1. Matching historic data is a poor test of the computer models, because all of the computer models were tested against historic data and tweaked to match that data before they were published.

   If there were a published model of the climate system in which all the parameters were empirically measureable constants, the historic matching would be a useful form of verification. None of the IPCC models come even remotely close to this. Most have hundreds of parameters which were set PRIMARILLY based on their ability to match the historical data.

2. In 1890, and in every decade afterwards, there were thinkers who postulated that the earth would get warmer or colder. All of them, including Svante Arrhenius were off by a substantial margin. There is no virtually no global climate pattern that, if it occured in 2007, could not be "validated" by pointing to a scientist from before the computer era who predicted it.
3. The IPCC models invalidate each other. For most combinations of Model A, Model B and Scenario S, Model A predicts global temperatures in 2100 which are more than two standard deviations from what Model B predicts.

   The IPCC is not confident enough in its understanding of future climate to identify one model which will produce accurate results for the future. Why should we believe any one.

   I wish the IPCC would get behind one model. If the model successfully predicted the next 20, that would help validate or disprove the model. If you choose a handful of models, and one happens to be right, you've just proven that if you pick enough models, one of them would be right.

4. Models based on historical results have the following property:

   The continue to accurately predict the future until they stop predicting the future.

   A few decades ago, Economists believed the Philips curve with such fervor that (as now) skeptics could not get a tenured position or find a journal willing to publish their work (poor Milton Friedman).

   The fact of the matter is that for a very long time the Philips curve was an accurate model of the economy. Then one day, it wasn't.

5. Global warming models suffer from one disadvantage that Keynesian economics did not. While Keynesians were predicting an economy that continued to operate within a linear extrapolation of historic values, global warming models specifically predict a nonlinear acceleration in the rate at which the world is warming.

   In other words, models based on historical results tend to be more accurate when the future they predict is similar to the historical data they are based on. Global warming models predict a future that is DIFFERENT from the historical data that they are based on. It is therefore reasonable to assume that they will have problems predicting that different world.

In sum, Human kind has never succesfully modeled and verified a system even remotely as complicated as the climate system. It is good that we are attempting to verify the models, but we have yet to produce a single model which is generally believed to be an accurate representation of the world around us.

When such a model is produced, it may accurately model the Earth's climate for an extended period of time. If so, it would be an historic achievement.

Until then, it reasonable to assume that, like all other models for systems as complicated as the global climate and human behavior, it will remain accurate right up until it stops remaining accurate, and no longer.

>>> models predict warming of ocean surface waters, as is now observed; >>>

Yes the seas are heating but
LOL what model ?????  The Global Warming based upon the greenhouse gases model is one drawn from historical data of natural events (ie high salinity sea)
but the current climatic situation is not natural
You can not appeal to past events to model the future in this case.

If CO2 was to blame for GCC,,,, Then if the sea heated up, guess what, water evaporation would increases thus cooling the sea, and the cloud cover would increase, thus cooling the sea.

To actually heat the sea is rather difficult because there are so many checks and balances

The only way the sea can be heated is by preventing water evaporation,,,,, heat then has to stay bound to the water, cloud cover would diminish and so causing a further heating of the sea.
In the past high sea salinity preceded an Ice age.... today the inhibitor of sea water evaporation is a petroleum oil film.

Nice to see selected BS being quoted as proof THEY have their act together.
More like the Laurel and Hardy buffoons.
 

Despite the incredible complexity of the models I cannot see any improvement on Arrhenius' postulations over 100 years ago.  I still see no justification for this violation of the principle 'entia non sunt multiplicanda praeter necessitatem'.

I see the models as little more than a distraction - regardless of how wrong the models are (and I view them as all being wrong unless proven otherwise), the fact remains that the earth is warming at a great rate, solar output and geometry alone cannot account for the rate, results from various experiments on "CO2 fertilization" and "greenhouse response" of various plants do not suggest that natural CO2 production is increasing at a rate which comes anywhere near anthropogenic increases in production (so nature isn't competing with humans for #1 producer of CO2), the anthropogenic production of CO2 is far larger than would account for the entire observed global increase (which means there is a large sink operating, but that sink is just not good enough).  Add to all that the fact that CO2 will cause increased warming.  Personally I think the IPCC statement that there is a 90% probability that humans are contributing to the warming is both misleading and nonsensical.  Since the increase in global CO2 is, for all practical purposes, entirely due to human activity, I would turn the matter around and say that humans are ultimately responsible for 100% of the warming which cannot be accounted for by solar output and geometry.

Yes, therefore nobody regards that as a valid test, including the IPCC et al. Therefore, people evaluate the models on their predictive ability; for instance, see
http://gristmill.grist.org/images/user/6932/hansen88.gif for a ten year validation of Hansen's model, from eighteen years ago. It's even more impressive if you continue to plot the data for the past 8 years. Or would you claim that although the validation data is from 18 years in the model's future to you it's still historical data?

"If there were a published model of the climate system in which all the parameters were empirically measureable constants, the historic matching would be a useful form of verification. None of the IPCC models come even remotely close to this. Most have hundreds of parameters which were set PRIMARILLY based on their ability to match the historical data."

I'd love to see any models which meet your constraints, in any field of science, economics, whatever. Models come in the form of
Estimate=(AX+BY)/CZ etc.
where X, Y, and Z are measured parameters and A, B, and C are the coefficients which are chosen to reduce the error of estimated vs. measured data to a minimum. Then the model is validated by feeding it new data and seeing how well the estimates coming out match the actuals. THERE ARE NO MODELS WHICH DO NOT HAVE COEFFICIENTS WHICH ARE ADJUSTED TO MAKE THE CURVES FIT THE MODELING DATA!!!!! IN ANY FIELD OF HUMAN ENDEAVOR!!! If you have something that does not have adjustable coefficients, it isn't modeling, it's just arithmetic!!!!! Furthermore, the coefficients used in the climate models are constrained to the ranges found in the published literature; if not, then there would be somebody pointing out the clinkers, instead of this general handwaving, "The models are fit to the data! That's cheating!".  

Once again: what part of this 10 years of validation of what a climate model from a state of the art which is now ancient, do you guys find unsatisfactory?
http://gristmill.grist.org/images/user/6932/hansen88.gif

"Models based on historical results have the following property: They continue to accurately predict the future until they stop predicting the future."

Right. And when the climate models stop accurately predicting the future, then we can add whatever term is THEN throwing them off. Until then, it would seem they ARE accurately predicting the future. No?

Actually hurricane modeling works just this way. Many parameters have been tweaked, even algorithms have been redone, and finally after years of marginal utility, they worked well enough to give cause to evacuate whole cities, and folks think on balance that using the models are worth the false positives. It is worth noting that there is a suite of models which are weighted according to skill scores and put together to get temporal/spatial probability distributions. NWS endorses no single model, but uses several (redundant systems), as one would expect in any important risk assessment.

In fact these models are not being developed just as a jobs program; the economy and national interests need to long term plans and climatological risk assessment is essential in many areas, for example water supplies and agriculture.  LLNL is hosting model comparison data, which is essential step on the way to computing probabilities similar to hurricane predictions. This will give realistic expected values drought/heat damage/morbidity, revised habitat ranges, Siberian agricultural yields, Northwest Passage seasonal transit, etc. In short we will be able to expected costs of AGW and compare them to current costs of mitigation.

I think we have a pretty good idea various damage and casualty costs of various scenarios; now we need to come to better agreement on probabilities.  It is too risky not to be using climate models, even if they are imperfect.

"First the snide answer: the engine processor in your car has numerous parameters that get tweaked to keep it running, but I don't hear any complaints."

Well, I can't evaluate that image better than by sort of squinting at it, so bear with me.  Assuming that the image was produced from 1988 models, I can see the models roughly correspond to reality from 1988 to 1998.  But again, just from looking at it, the correspondence is not striking.  In fact it seems to me that if you were extrapolate a linear fit on temperatures from 1960 to 1988, then add some random bumps, you would do about as well.  The trough around 1996 or so in models B and C does a bit better than that, granted, but not enough to make the resemblance particularly remarkable, as far as visual inspection goes.

So basically, what I'm looking for is a model that made a prediction that required CO_2 to be the cause of the warming.  Something that just extrapolates prior trends doesn't say anything about the causes of those trends.  Now, perhaps the graph you posted is an example of such a model, but I can't see that from just looking at it.  Do you have any more precise analysis?  For instance, something that successfully predicted a substantial change in an existing trend, repeatedly, with comparison to the performance of other models that were considered comparably serious at the time the original was released?

Mind you, I'm willing to accept that on the balance of the evidence, human-released CO_2 is probably causing global warming.  I haven't seen anything that convinces me that it's very likely, or almost certain, as various climatologists say it is.  I have read the IPCC AR4 FAQ, but that didn't really answer my questions.