A new study to be released in the Monthly Weather Review provides yet another nail in the coffin on the climate models that the claims catastrophic anthropogenic global warming AKA alarmism are based on. A group of Korean researchers did something that in retrospect is surprisingly hadn't been done before. They ran two major climate models the global model program (GMP), and the Global/Regional Integrated Model system (GRIMs) on 10 different computer systems to see just how much of the model results were the result of the actual physics being modeled and how much was simply the result of imperfections in how computers process, such as rounding errors. What they found was staggering. http://journals.ametsoc.org/doi/abs/10.1175/MWR-D-12-00352.1?af=R This is an issue that is well known in most fields. With all iterative computer models as you extrapolate further and further into the future rounding errors compound and compound to the point that they become extremely significant. Such problems caused absolute panic in the banking industry in the 90s. Climate models attempt to model the climate which is so chaotic that Edward Lorenz the pioneer of Chaos Theory used it as a prime example of a chaotic indeterminate system. However, never put it beyond some egotistical hack to think that they can model it. There is a good discussion going on at JoNova about the new study. http://joannenova.com.au/2013/07/oo...uce-different-results-on-different-computers/ What I cant get is that this hadn't been done before. Such tests are supposed to be common practice however, it never occurred to anyone in climate science that their models could be effected by rounding error. While this is far from a death blow its just another drop in the bucket of the drip drip that is destroying the credibility of climate models.
Very interesting, but do I believe this is the first time someone has noticed the variation? No I do not.
You do realize the study is saying that the variations introduced is simply approximately equal to the variation introduced by the marginally different initial climate figures. This does not indicate a difference that invalidates the predictions of tendencies or outcomes of the different models. Its simply affects, to a minor level, the intensity and amount of time in which to expect those results. Because different models of breathalyzers vary 5% or something like that in their measurements does INDICATE that a .12% alcohol blood level readings could mean the driver might be sober.
One must understand what the temperature anomaly is when talking about climate models. A climate model is attempting to model the climate in whole. Temperature in a climate model is in whole degrees. When we look at the anomaly we are only looking at how much that those whole degrees have varied from the base line. This is in tenths of a degree. When we run an iterative model out 100 years small rounding errors and computational approximations become significant. I think you need some preservative on how significant that is. When the IPCC runs a serious of model runs by adjusting the conditions the variance between the runs is greater than the actual average anomaly from the base line.
That is a good point. At times I forget just who I'm dealing with and fall back on the belief in the common good of men. The history of climate science is replete with examples of climate scientists failing to mention or publish negative results regardless of how significant they may be. Dr. Mann ran his model without the questionable bristle cone series and found that he couldn't get a hockey stick without them, proving that his hockey stick was simply the result of using a bad proxy, so he tucked the results away in a folder labeled "censored" and forgot about it. In the climategate e-mails we see that the scientists actually verified McIntyre's results and talked about it in private but never mentioned that publicly and continued to attack him. I can go on and on. A negative result should be discussed and published especially when it is as significant as this. You are probably right that at some point modelers did test their models to see just how much of the result compounded error through iterations. Sadly however, what ever results they got they decided to keep a secret.
Rounding errors? Are these programmers stupid? Such problems are a sign of sloppy programming, or bad math understanding, or insufficient numeric precision. They must have used floating point. And forgotten the order of math operations. That's very bad. I'm sorry to be rude, because I don't have access to the program code to see what they did wrong, but this kind of rounding error is a nit-pick of mine. Floating-point math is known to cause accuracy problems (and the math libraries that go with it). It shouldn't be used for advanced, scientific math, and actually it should also be used for nothing that results in an answer with more than 7-16 digits of accuracy and precision. Fortunately, the 1990s was the last time computers were slow enough that programmers still needed to use the native, built-in CPU limits of 32-bit/64-bit precision. And that's why there's the creation of Arbitrary-precision arithmetic. If they were to calculate the math to 100s or 1000s of digits of precision, while using the schoolbook method, there are few people who could dispute the results.
We have a winner! Time for Fun with (*)(*)(*)(*)ty floating point programming!!! Open your windows calculator (standard mode). Enter 4 then click "sqrt" or "√" depending on version, which will give us 2. Now subtract 2. Linux vs Windows Think that might have a significant effect on modeling? Comes from using rads in the trig instead of degrees or grads. 20 years later... and all MS stuff uses this methodology.
"Whole degrees"?? Sorry, climate modelling math goes so incredibly far beyond "whole degrees". Totally whacked. Where did you make that up from? Your source says the difference is approximately similar to the variance among the INITIAL TEMPERATURES used by different modelers. This article is in fact ONLY pointing out that concerns over initial starting temps given (which have some variation among data sets used) is a LESS significant issue than a particular operating system's 128 bit (minimum) math rounding errors. Sadly for the Denialists abuse of this study as some kind of massive indictment of climate modeling, the study is pointing out SMALL errors (on a planetary scale), not BIG ones
Strawman, I never said that climate models only model temperature. I said that the climate models model the climate in whole. And the temperature in climate models is in whole degrees. In other words temperature is only one part of a climate model. Stop with the bogus arguments. Where did you make that up from?[/quote] A climate model is not just initialized from the temperature many factors go into the initialization see here and how a model is initialized has a significant effect. http://hockeyschtick.blogspot.com/2013/07/new-paper-finds-in-retrospect-we.html Initial conditions is not just the initial temperatures. And nice try attempting to flip the study on its head. The title of the study is "An Evaluation of the Software System Dependency of a Global Atmospheric Model" Not 'evaluation on initial conditions'. This is just another attempt at warmmonger double speak that has been going on in warmmonger blogs to make themselves feel better sadly only Nick Stokes and a few others have dared to try and make the argument in public because its bogus.
All the initial conditions input into a particular climate model can have minor variations between the data sets, ALL the data, not just temperature. But your attempt to portray a relatively minor issue into something negates a "catastrophic anthropogenic global warming AKA alarmism" conclusion from the models is wrong ALL the models on ALL the operating system platforms unequivocally show a catastrophic effect of human activity on the current climate. This study conclusion that you quote is simply saying that the minor differences which set of initial data used is less significant than the different system's rounding errors. To put it simply, the catastrophe may be 20 years earlier or later, or maximum temp changes may vary .2 degrees, but NOTHING about this study indicates there is ANY possibility that the ""catastrophic anthropogenic global warming AKA alarmism" is NOT going to occur. This thread is just another attempt a Denialist double speak that is going on in Denialist blogs to continue a fossil fuel subsidized program of sabotaging any attempts to reduce mankind's effects on climate and the massive profits our fossil fuel cartels derive from remaining the sole source for an ever-depleting energy product.
