4th Jan 1998 quake have all the required potential and trigger

sun and Venus closest. please note here that, what we are taking is + or - 15 days period of closest position of Sun , Venus OR Sun. It is not necessary that on the day of quake (+ or - one day) these need to be closest
Moreover, Moon at Perigee on the same day
AMIT

Roger Hunter said...

Amit;

I'm considering sun,venus and jupiter as indicators and moon dec and distance as triggers. I also added mars for good measure (it didn't help)

A hit happens if the quake is within +/- 12 hours of the moon trigger

1998 does not qualify; there's no moon trigger anywhere near the quake.

There were 4 hits total in 20 years

Roger

Roger Hunter said...

Amt;

Perigee was 1998/1/3 at 14:00

Quake was on 1/4 at 6:12

Roger

AMIT said...

Roger
pl see the link

http://www.fourmilab.ch/cgi-bin/Solar
pl also see the distance of moon on 3 rd and 4th both.find for yourself ,when moon is at 57.00 ER or at perigee
AMIT

Roger Hunter said...

Amit;

Here's a better one.

http://www.fourmilab.ch/earthview/pacalc.html

Roger

Mr. Tony said...

Hi just testing

Mr. Tony said...

Hi there I have not posted here for some time. I have also developped my way of predicting Earthquakes and I think it is pretty good. I wonder if I can post in here some of my predictions too so as we can compare. I have done a lot of stats, but generally I have predicted within a day (as I day error is perfectly reasonable due to the fact people do not count the hour difference between earthquakes which accumulates to one and sometimes 2 days. Anyway I do not want to spoil your fun, but let me know. Cheers Be Good!

AMIT said...

Tony
Yes. You are most welcome to post your prediction on my blog. In fact,earlier I have invited one and all to give random dates and compare with my prediction,so as to evaluate them wrt to random dates
Amit

Roger Hunter said...

Hi Tony;

I can save you some time if your method is something calculable, as for example Amit's planetary position method.

I'll be happy to put it into a program and run it against all the quakes in the NEIC catalog. Then you'll know right away if you have something worthwhile.

Roger

Roger Hunter said...

Amit;

I'll extend that offer to anyone who has a set of rules for determining when an earthquake may be expected.

Roger

mvcejas said...

You guys are cool. I will be up to date to this site :)