Seismology for Dummies: Guesstimating nuke test yields

shiv · Post by **shiv** » 02 Jul 2002 20:09

SEISMOLOGY FOR DUMMIES: GUESSTIMATING NUCLEAR TEST YIELDS

Misinformation. or misleading information, like scum floating up in a still pool has a way of becoming the only available information if an attempt is not made to point out that other viewpoints exist.

When it comes to the question of estimating the yields of the Indian nuclear tests, there is a whole lot of information available, from a plethora of sources, but the only two names whose work gets quoted and propagated is one [url=http://www.geo.arizona.edu/geophysics/f ... Wallace</a> and a <a href="http://www.iisc.ernet.in/~currsci/jul10 ... Douglas</a>. I see this situation as akin to the "selling" of Chuck Yeager's "Invincible PAF" even after decades of dismal performance and defeats.

I want to add to the efforts of a <a href="http://www.bharat-rakshak.com/MONITOR/I ... .html">few concerned people</a> by writing something that can hopefully be understood by the seismic non techie, and by a person who can only understand as much math as I do, - i.e. elementary, high school maths.

MEASURING YIELDS OF UNDERGROUND NUCLEAR TESTS:

First let us get some things straight. The ONLY RELIABLE methods of estimating nuclear test yields (as per <a href="http://tejas.serc.iisc.ernet.in/currsci ... .pdf">open sources</a>) is to have instruments close to the test site, and post-test radiochemical testing AT THE TEST SITE. Anything else can be way way out.

But what if you are not invited to come near a nuclear test site? That is where seismic methods come in. Since most people are not allowed near nuclear test sites, the whole "scientific", "debating" and gossiping world depends on the speculation of seismologists reading squiggles caused on instruments thousands of kilometers from a test site. This fact, more than accuracy, makes seismology important.

SEISMOLOGY AND NUCLEAR TESTING:
According to one dictionary Seismology is the science that studies earthquakes. It so happens that underground nuclear explosions also cause little squiggles in the instruments that are set up to study earthquakes (seismographs), putting seismologists in the unique position of talking about nuclear tests.

So how do people arrive at the yield of a nuclear test from squiggles on seismographs?

The "magic formula" is given below:
<hr>
mb = attenuation constant + 0.75(Log of Yield)

or

mb = a + 0.75 Log Y

<hr>

For this formula, "mb" is measured from the squiggle on the seismograph, and "Y" is the Yield of the nuclear test that the magic formula will reveal after you fill in the value of "a".

Now what the hell is "a"? "a" is a number that is supposed to indicate the amount by which the seismic signal of the explosion has petered out as is gets to you. The value can be anywhere from <a href="http://www.geo.arizona.edu/geophysics/f ... .pak/">3.9 to 4.5</a>

Now that is really really funny, and its called science. If some seismographs indicate a value of 5.2 for mb you apply the magic formula and get a nuclear yield of 53 kilotons if you use a value of "a" as 3.9, and the SAME NUCLEAR EXPLOSION will show a yield of only 8 kilotons if you use the value 4.5 for "a". How convenient.

When it comes to using a value for "a" in calculating Indian Test Yields, let me quote a <a href="http://www.iisc.ernet.in/~currsci/jul10 ... ">paper</a>:

in the value of a: Sikka et al use 4.04 on the assumption that the yield of the first Indian test (18 May 1974 referred to below as 740518) was 13 kt, Wallace uses 4.45 on the assumption that a for Pokhran is the same as that of eastern Kazakhstan . In fact I have not cross checked to see if these alleged Kazakh values are accurate.

IT GETS WORSE THAN THIS:

Now all that is written above is just great for single shot nuke tests. Now what if some sneaky people set off two or more nuclear tests simultaneously? The seismologists of this world have next to zero experience of seismic signals from nuclear tests set off simultaneoulsy. And that was what was done on May 11th at Pokhran.

OK - if nobody has PRACTICAL experience, what does the theory say? What's in the textbook?

When two sources situated close together produce waves of any kind the waves interfere with each other and may add up or subtract from each other depending on the direction. Imagine a "double simultaneous nuclear test" conducted by Pakistan midway between Lahore and Islamabad 100 meters below the Lahore-Islamabad expressway. The people in Lahore and in Islamabad who are in line with the two devices will experience interference such that the signals will be weaker than the total energy of the two tests, but those of us watching from Jammu, at right angles to the road, will see the two test signals add up. That is the theory. What does the little practical experience available say?

In fact, the practical experience says that the theory is correct. East-West seismic stations have recorded smaller signals from the East-West oriented Pokhran tests, as opposed to stations situated North and South. This can be seen in the tables in <a href="a href="http://www.iisc.ernet.in/~currsci/jul10 ... .pdf">this paper</a> and in the <a href="http://www.bharat-rakshak.com/MONITOR/I ... a.html">BR paper</a>. The same thing is <a href="http://www.bharat-rakshak.com/MONITOR/I ... ">depicted grahically</a> on BR.

When such interference occurs - adding up the signals from different seismic stations around the world and taking an "average value" to calculate the yield is utter nonsense. Garbage in, garbage out. But what can we say? "Seismology experts" understand rocks, but perhaps not physics or maths.

I'll stop here, but I will leave you with some relevant notes and references (some are repeats, so check before you click)

<hr>

1)The inaccuracy and problems associated with trying to detect and calculate yields from seismic signals has led to a continuing search for better methods - as shown by these two papers:

<a href="http://www.nemre.nn.doe.gov/review2001/ ... /04-03.pdf]http://www.nemre.nn.doe.gov/review2001/ ... /04-03.pdf[/url]
and
http://www.nemre.nn.doe.gov/review2000/3-Seismic%20Source%20Characterization/03-08.pdf

The latter paper acknowledges the existing problems of innacurate yield estimation by seismologic methods with the words: "to address a long-standing problem in observational seismology, namely accurate estimation of the radiated seismic energy"

These papers are quite difficult to read and digest - especially for the seismic non-techie, but I would advise that at least some should try and read them, if only to try and pick holes in my interpretation.

2)Another , also available in <a href="http://216.239.39.100/search?q=cache:0aCNIib3HqYC:www.princeton.edu/~globsec/publications/pdf/3_3-4Adushkin.pdf+seismic+attenuation+and+mb+values+in+nuclear+yield&hl=en&ie=UT F-8&e=619">html</a> lists some of the unknown variables that can affect yield measurements:

the nature of the relationship between mb BRV and yield is tied to the considerable effect by the depth of a charge on seismic efficiency. This is because higher yield explosions are carried out at greater depths. The relationship is also linked to any change, based on depth, in the physical, mechanical and gas-forming properties of the enclosing rocks, which determine the
action of an explosion on the surrounding environment. It is obvious that the depth of an explosion influences not only the lithostatic pressure, but also the stability, density, porosity and wave velocity of elastic waves and the moisture content of rocks. All of these properties have an integral effect on the transfer of the energy from an explosion to the creation of seismic waves.
3)This fits in well with a quote from <a href="http://tejas.serc.iisc.ernet.in/currsci/nov102000/1359.pdf">another source:</a>

An underground nuclear explosion sets up a shock wave near the point of detonation which interacts with the surrounding geological medium. This shock wave vaporizes, melts, plastically deforms and fractures the surrounding rocks and then degenerates into elastic waves. Only a
small portion of the total energy released by an under-ground explosion is converted into seismic waves. The ratio of the elastic wave energy to the total energy is known as seismic efficiency. The seismic efficiency depends on the physical characteristics of the surrounding medium and the source parameters.
4)<a href="http://www.barc.ernet.in/webpages/milestones/drs_03.html">Indian Explosions of May 11, 1998 : An Analysis of Global Seismic Bodywave Magnitude Estimates</a>S. K. Sikka, Falguni Roy and G. J. Nair, High Pressure Physics Division, Bhabha Atomic Research Centre, Mumbai 400085, India

5)BRM article <a href="http://www.bharat-rakshak.com/MONITOR/ISSUE3-6/Figure2.jpg">graph</a> of North South east west mb value variation

6)<a href="http://www.bharat-rakshak.com/MONITOR/ISSUE3-6/ramana.html">BRM article</a>

7)Preliminary yield reults
<a href="http://www.barc.ernet.in/webpages/milestones/drs_02.html]http://www.barc.ernet.in/webpages/milestones/drs_02.html

8)Yield estimation of Indian nuclear tests of 1998 - S. K. SIKKA,FALGUNI ROY, G. J. NAIR Seismology Division, Bhabha Atomic Research Centre, Mumbai 400 085, India
"Further,the ratios of neutron activation products to fission products for the thermo-nuclear
and fission test of 11 May 1998 were also found to be in quantitative agreement with that expected from their radiochemical yields. BARC has also performed simulation studies of the
close-in acceleration values and surface features by finite difference and finite element codes (unpublished). These studies also corroborate the estimated yield of POK2."
9)The specific yields of Pokhran 2 were measured by the methods mentioned <a href="http://tejas.serc.iisc.ernet.in/currsci ... f">here</a>:

the yield of a nuclear explosion can be determined with a reasonable accuracy from the close-in acceleration and velocity measurements. Com-parison of the close-in seismic data pertaining to POK2 with the available global data from similar geophysical environment gives an yield value close to 58 kt for these explosions (see Figure 10)3 . Recent radiochemical analysis of samples from post-shot logging of POK2 32 has confirmed the authenticity of the estimated seismic yields
of POK2 explosions. The radiochemical yield for the thermonuclear device has been obtained as 50 ± 10 kt. A preliminary estimate of the radiochemical yield of the fission device, is 13 ± 3 kt (ref. 33). These studies have further substantiated that the fusion component was in accord with computer simulations.

10)From the Douglas paper:
[URL=http://www.iisc.ernet.in/~currsci/jul102001/72.pdf]http://www.iisc.ernet.in/~currsci/jul102001/72.pdf</a>

The principal difference in the yield estimate of Sikka et al.1 and others is in the value of a: Sikka et al use 4.04 on the assumption that the yield of the first Indian test (18 May 1974 referred to below as 740518) was 13 kt, Wallace 6 uses 4.45 on the assumption that a for Pokhran is the same as that of eastern Kazakhstan.

Tim · Post by **Tim** » 02 Jul 2002 20:35

Shiv,

As a non-physicist, my eye was drawn to the statement that Sikka et. al. used an "a" of 4.04 "on the assumption that the Pokhran I yield was 13 kilotons."

NOTE: what follows is purely theoretical and based entirely on a precursory textual analysis. I have not read Sikka et. al., and frankly don't have time to right now. However....

It might be worth digging through those papers to see if Sikka et. al. made an "assumption" - which implies a certain level of guesswork, not unlike what Wallace did - or were based on actual data.

There remain questions at quite high levels about the yield of Pokhran I, which has been questioned not only by Wallace but also by Indian scientists (see Perkovich) - the initial yield has been estimated at between 6 and 13 kilotons by a variety of different sources, and not all Indians agree with the 13 kiloton figure.

If, therefore, Sikka et. al. are basing their estimates on the "assumption" that the 13 kiloton yield initially claimed was accurate, they may have a large margin of error which would impact their later analysis. If they aren't using Pokhran I DATA but instead are making mathematical equations based on simple acceptance of the 13 kt claim, they are starting with a potential 100+% margin of error, and then building their equations on that.

OTOH, if they have based their estimates on the actual data, this would make their estimates much more convincing and would also confirm the initial announcements about Pokhran-1 yield - a point that would bear emphasizing in its own right.

Just a thought. There may be a future article in this for someone.

tim

Umrao · Post by **Umrao** » 02 Jul 2002 20:39

Excellent post, the first time I have seen Fourier Series, Log equations and the theory of intereference (which was lost on Messers (

) Wallace and Douglost) is easily explained.

The equation's constant seems to be have been made a variable of fancy by Wallace, and I dont find fault with him because if you really question Wallace he may not even know where the new madrid fault starts and ends (In all probabilty, if it suits him he will say it starts in Spain and ends in applachian mountains!!

without knowingly )

Sunil · Post by **Sunil** » 02 Jul 2002 20:43

I think the hardest part to explain to people is the idea of interference.

Most people do not immediately realize the strange nature of a wave.

The closest example in daily experience is the effects seen on the surface when we drop two stones simultaneously into a pond.

here is a picture of the interference pattern on the surface of water.

http://mediatheek.thinkquest.nl/~ll122/en/br-interf.shtml

there are a lot of java applets on the web that describe the basic physics of waves on the surface of the water and show the interference patterns.

http://www.physicsclassroom.com/Class/light/U12L1b.html

http://www.colorado.edu/physics/2000/schroedinger/

http://landau1.phys.virginia.edu/~snp9b/java/Ripple.html

http://www.bekkoame.ne.jp/~kamikawa/wave/wave_e.htm

The thing to point out is the alternating ridges and valleys that form on the surface and then explain that if the source of the waves was a nuclear explosion, the ridges would correspond to a high value on the seismograph and the valleys would correspond to a low reading.

At this point one also has to point out there is reflection and refraction from of a wave from a surface. Therefore is there is an obstruction in the path of the wave, then one has to account for that as it will change the shape of the pattern of ridges and valleys.

shiv · Post by **shiv** » 02 Jul 2002 20:48

Originally posted by Tim:

As a non-physicist, my eye was drawn to the statement that Sikka et. al. used an "a" of 4.04 "on the assumption that the Pokhran I yield was 13 kilotons."

The quote about "assumption" is from Douglas' paper - I hope the link is working - can't check now but will correct it later.

I am assuming that Sikka saw the actual yield of P1 - 1974 (from accelerometers etc). a is a figure that a seismologist can figure out I guess from actual rock samples and seismic squiggles in local Indian stations (Gauribidanur is not on CTBT circuit I think)

If you take yield as 13 KT and mb as 5.17, you will arrive at a figure of 4.04 for a.

Umrao · Post by **Umrao** » 02 Jul 2002 20:59

By knowing the orientation of the simultaneous explosions and the desired effect of (BARC) scientists to dampen the wave form thereby generated, can we fit a emperical equation with varying degrees of damping efficiency and then arrive at the consequent values of "a".

Because the value of "a" as arrived at (4.04) based on one single 1974 explosion, that too for the first time, is rather difficult to sell.

On the contrary is it possible to estimate the value of "a" by non nuclear explosions in the same site and also use those explosions to calibearte the equipment??

Has such caliberation of seismic equipment ever used elsewhere to determine the wave propagation characteristics. I know for oil exploration they do drill holes and dynamite the ground to read the wave form to determine oil bearing content.

In other words is there a way (or was it done )to establish the value "a" at 4.04 other than the sole nuke explosion in 1974.?
If it was done then great!!

bala · Post by **bala** » 02 Jul 2002 21:44

The question of actual yield is determined by chemical analysis of samples
obtained from drilling on site. However, the disclosure of the details also
implies clues to the nature of the device. BARC has a paper on the chemical
tests without going into actual details. Sufice to say, that a variety of techniques
including seismic are needed to determine actual yields.

On the seismic yield, there is a paper from an England based scientist that claims
that the stated yield of POK-2 is close to 60K tons when all inputs across the globe
are taken into consideration. Hence the only fellows propagating falsehood and
propaganda are the so-called US seismic experts. This is similar to the assertion
that Al-Qaeda is not in TSP Occupied Kashmir.

Kati · Post by **Kati** » 02 Jul 2002 22:00

For Shiv and others:

Estimation of Nuke test yields is based on, what is known as Inverse regression in Mathematical Statistics (or sometimes known as `Calibration Problem'), several model assumptions.

Yes, the value of `a' (4.04 or 1.1) can greatly affect the estimation procedure, and it depends on

1. Mathematical-Statistical Model assumptions like- (a)Gaussian error model,
(b) `homoscadasticity' of the model
(c) uncorrelatedness of the errors; etc

2. Robustness of the model

3. The optimality criteria (absolute error loss or squared error loss) in estimating the model parameters;

4. Validation procedures of the model; etc.

Usually, after collecting data, we the statisticians play with it, look at many possibilities, and then come-up with the most plausible model to explain the data. Believe me,
give me a data-set and I can make you confused
by varying model assumptions slightly.

Unfortunately, the atomic scientists (like other applied researchers) overlook the mathematical complexities involved in statistical
procedures and uses the ready-made standard class-room procedures, and that affects the analyses greatly. I personally went through volumes of publications (from various journals) and was apalled by the adhocness in statistical data analyses. These researchers did put a lot of energy in doing the other math parts (in nuclear physics) very elaborately, but made short-cuts when it came to statistical model fitting and subsequent analyses.

The spacecraft challanger (Isn't it?) blew-up because of a wrong stat model used.

Umrao · Post by **Umrao** » 02 Jul 2002 22:22

nabendu>> The challenger disaster happened because of design flaw in the metallic "O" rings, the design was centered around the concept that the "O" rings would expand due to temperature rise as the engines ignited and make it leak proof seal (which is normal design ) but the ambient temps were so low (lower than the design threashold on that fatefull day) that the pulme and heat leaked into places undesirable.

Yes statistically the (modeling) design might have given the wrong range of possible coefficients of expansion for the "O" rings because of the faulty rancge selection of possible ambient temp for the launch site (time & location).

Also the "O" ring design was subsequently changed to compress seal the parts incase of temp drops.

Prateek · Post by **Prateek** » 02 Jul 2002 22:36

WoW, this should have been posted in the technology forum, what is this doing in the strategic issue forum !

Just kidding!

Roop · Post by **Roop** » 02 Jul 2002 22:39

The spacecraft challanger (Isn't it?) blew-up because of a wrong stat model used.

Say what?!! The Challenger blew up because of faulty (i.e. leaking) O-rings. Where did the stat model come in?

Sunil · Post by **Sunil** » 02 Jul 2002 22:42

Hi,

AFAIK, the statistical aspects apply only to the measurement at a given site. The errors at each site have to be fitted to the appropriate statistical model.

The analysis of the yeild based on various measurements does not involve any statistical argument. It rests solely on the application of the *right* physics. The change in seismograph reading as a result of a intereference effect is quite in excess of the instrument error.

What is however quite painfully clear is that neither Douglas nor Wallace bother to propagate their errors. The result is a series of "yields" without error bounds, which may make for excellent coffee table conversation but reflect poorly on their understanding of physics.

Generally speaking, the statistical nature of measurement is included in an error term in the instrument reading and then physics begins. Now if an incorrect model of error is applied to the instrument then yes most definetely erroneous readings will result and the physics applied there will be imperfect. <B> But there is nothing here to suggest that the Indian seismic community (or for that matter the international seismic community) is prone to applying incorrect statistical models to their seismographs. [/b]

The presentation of `their motivations' does *not* constitute a real discussion on the merits of their measurements.

The discussion on the statistics of the seismic measurements is out of the scope of Shiv's post.

The claim here is that an inappropriate physics model has been applied by Wallace and Co. to the analysis of POK 2.

member_4218 · Post by **member_4218** » 02 Jul 2002 23:01

Seems like Deja-vu but we had a similar discussion some time ago and it was pointed out that in

mb = a + 0.75LogY,

the "a" can be factored out in case of POK because of data from the 74 test. But then Sunil Sainis pointed out that the ratio will not make sense because of instrumentation errors.

Incidentally after the previous discussion I exchanged a few emails with Prof. Wallace and he related an interesting ancedote:

Apparently in the early 80's the US nuclear establishment was trying to deliberately "overestimate" the yields of Soviet tests in order to get funding to carry out more elaborate tests. Terry Wallace and his colleagues were able to "convince" Congress that Soviet yields were lower than what the US establishement were quoting.

Umrao · Post by **Umrao** » 02 Jul 2002 23:32

I just got my subway and muching away I realise the constant "a" as used must be correct value of 4.04, and could have been arrived after numerous explosions with in very close proximity to the actual test sites. It is not based on 74 explosion alone. The reason being if the underground tunnels were to be blasted with dynamites to create the tunnels they must have used those occasions to calibrate the instrumentation and as well as as run emperical models to nearly zero down on the correct value of "a" at thos (proximate) distances. This could have been done in 1974 (as the shafts were being blasted ) as well for the POK II.

We do not have the situation US where in they always "Exaggerated" the SU threat to get funding.
SO even to suggest obliquely BARC would tamper with "a" yield deliberately is something I can not conjure.

Kati · Post by **Kati** » 02 Jul 2002 23:33

Two things:

1. The challanger case: estimating the stress that a particular metal (or material) can withstand is an old statistical problem. In general, the life time of a material is modelled by various probability models, like exponential model, gamma model, Weibull model etc. Inappropriate modelling of lifetime of a material
or a machine can cause disaster. That's why rocket or missile scientists spend a lot of time in testing components and fitting right model.

2. Back to nuclear yield estimation: The statistical theory of Inverse Regression (or Calibration) got a boost in mid eighties when CTBT
was hot and both the US and USSR were trying to determine each other's yield. Normally the data is collected at various measuring stations (listening posts) and these shock waves travel through different types of soils, rocks. Also the distance is a factor, so are the type of radio-active materials. Using all these, statisticians try to build a model which is supposed to describe the measured values (at measuring stations) at the best possible way using the existing own data on yields which are already known. Then do the back calculations using the Regression model to guess the actual yield. Usually the yield is accompanied by a standard error (SE) which can be used to get a 90% or 95% confidence interval. Statistical hypothesis testing is also done to verify whether a test had yield less than some agreed value (accepted by both sides).

Sunil · Post by **Sunil** » 03 Jul 2002 00:03

schawla:

> Seems like Deja-vu but we had a similar discussion some time ago

Oh I plan to make it a lot more than just a deja vu..

> and it was pointed out that in mb = a + 0.75LogY,

the "a" can be factored out in case of POK because of data from the 74 test. But then Sunil Sainis pointed out that the ratio will not make sense because of instrumentation errors.

No, thats *NOT* what i said. I said that comparison of yeilds method using the POK1 and POK2 readouts without appropriate error propagation was incomplete.

Such an approach attempts to solve an equation of the sort:

(mb_pok2-mb_pok1) = Log (Y_pok2/Y_pok1)

the `instrument' error is only the error in the (mb_pok2-mb_pok1) part. The other error is in the Y_pok1 term.

Assuming that the error in the (mb_pok2-mb_pok1) is small, the main source of error is the Y_pok1.

Which is what wallace and co haven't bothered to take into consideration.

> Incidentally after the previous discussion I exchanged a few emails with Prof. Wallace and he related an interesting ancedote:

I am so happy for you..

> Apparently in the early 80's the US nuclear establishment was trying to
deliberately "overestimate" the yields of Soviet tests in order to get funding to carry out more elaborate tests. Terry Wallace and his colleagues were able to "convince" Congress that Soviet yields were lower than what the US establishement were quoting.

Good for him, what relevance that has to the issue of the Indian tests i fail to grasp.

If Wallace is claiming that the US Govt. inflated the Soviet test results and ergo India has inflated its own test results, and as terry wallace showed the `real' results of soviet tests, the results he shows of the Indian tests are automatically `real'.. then i think that is a most absurd form of logic.

Sunil · Post by **Sunil** » 03 Jul 2002 00:08

Hi Nabendu,

Do you have a specific example where the validity of the statistical model used by the seismic community (in India or internationally) has been questioned?

member_4218 · Post by **member_4218** » 03 Jul 2002 00:27

Originally posted by sunil sainis:

> error in the (mb_pok2-mb_pok1) is small, the main source of error is the Y_pok1.

Is not Y_pok1 declassified ?

Good for him, what relevance that has to the issue of the Indian tests i fail to grasp.

If Wallace is claiming that the US Govt. inflated the Soviet test results and ergo India has inflated its own test results, and as terry wallace showed the `real' results of soviet tests, the results he shows of the Indian tests are automatically `real'.. then i think that is a most absurd form of logic.

The relevance is that he may not be in "bed" with the US establishment(as he has crossed swords with them too) as part of a concerted effort to deliberately question Indian test results.

Kati · Post by **Kati** » 03 Jul 2002 00:29

Yes, we looked into a few data-sets (not related to the Indian tests) where things could have been done in a better way. I can give you specific references in mid August after I'm back from vacation.

In the meantime take a look at this article:
http://www.llnl.gov/str/Walter.html

Kati · Post by **Kati** » 03 Jul 2002 00:43

http://www.ctbtcommission.org/blancpaper.htm

Check the second diagram (three scatter plots-red, blue and green). How well can we fit the scatter plots is the moot point. This is basically called `Regression'. A small change in curve fitting can lead to great deal of differences in final `inference'.

Sunil · Post by **Sunil** » 03 Jul 2002 01:30

Nabendu,

> Yes, we looked into a few data-sets (not related to the Indian tests) where things could have been done in a better way. I can give you specific references in mid August after I'm back from vacation

Cool, but before you leave could you tell us what the difference in error from the two statistical models was? thank you. I will go through the references you have provided.

schawla,

> Is not Y_pok1 declassified ?

Yes but how is that relevant? The main questions are:

what is the error on the western estimates of Y_pok1?

and how does this error propagate?

both are questions that Wallace etal seem not to want to address in their paper.

> The relevance is that he may not be in "bed" with the US establishment(as he has crossed swords with them too) as part of a concerted effort to deliberately question Indian test results.

I fail to see how you conclude this with near absolute certainity. Does it not seem plausible to you that precisely on account of his `crossing swords' with the USG earlier, Wallace could now have changed sides? or that in some complicated way Wallace and Co represent various factions within the USG and that the `debate' in the Congress was merely an internal turf battle?

All of this ofcourse has no bearing on what the accuracy of the physics used by Wallace and co. His relationship with the USG `in bed' or `out of bed' is not going to undo what is manifestly a flawed physical analysis.

Kati · Post by **Kati** » 03 Jul 2002 02:01

Sunil:
It is not a simple question of difference of errors. It was like approximating an apple by an orange.

The standard regression models assume equality of error variances (Called `homoscadasticity) and independence of individual errors. But when model was fitted and residual analysis was done (to check the asuumptions of the model), it indicated
heteroscadascity i.e., error variances were unequal. In such a case you have to make a transformation of the variable(s), output and possibly input, and then fit a regression model with transformed data. The resultant is a completely different model.

If a model is homoscadastic then the error standard deviation (square-root of the variance) is called the model error. But when the model itself is heteroscadastic then the concept of `model error', the way we understand it, is meaningless.

Sunil · Post by **Sunil** » 03 Jul 2002 02:28

Nabendu,

thats very interesting indeed.

so in the data you analysed, you felt that there was precisely such a change of model needed?

am i correct to assume that you feel the current way of estimating error on seismograph measurements is (at the very least) incorrect?

Kati · Post by **Kati** » 03 Jul 2002 02:44

Not totally.
Every time we have a data set, we need to play with it. Some times, the conventional methods work, sometimes not. if not then go thru careful transformations of data (popular one is- `Box-Cox
transformation') to see what makes the model assumptions valid.

member_4218 · Post by **member_4218** » 03 Jul 2002 06:23

Originally posted by sunil sainis:
Yes but how is that relevant? The main questions are:

what is the error on the western estimates of Y_pok1?

As I understand, Wallace used the published or announced Y_pok1 yield by the Indian scientists so there is no need for "western estimates of Y_pok1". This gave him the mb1 and Y_pok1. He then used the ratio formula along with mb2 to calculate Y_pok2.

shiv · Post by **shiv** » 03 Jul 2002 06:27

Sikka speaks of 1974 yields here:
http://www.barc.ernet.in/webpages/milestones/drs_01.html

Taking the yield as 12 to 13 kt for the 1974 explosion based on seismic data (4-5) and rock mechanics phenomenology calculations (6) which reproduce measured cavity radius, spall velocity and extent of rock facturing, (The IDC also shows the yield in their database as 10-15 kt)

Below are 2 papers about 1974. All of them are playing the usual games looking at teleseismic estimates and do their best not to accept what the scientists on the spot say. All subsequent cross references on information in scientific circles tend to be biased towards certain papers that are considered by "reputation" as "key" papers. This appears to be as true in this business as it is in medicine. Medicine at least has a periodic review process. With few new entrants to the nuke testing field, and decreasing numbers of tests, there appears to be no incentive to questioning or reviewing what becomes truth by accident or default.

http://nuketesting.enviroweb.org/hew/India/IndiaRealYields.html
http://nuketesting.enviroweb.org/hew/India/IndiaSmiling.html

Note the overbearing presence of Wallace and Douglas

One paper says "Iyengar conceded" that the yield was smaller. The other paper reveals that Douglas says that Iyengar said this. No direct quotes from Iyengar. Ha Ha.

What is my problem with all this? 20 years from now only Wallace and Douglas, who read squiggles in distant seismograms will be the only ones who are quoted. Nobody in the world is interested in listening to the claims of people who make on the spot measurements. Everyone argues about yield from imprecise teleseismic methods while performing pooja to the seismologists.

Still looking for more details . . .

Umrao · Post by **Umrao** » 03 Jul 2002 06:31

shiv>> The only way to uphold the findings of BARC is to drop one on top of Islamagood and ask mv ramana to publish findings.

shiv · Post by **shiv** » 03 Jul 2002 06:35

<a href="http://www.ca.sandia.gov/casite/gupta/r ... 1974">This paper</a> says the 1974 test was 4.9 on the Richter scale.

No mention of mb values.

I wonder whether the monitoring regime was geared up with all their monitoring staions and magic formulae in 1974? I think not.

shiv · Post by **shiv** » 03 Jul 2002 07:02

Originally posted by schawla:

Incidentally after the previous discussion I exchanged a few emails with Prof. Wallace and he related an interesting ancedote:

Apparently in the early 80's the US nuclear establishment was trying to deliberately "overestimate" the yields of Soviet tests in order to get funding to carry out more elaborate tests. Terry Wallace and his colleagues were able to "convince" Congress that Soviet yields were lower than what the US establishement were quoting.

Interesting, if true.

Wallace ascribes an agenda to other yield establishers. Apears that he too had an agenda in this "convincing" business.

If this email story is true - I would preserve the email as a direct admission of Wallace reading a pecuniary agenda in some yield reports from his peers in the US and waging a successful personal jihad against it.

It is now easy to allege that Wallace may have an agenda in much of his work on the Indian tests - on the basis of the argument that "absence of evidence of agenda is not evidence of absence of agenda" We now have a clear aganda story with regard to the Soviet tests.

Kuttan · Post by **Kuttan** » 03 Jul 2002 07:25

It is a part of my day job to sometimes have to read the life histories of people who walk on water ( at least according to themselves and those who write letters glorifying them. Reading one of those, a light bulb finally went on in my brain.

There was this guy who was being touted as "one of the best in the world in his field". Reading carefully, I saw this quote:

**** is one of top people in the Interior Point community

The key to declaring oneself to be a "the world's Top Authority in the field" is to define one's field appropriately. So I guess this Wallace dude is the Top Authority in the world (meaning US) in Foreign Nuclear Weapon Yield Underestimation (FNWYU) He's done it with soviet weapons (with some data that he claims came from the CAR. Now he's also done it with Indian nuclear devices. :p

shiv · Post by **shiv** » 03 Jul 2002 07:29

Originally posted by narayanan:
I guess this Wallace dude is the Top Authority in the world (meaning US) in Foreign Nuclear Weapon Yield Underestimation (FNWYU) He's done it with soviet weapons (with some data that he claims came from the CAR. Now he's also done it with Indian nuclear devices. :p

Narayanan - my own "light" came on when I realised the lack of muckraking in the "teleseismic yield guessing pseudoscientific game"

If one were to dig deeper I am certain that it would be dead easy to come up with contradictions and perhaps even outright lies that are passed on for all time as science.

Tim · Post by **Tim** » 03 Jul 2002 07:54

Shiv,

My copy of Perkovich's book is currently in a box somewhere so that I can ship it to Rhode Island. However, I know that he has several citations of different yield estimates on Pokhran I, and I believe one of the "revisionist" estimates is attributed to Iyengar, and again IIRC that estimate was 8-10 kilotons rather than the original 12-15. That's a significant differential (50+% lower).

It still sounds to me, again just from a text analysis, as though the Sikka estimate may have been based on an "assumption" rather than actual DATA. This is worth following up, and if I knew where I'd buried the @$#$@$# book I'd do it myself.

However, my memory may be faulty. IIRC, Wallace himself also posits lower yields for Pokhran I. Does anyone have any idea how he may have arrived at them? Or did he just use his 1998 assumptions and look backwards?

Hope that helps a little.

Tim

Umrao · Post by **Umrao** » 03 Jul 2002 08:24

Ok I pulled Perkovich out from where he was perched, I look at page 181 through 183.

<I> Questionable Yield.</I> Another potential controversy that never arose concerns the explosive yields of the test. AEC Chairman Sethna told journalists on the evening of the test that its yield was between 10 to 15 Kt. Since then practically all descriptions and commentatires on the test describe the yield at 12 Kt or between 12 to 15 Kt. In a authoritative paper Ramanna and Chidambaram prepared for IAEA meeting on peaceful nuclear explosions in January 1975, the Indian scientists "estimated" the yield at about 12 Kt.. Coincidentally or not the 12 Kt figure matches the yield of the American bomb dropped on Hiroshima....
Since 1974, Indian journalists have periodically reported that measurements of the yield were inflated. In june 1975 the Hindu cited sources claiming that the yield has been considerably less than predicted. A detailed august 30 1981 article in the SUnday Observer (Bombay) reported that "sources at the BARC maintain that independent measurements by some scientists put the yield at Pokhran to be as low as 2000 tons of TNT or 2 Kt. Despite these journalists claims of inside sourcess, their reports of a lower yield have never shaken the accepted truth of a 12 Kt yield.
(here comes the gem)
The uncertainity seems in part from the inexactness of sciesmic monitoring capabilities and interpretations. Technically, the most accurate way to determine the yield of a nuclear explosions is to conduct radio chemical analysis of the radio active debris at the test site.

In interviews with the author, two former chairmen of the Atomic Energy commision said that the radio chemical analysis showed a lower than advertised yield. "The yield was much lower than that has been stated" Sethna acknowledged in 1996.
PK Iyengar reported that "The yield was checked against the measurements we had taken. It checked outr very well between 8 - 12 Kt. When asked why the the results of the radio chemical analysis had never been published, Iyengar responded "what does it matter if it was 8 or 12 Kt?"

jaikiran · Post by **jaikiran** » 03 Jul 2002 08:26

Shiv ji,

Great Thread!

No...! i am not trying to contribute my pearls of wisdom in this very technical thread!(never was good at math & physics... so chose medicine!!

)

few general issues with regards to Phoren experts:-

1)TSP with less money,funding, academic institutions etc... still have better quality everything

b)we will not be too wrong.i think we will be as good as phoren experts as i think their guess is as good as ours.

admin!delete if my post is too non-technical.. i agree it is.. but i am making a general point i.e:- lack of self confidence in ourselves and our experts plus our DDM playing a role in perpetuating this mentality.

Jai.

Sunil · Post by **Sunil** » 03 Jul 2002 08:37

schawla

> wallace quotes the Indian estimates.

Where does he do this? He mentions the Indian estimates but never sources any of his claims of POK1 having an mb=4.9 .

What makes matters more hilarious is Fabian and Gupta listing Dr. P K Iyengar's statement to Douglas about POK1 being `exactly 8 Kt as predicted'. This is a statement that Dr. P K Iyengar has never corroborated. It seems awfully dumb to me that the entire `Low-Yield-of-POK2'jamboree rest on this one statement.

All this does not do much to improve my opinion of Douglas ofcourse.. not that anything really could but still, there is allways hope.

shiv · Post by **shiv** » 03 Jul 2002 09:02

Originally posted by Tim:

However, my memory may be faulty. IIRC, Wallace himself also posits lower yields for Pokhran I. Does anyone have any idea how he may have arrived at them? Or did he just use his 1998 assumptions and look backwards?

Hope that helps a little.

Tim, this helps in showing that given the right circumstances apples can be confused with oranges. When one speaks of accuracy or truth, one must judge if such confusion is desirable. Does it matter? That is perhaps more of an ethical question - but if projections and analyses are based on questionable data, it is going to matter somewhere down the line.

The GOLD STANDARD for estimating yields is close in measurements during and after the tests.

All teleseismic methods are notoriously innacurate.

How is it that all papers, references, verification regimes, think tank analyses, projections, speculation base all their data and arguments on the results of the notoriously innacurate teleseismic methods?

How is it that nobody stops to think that all these proliferating references and cosy analytical scenarios are all based on a questionable method of establishing yields? This whole business of detecting explosions and establishing yields from teleseismic data is fraught with error, and the whole structure appears to me to be resting on the claims of one or two people with big egos. This sounds not a little problematic to me.

Calvin · Post by **Calvin** » 03 Jul 2002 17:34

This thread is attracting a lot of whiners, who think like to bring in their irrelevant comments into essentially a thread on physics.

IMHO, and this was discussed in the BRM article, is that none of the foreign physicists have adequately accounted for the interference effects. Further, these interference effects can be shown to have a significant impact on the interpretation of the measured M_b values. Finally, a difference of 0.1 in M_b values, since we are dealing with exponential functions can result in ~25% difference in calculated yields.

If we are going to spend time discussing this topic again, then perhaps we should resolve to have an objective - which is to create an easily readable and understandable article for BRM that dissects the math, and addresses the objections of the dissenters.

Tim attempts to be objective by assigning equal credibility to the ignorant, the propagandists and the scientists. One does not really respect this methodology. As intelligent people we should have the ability to separate the chaff from the wheat and recognize that neither Perkovich, Sublette nor Vanaik are physicists that would recognize the difference between natural log and log to the base 10.

In reality, it is possible to cross reference the M_b and the ground signature, and the radiation isotope data.

Kuttan · Post by **Kuttan** » 03 Jul 2002 18:18

Calvin: The point here is that some of what is passed off for "physics" in the "prestigious journals" is actually biased, dishonest, "lifafa" science. This escapes censure because the truth can never be shown - enabling all these, yes, political and ego-based arguments based on calculations which are uncertain by orders of magnitude.

The thread is properly titled. The debate on these sorts of supposedly intricate scientific/mathematical issues should indeed be subjected to the light of common sense.

What I've learned so far here:

1. The only way to really measure the "yield" reliably is to combine data on

a) the concentrations of various products of the reaction (why? is this because it gives a good idea of the temperatures and the intensity of the radiation?)

b) the cavern size (though this clearly should depend on the composition, hardness etc. of the earth in which the device was buried).

2. Given the above, the seismic propagation method is indeed subject to order-of-magnitude error.

3. Data from (1) are not going to be released - and if they are released, I would ask "why are these being released? and hence, how do I know they have not been fudged?"

4. Given (2), the "conclusions" expressed by these so-called "Authorities" are worthless.

5. Given that they pompously publish worthless conclusions based on nonsense data and formulae, these experts are not scientists, any more than snake-oil salesmen are chemists.

6. And hence, Calvin, the whole business is not about "physics" but about egos and the absence of ethics in this field.

7. And so the thread is also not merely about "physics" but about the gross abuse thereof by charlatans out to cheat the general public - "lifafa" physicists. And the use of common sense to show this.

Q.E.D.

As such, it is vital that a free discussion on this issue continue on BRF (not on BRM) and show the maximum number of people possible, how so-called "international experts" are used, or use their own credentials, to advance lies. And how the DDM are taken in so easily by this - because they judge credibility, not by the common sense of what is stated, but by "who" is stating it.

member_4218 · Post by **member_4218** » 03 Jul 2002 20:47

Originally posted by sunil sainis:
schawla

> wallace quotes the Indian estimates.

Where does he do this? He mentions the Indian estimates but never sources any of his claims of POK1 having an mb=4.9 .

You maybe right. In the Seismic Research Letters, vol 69, p 386-393, the claim that mb=4.9 is not cited, atleast in the web reprint.

Kuttan · Post by **Kuttan** » 04 Jul 2002 00:08

Question to the gurus here, just to show my level of expertise in these matters: What is the shape of a "typical" nuclear blast "cavern"?

My understanding is that the radiation and the blast wave propagate in all directions (spherical) regardless of the shape of the bomb (anything that "shadows" the radiation nearby is instantly vaporized). The radiant heat melts and vaporizes the surrounding rock and soil - then the shock wave comes through and raises the pressure to some gazillion atmospheres, and compresses the rock all round. The rock is shattered, melted, vaporized. Eventually the radiation and shock wave are attenuated, so stuff is merely melted, not vaporized. This stuff must ooze down, so the top portions of the cavern collapse around the edges, and the molten rock settles at the bottom. What happens to the vapor? Does it leak out through fissures above? Does it get absorbed into the surrounding rock? Do the cavern walls harden into something with a lot higher density?

IOW, how is conservation of mass satisfied for the mass that used to be inside the "cavern" region?

If the top bulges, as it did at Pokhran-2 (or was there just a depression? ), isn't that a good indication that the yield was nearly on the point of exceeding design limits, and they came close to having an environmental disaster (i.e., "fallout" getting tossed into the air)?