Someone told me back during the days when I took the anthropogenic global warming claims at face value that if I wanted to completely lose faith in the climate models I should take a look at them. And so I did, and he was right.
I looked at the code for the model used by GISS to do the predictions for the IPCC report of 2001. What is striking is how crude the model is. It divides the earth into a grid with about 8000 squares. The continental US is therefore represented by a 16 by 36 grid. Then there is what is not in the model. There's no way to model clouds and rain, so they have to make a guess at what those will be doing in response to changes in other factors. Consider that a 2% change in average cloud cover would completely negate any warming due to CO2 doubling. And the natural variability of cloud cover year to year and decade to decade is much larger than 2 percent.
The assumption of the modelers has been that a slight increase in temperature due to CO2 would be magnified by the resulting increase in water vapor in the air. But that might also result in an increase in cloud formation, which would offset the water vapor feedback.
Satellite measurements of clouds, water vapor, and temperature in the tropics show cycles. Heat forcing increases, temperatures increase, water vapor increases, and then clouds increase resulting, apparently, in a fall in temps. Rinse and repeat.
And the distribution of water vapor in the atmosphere is not uniform. Even over the open oceans the humidity is less than 100% most of the time, and it varies constantly in response to other factors. Satellite pictures of water vapor show that it swirls around in a more or less random fashion in the atmosphere. And CO2 does the same thing. Simple assumptions about these things won't do. And the models have a host of other fudge factors that can be tweaked endlessly until the model coughs up the desired results. The agreement between various modelers is not evidence of the validity of the results.
In any case, I have no faith in the computer models, having seen how they work.
More recently I became interested in surface temperature measurements, as reflected in previous posts. I wanted to know why satellite measurements of global temperatures don't agree with surface measurements. And so I looked at the data from GISS. I didn't appreciate the problem until I started trying to come up with a global average from station measurements. And then I found that the stations only represent about 25% of the earth's surface. I don't care how clever you are with statistics, there is no way in hell that you can come up with an accurate measurement of global temperatures if you're not even looking at three fourths of the globe, and that's really all I need to explain the difference between surface and satellite results. Trends can be determined by using a limited set of measurements, but as one area goes up another might be going down, as we have often found is the case.
Therefore, the four alarm increases in temperatures, according to GISS, NOAA, and others, over the last 30 years are probably not valid. Much closer to accurate is the satellite data, which shows a much more sedate rate of increase on the order of 0.14 degrees C per decade since 1972. The disagreement between satellite and temperature data has given rise to efforts to reconcile the two mainly by "adjusting" the satellite data using computer models(!) among other things. Which is doing it exactly the wrong way around.