Home > AGW, Catastrophism, Climate Change, Global Warming, Omniclimate, Science, Skepticism > So You Believe In Computer Models…

So You Believe In Computer Models…

Shub Niggurath tries to explain what makes Nature magazine reluctant from asking authors to make their computer code public. I am not too convinced by the arguments listed as possible reasons, and prefer to blame sheer ignorance on the part of people with little experience in programming.

On the other hand, SN’s post has made it very clear to me what’s wrong with my approach to “climate science”. The problem is that I became a “computer scientist” before I became a “scientist”. This makes any computer code highly suspicious in my eyes, and therefore it also makes next-to-impossible for me to trust “computer models” of any kind.

Nowadays, more than 18 programming years later, I am amazed at learning that there are intelligent people out there unaware of the infinite range of results that can come out of a relatively simple group of programs. Those people have never supported an application written by somebody else. They have also never experienced the frustration of having to debug a piece of code.

A computer, as the saying goes, is an ass, i.e. it will only do what asked and will only do it literally. Anybody programming it without the necessary level of expertise is bound to be lead around by the ass. In other words, the vast majority of scientific papers based on computer models are very likely to be elegantly-thought garbage.

Advertisements
  1. UzUrBrain
    2011/03/29 at 19:54

    Years ago I was taught the ramifications of computers. I Got My BS with a slide rule and a few years later when taking graduate courses the professor allowed the use of the new fangled pocket calculator (HP-35). I was glad to use it as I had been using it on my job. However, at least half of the problems came up with an answer different than that in the back of the book. Some were as much as 50% different. I spent hours trying to find what I did wrong. Upon asking the question in class he explained how the answers were based upon “slide rule” calculations, and that since the calculator takes angles and the sin/cos, tan/cot out several places before they are added/subtracted/multiplied, etc., the small differences can cause a big difference in the answer. The same problems arise with computer languages that only use a limited number of digits of accuracy and the conversion problems caused by these limitations from decimal to binary. I remember using a “trick” program that upon reading its code, for all practical purposes would give the correct answer. However, once programmed, when you put in the numbers you got out the wrong answer every time! Even after debugging, flowcharting, going through step by step and mentally determining the result of each computer action, etc, you could not find anything wrong with the code. The program relied upon the known deficiencies of the decimal/binary conversion of the numbers in that language (IBM Fortran.)

  2. Paul Penrose
    2011/03/03 at 22:09

    I agree with the head post completely. After over 25 years as a software engineer I can confidently say that unverified computer code is highly suspect – to the point of being virtually useless for anything other than playing around. So even if you believe the GCM designs are correct (ie. they got the physics right), there is no way you should trust the output since none of them have been validated and verified. It’s really that simple. You wouldn’t drive across a bridge that had been designed on the back of a cocktail napkin, would you?

  3. 2011/03/02 at 13:57

    Well, the reasons offered in my post for the ‘reluctance’ are not mine. 🙂 But they must be real because Nature is, after all siding with those who don’t want to share their code.

    It is a toxic synergy in the end. There are obviously a range of (ultimately stupid) reasons why researchers do not want to give out their code. These lead immediately to very sound business reasons why journals do not want to make code requirements a condition for publication. Both parties however cover their asses by providing the same high-flown reasons to why they don’t want to distribute code. ‘It is the principle in science. Any researcher should be able to build back what we did, from our description. If he or she cannot, they are not good scientists’.

  1. 2011/03/02 at 06:47

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: