Over a year ago there was correspondence in this blog over Professor Sienicki’s paper on the subject of the temperature on the Antarctic Ice Barrier as Scott and his companions made their fated return and the role of the temperature in relation to Scott and his companion’s deaths. The paper put forward the idea that Scott and Bowers, having decided on suicide, fraudulently doctored the Barrier temperature recordings on the last part of their return, so as to suggest unusually cold weather. The authors suggested that an important motive for doing this was that Scott was fearful of powerful enemies in Britain who had slandered both himself and Lieutenant Royds over the latter’s meteorological results from the ‘Discovery’ expedition.
In May 2014 Kristoffer Nelson-Kilger defended the team’s conclusions and mentioned a book “Captain Scott’s Fatal Antarctic Expedition- Slanted Truths–Centennial Account’ which was to be published.
I refuted the suggestion as did Karen May, a polar researcher who has gone minutely into Scott’s last expedition; her detailed and strong reply (7 pages), was posted on 02/05/14.
I have been looking out for the definitive book but as yet have not seen it, or been sent a copy. I am interested in knowing, when this book, which in my opinion goes against the evidence and which makes suggestions contrary to the characters of Scott and Bowers (not to mention Edward Wilson who was with the mion the final part of the expedition), will appear.
Isobel Williams – Author, Cruise Speaker and Lecturer 
Hello Isobel,
Before I answer your question, I have one clarification I have to make. Because of your presumption that Sienicki was currently a professor, I mistakenly thought so as well. In actuality, he is retired from being a professor, and is currently self-employed quite successfully.
To answer your question, Sienicki’s book is nearing completion. Negotiations are in progress with a highly respected independent publisher who is interested. The only other information I can say about the publisher’s identity is that no British publisher he has contacted showed enough interest.
I also wish to note three more things:
1) I am not an co-author, but rather a research assistant;
2) that the previous book title was a provisional title, a new title has since been implemented, and I will not reveal the new title here;
3) your own propensity to errors, like thinking Sienicki had made a neural network across the Barrier, thinking that I am a co-author to Sienicki’s arXiv papers when I am not credited + they were created before I began assisting him, and several egregious errors I noted from a brief look at your book “Captain Scott’s Invaluable Assistant, Edgar Evans,” would preclude you from review of his book.
Hey Kristoffer
Are you able to describe the method by which you and the retired professor have validated the accuracy of the computer model which is the basis of your claims about Scott’s suicide pact?
Given the startling claims you are making about Scott’s motives, I’d suggest a convincing peer review of your model would be in order. )It would of course be most unfortunate if your claims turn out to be based upon a flawed model).
Perhaps you could provide a list of peer reviewers, their credentials and a description of their review process. Just a thought …
I look forward to your response
Bill Alp
I think this is important and await the reply with interest
Isobel Williams
Hello,
Sienicki’s response is below. I will also add that his thesis that Captain Scott and his comrades committed suicide does not solely depend on the falsification of temperatures by Captain Scott and Lt Bowers.
——————————————————————————————-
Hello Bill, all scientifically pertinent descriptions of methods used in my studies are sufficiently described in my paper, available at http://ocean.am.gdynia.pl/p_k_p/pkp_21/Sienicki-pkp21.pdf. The provided description will allow you (or anyone else) to develop your artificial neural network engine and make the required simulations. (If you provide me with your snail postal address, a CD with my software will be forwarded to you.) As for editorial matters, I advise you to follow the customary procedure and enquire about matters of your interest by contacting the Editor of this journal. All information is available at http://ocean.am.gdynia.pl/p_k_p/pkp_sg.htm The e-mail address for the Editor, Dr Styszynska, is astys19@wp.pl
I have commented with Sienicki’s reply, and am awaiting the comment’s approval.
Just in case my previous attempt failed, I will post his reply again:
————————————————————————–
Hello,
Sienicki’s response is below. I will also add that his thesis that Captain Scott and his comrades committed suicide does not solely depend on the falsification of temperatures by Captain Scott and Lt Bowers.
Hello Bill, all scientifically pertinent descriptions of methods used in my studies are sufficiently described in my paper, available at http://ocean.am.gdynia.pl/p_k_p/pkp_21/Sienicki-pkp21.pdf. The provided description will allow you (or anyone else) to develop your artificial neural network engine and make the required simulations. (If you provide me with your snail postal address, a CD with my software will be forwarded to you.) As for editorial matters, I advise you to follow the customary procedure and enquire about matters of your interest by contacting the Editor of this journal. All information is available at http://ocean.am.gdynia.pl/p_k_p/pkp_sg.htm The e-mail address for the Editor, Dr Styszynska, is astys19@wp.pl
Hi Kristoffer
Thank you for responding to my 20 October post on Isobel Williams’ blog. I appreciate your promptness.
Unfortunately the response provides information which doesn’t address what I thought I’d asked for. I apologise for my request not being clear enough.
My concern is about formal testing of Professor Sienicki’s software (the “artificial neural network engine” to be more precise). We all know that newly developed software often has serious defects or “bugs”. There is a large body of information technology professionals who specialise in software testing, and they have well established processes for testing whether software actually performs in accordance with what has been specified in “functional requirements” documentation. This goes under a generic heading of “best practice” and these people talk about “testing maturity model” for assessing organisations which specialise in testing of software.
The website link you provided took me to a pdf document which could be broadly seen as “functional requirements” for the neural network, from which the software model would be built and tested. But it did not address independent testing of the model.
One of the key principles of software testing is that of independence – the tester must be totally independent of the person who built the software. Bitter experience shows it is all too common for a software developer to not detect their own blind spots. This is of course akin to peer review of written documents.
As well as understanding what has been done to objectively validate the model’s functional behaviour, it would be helpful to understand if independent validation (to the same level of confidence) has been completed on the meteorological data that has been fed into the model. Are you able to provide information about any independent assessment of the meteorological data that has been undertaken? It would be great if that information could include validation of the full range and sequence of meteorological data used. Would an objective third party be able to access the data for themselves?
I hope the sentences above clarify my request. Please let me know if any of this is not clear.
I look forward to your response
Regards
Bill
Here is Sienicki’s response:
Hello Bill, my paper stands by itself as a scientific publication, and it contains all information pertinent to a scholarly research work, provided that it is read carefully. In this work, a scientific method was systematically used. The method was based on computer experiments of simulations of an artificial neural network (ANN), more specifically simulations of the back-propagation neural network. It is the basic requirement of scientific work to test methods and tools used in particular research. The method of developing the optimum architecture of ANN and method of ANN performance testing (by de-selecting data input chain) is described on pages 52-56 of my publication. The ANN used in my work was validated from a full range of architectures (# input neurons and # input layers) and input temperatures. A description of data sources is also presented in the paper, and the data (minimum daily temperatures) are freely available to anyone interested. See the Acknowledgements and references provided in the paper. As a matter of fact, two years after my original publication, the used ANN was one more time independently tested by Dr L. Kulak and M. Sc. student Adrian Piasecki at the Faculty of Theoretical Physics and Quantum Informatics, Technical University of Gdansk. These tests fully confirmed the original work.
Hi Kristoffer and Professor Sienicki
Thank you for identifying your independent testers: Dr. L Kulak and Adrian Piasecki.
Are you able to share their Test Exit Reports via this blog? Hyperlinks will be fine as I’d expect the full text to be too large to post on a blog.
Thanks again.
Bill
Here is Sienicki’s response:
Dear Bill, since the names were identified as you mentioned in your post, I suggest that you forward your questions to the named people. If you want to contact me, just drop an e-mail into my box at kris_sienicki@yahoo.com Best regards.
Hi Kristoffer and Professor Sienicki
I do not think it is appropriate for me to contact your independent software testers directly. As it is your ANN software model, I have asking that you share the extent of validation undertaken. Your tester’s “Test Exit Report” or “Test Summary Report” should serve that purpose admirably. I believe such information should be shared with all participants and observers on the blog.
Once I have seen the full extent of information that can be made available, I will start by checking out the way that modelling of height variation between observation points has been validated. I can see from your paper (linked in prior email) that you have built and tested your model using data from automated weather stations that are all about 4.5 metres above surface level; I’d call that a 2-dimensional model as there is no significant variability of height. You have then applied the model to historic data from the Terra Nova expedition which has considerable height variation. The Cape Evans data was recorded on top of Wind Vane Hill which is 66 feet above the surrounding surface, and when you add the height of weather station equipment (your paper says 5 feet) then that comes to over 21 metres above the surrounding surface. The observations take by the Polar Party, on the other hand, would have been made at about waist level at (say) 1 metre above the surface. I’d therefore characterise the data from the Terra Nova expedition as being 3-dimensional.
So in reviewing the test reports I’ll start by looking at how Dr. Kulak and Adrian Piasecki evaluate and validate the usage of a 2-dimensional ANN model against 3-dimensional data.
I suspect this may become important when temperature inversion layer conditions are addressed. Say, for example, the Polar Party was within a temperature inversion layer whilst the Cape Evans observations were being made on a 20 metre hill which stuck out above the temperature inversion layer. The validity of your ANN model when processing that type of situation seems important to me.
I look forward to gaining a more complete insight into testing of your ANN model in such “boundary cases”, as they are called by professional testers.
Best regards
Bill
Here is Sienicki’s reply:
Here is Sienicki’s reply:
Dear Bill, thank you for your persistence. However, you are heading in the wrong direction. I am sorry to say this, but you do not understand the elementary issues directly related to Artificial Neural Network (ANN) simulations. Would you first read and get at least a basic understanding of ANNs, and then follow with educated comments? From your writing like “It would of course be most unfortunate if your claims turn out to be based upon a flawed model,” I read your negative feelings (only feelings without sound argument) toward the whole issue. And please do not think that you are very smart, while the others are unable to get the things right. First read, then study, and think, finally discover, and at the end communicate to fellow people, researchers, etc. Having done so, you will be able to communicate with others at a civilized level, meaning exchanging educated knowledge and at an educated level. That is how at least scientific progress works and advances knowledge. Rambling over the Internet and posting pop-corn thoughts would do nothing.
Now, back to the ground and your last post. Let me tell you that your comment:
“So in reviewing the test reports I’ll start by looking at how Dr. Kulak and Adrian Piasecki evaluate and validate the usage of a 2-dimensional ANN model against 3-dimensional data.”
associated with your line:
“I’d therefore characterise the data from the Terra Nova expedition as being 3-dimensional.”
is entirely incorrect (uneducated), since temperature is a physical scalar and not a vector.
Let me for your convenience cite here Wikipedia: “In physics, a scalar is a one-dimensional physical quantity”. Thus, you are confusing elementary issues.
Moreover, why the back-propagation ANN in my work is according to you “2-dimensional” does not find an answer in my wildest imagination. An ANN is a dimensionless entity (computer code originally written on piece of paper). I am sorry Bill, but I am losing you on this point too.
The back-propagation ANN used in my work is not only dimensionless, but what is more importantly model-less, meaning no specific theoretical framework (theoretical model, theoretical algorithm) was assumed about the simulated quantity. Quite possibly, this ANN can be effectively used in your washing machine and/or other appliances (do not be surprised by that).
What counts is the ANN training. You feed the ANN with data (whatever data), and due to the explicate algorithm (back-propagation), the ANN is “learning” to guess the output. It means that the ANN is not looking at a specific place, situation, etc. it is looking at digits and it is trying to learn how properly respond to input digit. The washing machine is not washing yours or my socks, it is washing clothes using a certain ANN. In this sense, an ANN is a model-less algorithm. Of course, one using an ANN to control a washing machine can build one in which another ANN will be trained to recognize whose socks are being washed.
Now, that is exactly how my ANN is working in my paper. It knows nothing about Scott or anything related to him. The ANN algorithm gets the numbers, crunches them, and learns to crunch them better and better. Usually, it takes 105 iterations (each neuron weights) to get it right. Getting it right means that the difference between expected and actual output reaches a minimum value, called prediction error (in terms of statistical error).
In terms of Captain Scott, I was able to figure an ANN that performed in a certain way using minimum temperature data from Antarctica. However, I encourage you to make a shot. One day, it will be possible to find an engine to predict say the temperature in Antarctica using the actual temperature in London. You could be the next L. Fay Richardson.
Now, I hope you are able to understand that all particulates related to temperature data (in case under consideration) are loaded into a given ANN with given number of layers and given weights of neuron connections. In the case of an ANN, an input scalar is non-linearly mapped (in mathematical sense) into output scalar. There is a large body of book and research papers related to ANNs, and I encourage you to research them.
Here is a Synopsis of Chapter 7 of my book:
“The weather patterns in the Antarctica are neither completely regular nor completely irregular. These weather patterns are rather a self-organized criticality. I have not arrived at these conclusions by analyzing logical structures of particular weather descriptions from original journals. In this chapter, I have arrived at the conclusions presented because I have rigorously and scientifically analyzed modern and historical weather data. All modern meteorological data, as well as the historical data of Cherry-Garrard, has shown high correlations between temperatures at different locations at the Ross Ice Shelf, and precision retrodiction of modern and historical data all point to the oddity of Captain Scott’s temperature recordings from late February 27th through to March 27th, 1912. On the basis of the above-mentioned evidence, I conclude that the actual minimum near surface temperature data was altered by Lt Bowers and Captain Scott to inflate and dramatize the weather conditions.
Therefore, our understanding about the decisive causes of the Main Polar Party’s deaths must be entirely reassessed. As a result of the findings included in this chapter and the two chapters to come, I conclude that their deaths (Scott, Wilson and Bowers) were a matter of premeditated choice rather than severe weather conditions. The choice was made long before there was an actual end of food and fuel, and long before the end of physical strength needed to reach the imaginary and delusive salvation at One Ton Depôt.”
Bill, the decision to not return alive to Hut Point was made by Captain Scott’s party on or around Feb. 2nd, 1912. In my book you will find how I was able to deduce this date. I suppose that my upcoming (spring 2016) book containing the two volumes, each about 400 standard book pages with 149 plots of different data, will provide you with much to munch on. At this moment, I can assure you that my initial thesis (made in 2010) holds up, and is not solely based on Captain Scott’s temperature falsifications.
Throughout this thread I have been focussed on one thing only – gaining an evidence-base understanding of what testing has been carried out on your ANN software and the conclusions reached about the Terra Nova expedition.
Here is a summary of what I believe has been disclosed about your testing to-date. I have numbered each items so we can ensure all aspects get resolved in a systematic way:
1. You have built an ANN software model based upon sound principles of theoretical physics and neural network modelling, as set out in your paper at http://ocean.am.gdynia.pl/p_k_p/pkp_21/Sienicki-pkp21.pdf
2. You have populated (or “trained”) it with a large volume of data from automated weather stations at various locations on the Ross Ice Shelf. Terra Nova data was not been included in this phase of development.
3. You have characterised the ANN software as being “model-less”. In particular, it does not take vertical height of observation points into account. Neither is it aware of which individual observations may have been recorded under temperature inversion layer conditions or other meteorological phenomena which are known to influence near-surface temperatures. The ANN software is simply concerned with time-series of temperature observations at fixed locations.
4. You are confident that your (well-trained) software can be validly used to analyse temperatures recorded by the Terra Nova expedition, in order to search for inconsistencies in temperature data recorded by expedition members.
5. Dr. L Kulak and Adrian Piasecki have independently tested and confirmed items 1 thru 4 above. I have requested that you make a copy of their test summary documentation available for review via this blog. This interest in disclosure of independent testing information comes from my grounding in “best practice” in the information technology industry.
6. You have not commented on the consequences of not taking vertical height of observation points into account. As we are talking about “near-surface wind events”, it would be helpful to understand your view of the flow-on effects of the McMurdo AWS (approx.. 4 metres above surface) being taken as the proxy location for Wind Vane Hill (21 metres above surface). Your paper describes a variation to your model which introduces an uncertainty range of +/- 2 degrees Celsius to compensate for the (x,y) distance between those 2 locations. I presume your independent testers concurred with that approach and the 2 degree value? Did they perceive any need for a similar uncertainty range to account for differences in height above the surrounding surface?
7. You have not commented about temperature inversion conditions where the Polar Party would be experiencing reduced temperatures within the inversion layer, whilst the Wind Vane Hill observations would be taken at a vertical height above the inversion layer.
Please let me know which of these numbered items you agree with completely, also any changes required and omissions I may have made.
Thankyou
Bill
Here is Sienicki’s response:
Hello Bill, since your account was not clear to me, I did not comment about your observations concerning “temperature inversion layer” – point 7 in your last post and as mentioned in before. Could you say in a more descriptive way what you have in mind by saying “temperature inversion layer”?
Hello Kristoffer and Professor Sienicki
You asked about the meaning of temperature inversion layer. Here is a link: https://en.wikipedia.org/wiki/Inversion_%28meteorology%29 for an overview of temperature inversion, but I guess you know this stuff already!
We also know that the Terra Nova scientific staff undertook work in winter 1911 to study the temperature difference between the top of Wind Vane Hill and the Cape Evans hut. The temperature at hut level was found to often be 4 or 5 degrees colder than the hill-top; usually in calm conditions. These scientists had at least some understanding of temperature inversion layers and referred to the phenomenon as “inverted temperature”.
My point is that by knowing only the temperature at the top of wind vane hill one cannot predict the temperature 21 meters away, let alone 200 miles away. In order to use wind vane hill observations to predict temperatures at another location one needs an indication of the presence or absence of inversion layer at the remote location. Maybe wind conditions could provide that indicator, but I’m digressing into meteorology and should confine myself to software testing!
We now enters the realm of testing and validating design assumptions. Many fine software projects have foundered because they did not test and validate their design assumptions. It seems you have an undocumented design assumption underlying your ANN network, something like “height and wind conditions do not need to be taken into account in this model”. Like all design assumptions, this one will limit the range of situations where the model can be trusted to provide valid results. Your paper http://ocean.am.gdynia.pl/p_k_p/pkp_21/Sienicki-pkp21.pdf describes how you validated all your design assumptions against datasets from automated weather stations that are all at the same height above the surface. But the paper appears to be silent on validation of the model using datasets from Wind Vane Hill, or any other location at a different height above the surface. Should I be looking elsewhere for this evidence?
It would help to round-out our understanding of testing undertaken (both by yourself and by the independent testers) if you could list all your design assumptions and how they have been validated. Maybe this should be added as an eighth point to the list in my previous post.
It could also be instructive to hear from any meteorologists following this blog. Expert opinion on the applicability and limitations of an Antarctic weather model which does not pay regard to height or wind would be appropriate at this stage.
Cheers
Bill
Here is Sienicki’s response:
Hello Bill,
Thank you for making reference to Wikipedia, which hardly may serve as a reliable source of information, and it makes no reference to meteorology and/or inversion in Antarctica, more particularly at Ross Island and the Barrier. Would you please let us know your sources and references (research papers) in relation to inversion of temperature in the mentioned regions?
While waiting for the above, let me just point out that you contradict yourself. First, you say that it was established that the temperature at Cape Evans hut is “often 4 or 5 degrees colder” [what degrees] than at Wind Vane Hill. Then you say that, and what appears to be your main point, “My point is that by knowing only the temperature at the top of wind vane hill one cannot predict the temperature 21 meters away, let alone 200 miles away.”
Whoops! The prediction equation is indeed simple and follows from your first cited comment
Cape Evans Temperature = Wind Vane Hill Temperature – (4 or 5) [please specify degrees]
Thus on the contrary to your “unpredictability at 21 m range” comment, the prediction in a range of 21 meters works in a simple way: reduce the hill-top temperature by 4 or 5 and you get the predicted temperature at Cape Evans.
Let us know what you think. Cheers
Throughout this thread I have been focussed on one thing only – gaining an evidence-base understanding of what testing / validation has been carried out on your ANN software and the conclusions reached about the Terra Nova expedition.
Here is a summary of what I believe has been disclosed about your testing to-date. My update today (8 November) includes a new item 8 about validation of software design assumptions and re-casting of item 6 to better align with items 7 and 8. I have retained the numbering of each items so we can ensure all aspects get resolved in a systematic way:
1. You have built an ANN software model based upon sound principles of theoretical physics and neural network modelling, as set out in your paper at http://ocean.am.gdynia.pl/p_k_p/pkp_21/Sienicki-pkp21.pdf
2. You have populated (or “trained”) it with a large volume of data from automated weather stations at various locations on the Ross Ice Shelf. Terra Nova data was not been included in this phase of development.
3. You have characterised the ANN software as being “model-less”. In particular, it does not take vertical height of observation points or wind conditions into account. Neither is it aware of which individual observations may have been recorded under temperature inversion layer conditions or other meteorological phenomena which are known to influence near-surface temperatures. The ANN software is simply concerned with time-series of temperature observations at fixed locations.
4. You are confident that your (well-trained) software can be validly used to analyse temperatures recorded by the Terra Nova expedition, in order to search for inconsistencies in temperature data recorded by expedition members.
5. Dr. L Kulak and Adrian Piasecki have independently tested and confirmed items 1 thru 4 above. I have requested that you make a copy of their test summary documentation available for review via this blog. This interest in disclosure of independent testing information comes from my grounding in “best practice” in the information technology industry.
6. You have used McMurdo AWS as a proxy location for the observations made at the top of Wind Vane Hill. Given the distance between them and the difference in height and wind conditions between these two places, how has the correctness of ANN results based upon Wind Vane Hill observations been tested and validated? Does your independent testers’ report concur?
7. You have not commented about temperature inversion conditions where the Polar Party would be experiencing reduced temperatures within a temperature inversion layer, whilst the Wind Vane Hill observations would be taken at a vertical height above the inversion layer.
8. In a recent post, I asked you about a possible software design assumption concerning height and wind independence of your ANN model. Please provide a list of your software design assumptions and for each assumption produce evidence about how it has been tested and validated by yourself and your independent testers.
Please let me know which of these numbered items you agree with completely, also any changes required and omissions I may have made.
Thankyou
Bill
Here is Sienicki’s response:
Hello Bill, without your response to my last post (Nov. 7th), and in particular without referencing the sources of temperature inversion research on Ross Island and the Barrier, it is impossible to proceed and to respond in an educated way to your comments.
Let me repeat my question from the last post:
“Would you please let us know your sources and references (research papers) in relation to inversion of temperature in the mentioned regions?”
Your reference of temperature inversion from Wikipedia does not make a reference to Antarctica, Ross Island, or the Barrier. It makes reference to San Francisco.
Let us proceed with discussion. However, let us base it on sound and definitive scholarly works. Your temperature inversion must be referenced. Therefore, your observation
“You have not commented about temperature inversion conditions where the Polar Party would be experiencing reduced temperatures…”
cannot be commented on in an educated way.
Bill, if you built the argument in the following way: because of that … based on that … and observing this … I arrive at a conclusion that …
Transfer your argument into numbers, supported by historical and modern measurements and references, and share it with us. Then I will respond to every minute number and detail. That is how new discoveries are made, and quite possibly you are about to do, but provide us with numbers to crunch on.
Hello
I was not expecting you would offer to “crunch” numbers or revise your model. Upon reflection I believe this thread should be confined to your 2011 paper http://ocean.am.gdynia.pl/p_k_p/pkp_21/Sienicki-pkp21.pdf and the evidence which directly supports it. Any refinement or rework of the paper or the ANN model would be best addressed via a new thread.
I note the file creation date of your paper is 23 December 2011. All I am seeking is for you to share the evidence of testing / validation that supports your paper. This would be readily achievable by sharing your own test reports, plus the reports of your independent testers, provided they are dated no later than 23 December 2011.
Cheers
Bill
Here is Sienicki’s response:
Hello Bill, let me note that my effort to move along well-defined lines using referenced comments about your thesis of “temperature inversion” concerns is being obstructed. If you give the numbers, they alone would enable me to respond how these numbers were implemented in calculations and analysis. What heights are you referring to? Give us your height estimations. It will be helpful for me if I know what figures were included in your calculations. Therefore, I urge you to give us your figures. Otherwise, it is pointless rambling, as all issues were already addressed before.
Bill,
Sienicki has played along with your deliberate refusal to contact him for the contact info of the testers long enough. I’m stepping in with his approval to finish this.
Unlike you, Sienicki and I are throughly familiar with temperature inversions on Ross Island. This subject was covered by expedition meteorologist George C. Simpson in his work Meteorology Vol. I Discussion. I now quote from page 41 of it:
“It will be noticed that all the ascents made in November and December, summer months (Nos. 5 to 10), show temperature decreasing with height and the mean gradient for the first 2 1/2 kilometers is 6-8°C. per 1,000 metres. It is of great interest to notice that this is practically the same gradient as that found in summer months in Europe and America although the actual temperatures in the Antarctic summer are similar to those of winter in those regions.
The conditions are entirely different during the winter. On account of darkness upper air observations could not be made before August. In this month, however, four successful balloon ascents were made and the instruments recovered.”
Simpson’s results from these measurements are depicted on page 42. It is quite clear that in the Antarctic summer, temperature inversions do not happen on Ross Island. Even in late winter, the strongest inversion recorded was 5° C/9° F over several hundred meters, not your ridiculous 21 meters. This by itself dismisses Captain Scott’s ridiculous statement in his diary entry for May 3, 1911, which you quote, and raises the question of if he fabricated it for the sake of appearing scientific. After all, such an extraordinary event was noted by no one, including Simpson. At this time, Captain Scott was shifting the definition of his expedition to include more science, in response to Captain Amundsen’s presence. Sienicki will cover this process in a part of Chapter 12 titled Resurrection of Captain Scott’s Deus ex Machina.
Simpson Vol. I page 43 demolishes your “explanation” by showing that temperature inversions do not happen in summer, period:
“In the Antarctic summer the snow surface is warmed relative to the overlying air by the almost continuous solar radiation and convexion currents produce a normal temperature gradient. In the winter there is little or no sunshine, while rapid radiation takes place from the snow surface which cools the air in its immediate neighborhood. A cold layer of air is thus formed, which, under favorable circumstances, may be many degrees colder than the air a few hundred metres above.
These conditions obviously can only obtain during the absence of wind. If there is a wind it produces vertical mixing of the air without the aid of convexion currents. The effect of a wind is different in summer and in winter. In the summer there is already a large temperature gradient which is little affected by the wind. In the winter the wind removes the cold surface layer and produces a normal temperature gradient. This process will be considered more fully later ; it is mentioned here to point out that the temperature gradient shown by the curves on figure 13 during the summer probably exists during all kinds of weather, while that shown for the winter exists only during calm weather.”
It is crystal clear from your refusal to contact Sienicki for the tester’s results that you are looking for an excuse to disparage Sienicki’s work. No one reading you believes you when you say “Throughout this thread I have been focussed on one thing only – gaining an evidence-base understanding of what testing / validation has been carried out on your ANN software and the conclusions reached about the Terra Nova expedition.” Least of all yourself, since you refuse to do any work yourself and simply make demands. In addition, your temperature inversion “explanation” does not hold water.
Hello Professor Sienicki and Kristoffer
Counting back through this thread, I see 8 requests for you to provide documentary evidence that describes the testing undertaken to validate your ANN software (testing both by yourself and also by independent testers). You’ve provided plenty of words (some quite unkind) but no documents or links to documents.
I had expected that in preparing a scientific paper for publication any credible scientist would carefully record comprehensive evidence of testing and validation, and make it freely available for review.
Clearly my expectations were too high!
In any event, the conclusions of your paper are controversial and the burden of proof is on you, not me. Whilst you have both been keen to parade your superior knowledge, that is no substitute for documentary evidence about the validity of your (2011?) software testing.
In the absence of well documented evidence, readers of this blog will form their own judgement, I’m sure.
Cheers
Bill
Here is Sienicki’s response:
Helol Bill, the normal and expected procedure in scientific publications (and the scientific community) and progress is that an already published paper is accepted as correct (which translates to innocent until proven guilty), until a following publication on the same issue makes a correction or verification of the past publication. For this reason, the ball is in your corner. And we look forward to reading your scholarly publication.
One small addition from Sienicki:
But even this small difference (if occurred) was included (present) in my ANN through nonlinear neuron threshold connections in the presented simulations.
Also, Sienicki wishes to add that as per Line 7 in Simpson’s Vol. I, page 42, Figure 13, the 21 meter difference Bill cites would amount to a temperature difference of only 0.089° C.
Hello Professor Sienicki and Kristoffer
How is the book coming along?
It has been over 6 months since your previous posting on this forum. I am wondering whether you have reconsidered your stance on validation of your ANN weather simulation model software, in particular the assumption that the McMurdo AWS automated weather station is a valid proxy for the location where Simpson made his measurements at the top of Wind Vane Hill in 1911/12.
Personally, I was disappointed that a professional meteorologist did not join the blog to provide an expert opinion on your assertions. That would have been the natural complement to my well-founded concerns in the area of software testing. Never mind!
I guess the crucial thing is whether you have the confidence to publish your opinions without providing evidence of independent verification of your software model.
Cheers
Bill
The book Content is below. The signed hardcover copy of the book ( £67.98 + shipping) will be available from the author at krissienicki@poczta.onet.pl from September 24th, 2016.
Captain Scott: Icy Deceits and Untold Realities
by Krzysztof Sienicki
[Book Info]
[808 pages, 8vo (18×28)cm., 164 figures, 1417 references, 39 tables and 2 schemes]
Contents
List of Figures, Maps and Art Works xiv
List of Tables and Schemes xx
Author’s Note xxii
Acknowledgements xxiv
Prolegomenon xxv
Volume I
1. General Introduction to the Earth’s Air Circulation 3
1.1. Early Development of Atmospheric Circulation Knowledge 4
1.2. Importance of Polar Meteorology 20
1.3. Meteorology of the Ross Ice Shelf and McMurdo Sound 33
1.4. Self-organized Criticality Wind Regime over Antarctica 37
1.5. Meteorological Games – False Charges Against Lt Charles W. R. Royds 56
1.6. Synopsis 65
2. Analysis of the Weather Account from the Terra Nova Expedition 67
2.1. Captain Scott’s Journals 67
2.2. Leonard Huxley’s Adjustment of the First Edition of Captain Scott’s Journals 70
2.3. Expedition member’s accounts and descriptions 75
2.4. Synopsis 104
3. Dr George C. Simpson’s Weather and Climate Tantamount 105
3.1. Weather vs. Climate 107
3.1.1. Daily and Annual Variation of Temperature 114
3.1.2. Ubiquitous Friction 119
3.1.3. Day and Night Sledging 126
3.2. How Cold Can it Get on the Barrier? 128
3.3. MCMXII 137
3.4. The Never Ending Gale or Blizzard 142
3.4. Synopsis 146
4. Dr Solomon’s Fabrication of Meteorological Data, Fallacious Analysis, and Temperature Mania 148
4.1. Proceedings of the National Academy of Sciences 150
4.1.1. Temperature – The Cold March of 1912 153
4.1.2. Wind Data Dragging 166
4.2. The Coldest March – Something out of Nothing 167
4.2.1. Dr Solomon – Weather and Climate Tantamount 170
4.2.2. Dr Solomon’s Fabrication of Meteorological Data 174
4.2.3. Dr Solomon Nullifies Captain Scott’s Responsibility 179
4.2.4. Dr Solomon’s Lack of Scientific Methods of Analysis 193
4.3. Dr Solomon’s Hoax Epiphany – Something out of Nothing 194
4.4. Biased Perception of Captain Scott as a Scientist 197
4.5. Temperature Mania 204
4.6. Dr Solomon at the Royal Society of Chemistry – Note Added in Proof 207
4.7. Synopsis 211
5. Historical Scrutiny of the Meteorological Record of the Terra Nova Expedition 213
5.1. Huntford’s The Last Place on Earth 216
5.2. Sir Ranulph Fiennes’ Faithful Enthusiasm 219
5.3. Barczewski, Jones, Crane, and Murray 229
5.4. Synopsis 239
6. Meteorological Data and Weather Forecasting 241
6.1. Sources of Meteorological Data 242
6.1.1. The Ross Island Historical Weather Stations 242
6.1.2. Sledging Parties’ Weather Records 245
6.1.3. The Ross Island Modern Weather Stations 246
6.1.4. The Ross Ice Shelf Automated Weather Stations 247
6.2. Time 248
6.3. Historical Meteorological Tools and Measurement Methods 249
6.3.1. Temperature 249
6.3.2. Wind Velocity and Direction 254
6.4. Modern Meteorological Tools and Measurement Methods 254
6.5. Historical and Modern Data Acquisition 255
6.6. Weather Forecasting 257
6.6.1. God and Arrogant Humans 259
6.6.2. Nansen’s Connection 263
6.6.3. Artificial Neural Network 270
6.6.4. The Back-propagation ANN Algorithm 271
6.7. Synopsis 273
7. February 27th through March 27th, 1912 – Extreme Cold Snap? 274
7.1. Extreme Cold Snap Hypotheses 276
7.2. Orography-driven Weather at the Ross Ice Shelf 278
7.3. Artificial Neural Network Development and Testing 282
7.4. The Extreme Cold Snap 290
7.4.1. Captain Scott Temperature Record Retrodiction 290
7.4.2. Inaccuracy of Retrodiction Method 292
7.4.3. Location Differences 295
7.4.3.1. Schwerdtfeger vs. Elaine Temperature Gradient 295
7.4.3.2. McMurdo vs. Cape Evans Temperature Gradient 298
7.4.4. Thermometers Malfunction 301
7.4.5. Global Warming 301
7.4.6. El Niño Teleconnection 302
7.5. Captain Scott’s Temperature Data Fabrication 303
7.6. Particulars of Temperature Data Differences 304
7.7. Synopsis 307
8. March 21st through 29th, 1912 – Never-Ending Gale? 308
8.1. Nature of Near Surface Winds in Antarctica 309
8.2. The Never Ending Gale 311
8.3. Synopsis 320
8.4. Appendix to Chapter 8 – Scaling Parameters of Wind in Antarctica 320
Volume II
9. Food, Fuel and Depôts – An Antarctic Menu 325
9.1. The 144-Days Plan and Inward Journey 326
9.2. Outward Journey 330
9.3. Captain Scott’s Food Supply and Glossopteris Indica 345
9.4. Food Shortages on the Barrier 365
9.5. Lt Shackleton’s Shadow 376
9.6. Synopsis 377
9.7. Appendix to Section 9.3 379
10. Silent Mutiny at Cape Evans 385
10.1. George C. Simpson 387
10.2. Cecil H. Meares 392
10.3. Edward L. Atkinson 397
10.4. Apsley Cherry-Garrard and Edward Atkinson 420
10.5. Historical Scrutiny 427
10.5.1. Huntford’s Blunder 430
10.5.2. Thoughtless History Re-writing 432
10.6. The 11 Miles Myth 439
10.6.1. The Myth 439
10.6.2. Tacit Cover-up: How 22 Became 11 444
10.7. Synopsis 453
10.8. Appendix to Chapter 10 457
10.8.1. IV. – Instructions for Dog Teams 457
Volume III
11. Captain Robert F. Scott: An Apology 463
11.1. The Causes of the Disaster – Rebuttal 465
11.1.1. The Loss of Ponies in March 1911 466
11.1.2. (79°28½′S, 170°E) 470
11.1.3. Complexity of Transportation Methods 472
11.1.4. Misuse of the dog team 482
11.1.5. Navigation and Navigation Methods 494
11.1.6. Gale at 83°S 502
11.1.7. Soft Snow at Beardmore Glacier 506
11.1.8. The Fifth Man 510
11.1.9. Fuel Leakage 523
11.1.10. Food Shortages on the Barrier 524
11.1.11. Collecting and Hauling Geological Specimens 529
11.1.12. Vitamin Deficiency 533
11.1.13. Neglecting the Sick 549
11.1.14. Route Marking and Depôt Laying 558
11.1.15. Summary 562
11.2. The Two Black Flags Axiology 565
11.3. Surprises which did not Await Captain Scott on the Barrier 575
11.3.1. No Food Shortages 577
11.3.2. No Fuel Shortages 579
11.3.3. No Extreme Cold Snap 580
11.3.4. No Never Ending Gale 583
11.3.5. Captain Scott’s Meteorological Data Fabrications 584
11.3.6. Historians’ Weather Data Fabrications 585
11.4. Synopsis 585
12. Etiology of Captain Robert F. Scott’s Death 588
12.1. Captain Scott’s Deus ex Machina 600
12.1.1. Utilitarianism Crucible 600
12.1.2. Enough Scientific Tilt To Be Convenient 604
12.2. Captain Scott’s “Message to the Public”: Submission to Nature, Nation and Deity? 606
12.3. Resurrection of Captain Scott’s Deus ex Machina 610
12.4. Denouement – I Have a Tale to Tell 621
12.4.1. Scientific and circumstantial evidence of altruistic suicide 628
12.4.2. Was Captain Scott a Scientist? 644
12.5. Synopsis 646
12.6. Appendix to Chapter 12 and Subsection 12.2.1. 647
12.7. Appendix to Chapter 12 and Section 12.3. 650
13. Synopsis – Never Again 652
14. Appendixes 656
14.1. Appendix 1 – Geographical Locations 656
14.2. Appendix 2 – Errors and Fallacies in Drs Solomon and Stearns paper On the Role of the Weather in the Deaths of R. F. Scott and his Companions 658
14.3. Appendix 3 – Data Dragging and Fabrication in Dr Solomon’s book The Coldest March: Scott’s Fatal Antarctic Expedition 668
14.4. Appendix 4 – The Whistleblower 683
Coda 691
Interview 694
Notes and References 709
14.1. Appendix 1 – Geographical Locations 656
14.2. Appendix 2 – Errors and Fallacies in Drs Solomon and Stearns paper On the Role of the Weather in the Deaths of R. F. Scott and his Companions 658
14.3. Appendix 3 – Data Dragging and Fabrication in Dr Solomon’s book The Coldest March: Scott’s Fatal Antarctic Expedition 668
14.4. Appendix 4 – The Whistleblower 683
Coda 691
Interview 694
Notes and References 709
“It has been over 6 months since your previous posting on this forum. I am wondering whether you have reconsidered your stance on validation of your ANN weather simulation model software, in particular the assumption that the McMurdo AWS automated weather station is a valid proxy for the location where Simpson made his measurements at the top of Wind Vane Hill in 1911/12.”
Far from reconsidered. In section 7.4.3.2 of his book, Sienicki deals with the temperature gradient between the McMurdo AWS and Cape Evans. I quote from it with his permission:
“Due to the obvious shortage of temperature data, I am not able to do a precise evaluation of the observed negative gradient on artificial neural network training and subsequent data retrodiction. However, since the presence of the negative temperature gradient between Cape Evans and Hut Point (i.e. McMurdo) was deduced, I can use my artificial neural network to simulate an inferred temperature difference.
My neural network is ideally suited to answer any question as to how an eventual temperature difference would affect the overall prediction and retrodiction procedures related to the discussed case.
Provided that the calculation of the average temperature difference for the historical data depicted on Fig. 7.12 is representative, one can obtain the difference of 2.45°F for Hut Point (McMurdo). In order to be on the safe side, one can add an extra margin for temperature difference – say half of this value – and obtain a difference of about 3.6°F. Additionally, because temperature fluctuations may occur, it is safe to assume that the expected temperature difference between these two sites will vary between ±3.6°F (±2°C).
The whole procedure of neural network training and retrodiction was again performed, but this time with the plus/minus perturbed minimum daily temperatures. Thus, the whole set of data was obtained and an additional neural network retrodiction run on this historical data was performed. The results are depicted on Fig. 7.13.
The upper dashed curve was obtained after the McMurdo data was perturbed by +3.6°F and the lower dashed curve by –3.6°F, respectively. This numerical experiment shows that even if there was a slight temperature dependence due to intricate meteorological and/or physical features at Ross Island and McMurdo Sound, it was negligible and insignificant in accounting for the differences as depicted on Fig. 7.8 between Captain Scott’s and the retrodicted temperature data. Simultaneously, this numerical experiment further confirms that the artificial neural network used in this work is sensitive and capable of responding to fine temperature changes in training and analysis of data.”
As of currently, I have yet to see you provide an answer to Simpson’s discussion of how temperature inversions in Antarctica are formed. Your argument that no one can predict temperatures 21 meters away if an inversion happens is a joke, because of Sienicki’s above refutation, and because you mix vertical distances with horizontal distances. The latter alone discredits you when you happily use 21 meters of vertical distance as comparable to 200 miles of horizontal distance. Not that your argument matters in the case of an artificial neural network, which is not ruined by less than perfect accuracy.
I believe that my previous suspicion of Scott’s May 3, 1911 observation was incorrect. Let me draw your attention to Scott’s diary entry for June 3, 1911:
“The wind dropped last night, but at 4 A.M. suddenly sprang up from a dead calm to 30 miles an hour. Almost instantaneously, certainly within the space of one minute, there was a temperature rise of nine degrees. It is the most extraordinary and interesting example of a rise of temperature with a southerly wind that I can remember. It is certainly difficult to account for unless we imagine that during the calm the surface layer of cold air is extremely thin and that there is a steep inverted gradient.”
Examination of Simpson’s Volume III Table 1 confirms Scott’s observation, with a temperature of -5 F at 3 A.M. rising to +3 F at 4 A.M. What further proof do you need that wind prevents inversions, and therefore Scott should have experienced and recorded fluctuations in temperatures in his diary?
And you bring up the validation canard again. Again, software validation applies only to software which demands absolute accuracy in output, like a program which performs 1+1 and gets 2. If it performs 1+1 and gets 3, then there is a bug. Software validation does not apply to artificial neural networks, since they do not have a need for absolute accuracy in output. Additionally, software bugs are irrelevant if the ANN achieves a desired accuracy anyway. Your ignorance is not surprising. As your probable LinkedIn page shows, you do not have the smallest experience with ANNs.
“Personally, I was disappointed that a professional meteorologist did not join the blog to provide an expert opinion on your assertions. That would have been the natural complement to my well-founded concerns in the area of software testing. Never mind!”
Why is a professional meteorologist necessary? For all of your rhetoric, you certainly haven’t brought a professional meteorologist in yourself. That says it all about your lack of desire for genuine critique.
“I guess the crucial thing is whether you have the confidence to publish your opinions without providing evidence of independent verification of your software model.”
I certainly haven’t seen you do any validation yourself. All we have seen from you is rhetoric, and repeating debunked arguments.
Kristoffer
It is great to see that a new research article, ‘An Exceptional Summer during the South Pole Race of 1911-1912’, has been published by the prestigious American Meteorological Society. It may be viewed at: http://journals.ametsoc.org/doi/pdf/10.1175/BAMS-D-17-0013.1
I like the article and recommend it to others because:
1. It provides an even-handed non-partisan analysis, including Amundsen’s meteorological observations, and it addresses the exceptionally warm period that Amundsen experienced in December 1911.
2. It does not propose any bizarre theories about Scott falsifying his temperature records in order to pave the way for his own suicide. Fogt et al simply show that most of the extreme temperature and pressure observations during this exceptional season are about two standard deviations above or below the mean value in the climatology model they used, called ERA-Int. Scott’s cold March temperature observations are no more extreme that Amundsen’s high December temperature observations. [Noting that observations of under-sledge temperatures were excluded because pooled cold air could cause a cold bias].
3. With a lifetime of experience in IT and software project management, I have wide personal experience in reviewing and probing the adequacy of testing that has been carried out in development of complex systems. Fogt’s article rings true for me.
4. The article has been through the peer review process of a respectable research institution.
Thank you
I am really pleased at your balanced, informed comments and will comment further when I have re-considered this contentious subject
Isobel
I have read Bill’s comments several times and am delighted at the way he summarises the paper from the American Meteorological Society about Scott’s last fatal journey, recording the exceptionally warm weather experienced in December 1911 and ignoring theories that Scott and Bowers falsified their meteorological results in the latter part of their expedition. I am going to put it in full on my blog.
https://journals.ametsoc.org/view/journals/bams/99/10/bams-d-17-0282.1.xml
Thanks for your comment about Kristoffer and Professor Sienicki being the same person.
Are you willing to share your evidence via this blog? If so, I suggest it would gain greater readership by being posted as a comment against the latest entry of the post “Further comment on Scott’s suicide”, rather than its current position in the middle of an older post.
Thanks again for your insightful comment.
Sorry, I should have said that my previous reply is in response to a comment by Sock Puppet dated 28 December, positioned much further up this thread. I was expecting my comment to follow directly after Sock Puppet’s.