SS Myth: Models can Hindcast. NO - they cannot - Massive Fail!

Another in our series debunking UNSkeptical UNScience's so-called Myths of Global Warming. This post debunks their "Myth" No 6.

"Models are unreliable"Models successfully reproduce temperatures since 1900 globally, by land, in the air and the ocean.

And their explanation:
Climate models have to be tested to find out if they work. We can’t wait for 30 years to see if a model is any good or not; models are tested against the past, against what we know happened. If a model can correctly predict trends from a starting point somewhere in the past, we could expect it to predict with reasonable certainty what might happen in the future.
Donald C. Morton on Dr Judith Curry's site writes:
The Validation of Climate Models 
How do we know that the models representing global or regional climate are sufficiently reliable for predictions of future conditions? First they must reproduce existing observations, a test current models are failing as the global temperatures remain nearly constant. Initiatives such as the Coupled Model Intercomparison Project 5 (CMIP5) can be useful but do not test basic assumptions such as linearity and feedback common to most models. Matching available past and present observations is a necessary condition, but never can validate a model because incorrect assumptions also could fit past data, particularly when there are many adjustable parameters. One incorrect parameter could compensate for another incorrect one. 
Again, from the SS explanation:

So all models are first tested in a process called Hindcasting. The models used to predict future global warming can accurately map past climate changes. 
Actually SS was correct but they left out two keystrokes from their explanation - added here:
So all models are first tested in a process called Hindcasting. The models used to predict future global warming can't accurately map past climate changes. 

An Extract from paper by S-I. Akasofu, reported by Science Heresy:
If 14 GCMs cannot reproduce prominent warming in the continental Arctic even qualitatively, perhaps much of this particular warming is not caused by the greenhouse effect of CO2 at all. That is to say, if it is not caused by the greenhouse effect, the warming of the continental Arctic cannot be reproduced even qualitatively by the GCMs. This would be because 14 GCMs do not contain the processes that caused the continental Arctic warming/cooling. 

More information about Dr Akasofu can be found here. 
The complete paper can be downloaded here (warning - it's 52 MB!).

Tinghai Ou
Tinghai Ou from the University of Gothenburg’s Department of Earth Sciences: 
Tinghai has analysed the model simulated extreme precipitation in China over the last 50 years.
“The results show that climate models give a poor reflection of the actual changes in extreme precipitation events that took place in China between 1961 and 2000,” he says. “Only half of the 21 analysed climate models analysed were able to reproduce the changes in some regions of China. Few models can well reproduce the nationwide change.”  (Abstract here)
Anthony Cox, agrees with the clowns at UNSkeptical UNScience (SS) in their myth mode:
Models are unreliable. Completely and utterly. Even Gavin Schmidt thinks so.
Another Shrill Climate Clown


  1. Simon, I hope you have seen this

    Andy G

  2. Even hindcasting is no proof of validity. Just for fun, I tried to generate the GISS temperature data from 1880 to the present via an iterative random walk. It does very well with only two assumptions: a step size commensurate to the average yearly difference and a criteria to iterate until the deviation from the signal is minimized. No thermodynamics or physics involved at all. That is essentially the same process used to "tweak" the inputs to the CMIP models. In fact a unit ratio test of the GISS data indicates it is well modeled by a random walk. For a reference see
    Global Warming as a Manifestation of a Random Walk, A.H. Gordon, 1991, AMS. At

    A random walk has more validity than the CMIP models.

  3. Any curve can be fit by a polynomial model to an arbitrarly acurrate level depending on how high you want to make "n" in the polynomial. Of course it has no predictive capacity at all. Ask any mathematician. To make a better fit to past temperatures back to 1900 than any climate model is a trivial matter that can be done using a programable calculator to fit a polynomial. Again, it has no predictive capability.

    1. Very easy if you are also able to adjust the data to fit the desired curve.

      You can then predict your adjusted data.

      We have seen that recently in NOAA/GISS data with the "2015 warmest evah" memo being matched by adjustments.


Post a Comment

All serious comments published after moderation.
Comments should be polite, and respect all views.
No bad language. Spam never makes it!