I wouldn't call them fudge factors. These are what are called finite element models. Gross oversimplification. But they break the atmosphere up into boxes and then treat each box as a point. Then they model the interactions between different points. So for each time step, each box affects the boxes around it and vice versa.
The problem is, they do not understand the physics of a lot of the atmospheric dynamics. So they have to "parameterize" the model. A parameter is a number that you multiply times something or add to something in the model. It doesn't have any particular value; so you have to set it from the known data. There's nothing inherently wrong with this approach, as long as the process of setting it from the data is done in a statistically rigorous manner.
But this is not possible for the global warming models because they need lots of parameters and they don't have much historic data to fit. What data there is is VERY noisy.
As a general rule, the more parameters you have and the more noise you have, the more data you need to fit your parameters in a proper manner. Loosely speaking, when you don't have enough data to do that, you have an "overdetermined" model. Another term we use is that the model is "overfit." There are ways of avoiding overfitting but I see none of them being talked about.
There is a further problem. Nothing about the AGW models is surprising. That is, you can do a pretty good job of fitting the warming data since 1930 with a ruler and a pencil. In other words, any model that predicts global warming will increase fits the data pretty well. The ruler and pencil is a linear equation with only two parameters. You can do a better job than that using solar activity as a predictor, again with few parameters.
As a general principle, if two models do an equally good job of fitting the data and one is simpler, you choose the simpler model--it is more likely to be robust.
A further problem. When the models don't fit the data (the cooling or at least the leveling off of temperature increases since 1998), there is no reexamination of the models by the AGW folks. They just invent new factors that are temporarily 'masking' the effects their models predict. This is exactly what the Copericans did to defend their celestial spheres models of the Solar System. Add more spheres. In effect the AGW folks add more parameters.
Another problem: these are feedback models. Such models usually are exponential--in this case they will predict huge increases in the future. But exponential models start with an effectively linear section. What the AGW folks are doing is fitting the linear section of their models to the known data and then presuming to project into the future using the exponentially increasing portion of the models. Outrageously bad science.
Finally, the only way to actually validate these models is to make a fixed prediction at one point in time that says, given the following inputs, the following will be temperature and we predict that the 95% confidence interval around that prediction is xyz. Then you monitor the model (along with simpler models) over the years. But the AGW models are a moving target. They have new models every couple of months. Even for those, they don't provide rigorous forecasts with confidence levels. But at a more fundamental level, complex models like this need to predict something surprising, something the simpler models do not predict. They need to predict them in advance and then come true. A classic example is Einsteins prediction that mass would bend light waves--something Newtonian physics did not predict. So far, we have seen nothing remotely approaching that.
PS: you explain things well!
When I get home from work, I’m going to reread your reply.
But as I read your statement I’m reminded of a chapter I read some years ago in a book called “Chaos” about a guy named Lorenz and his “butterfly effect”. It seems to me that those AGW folks are about 12 decimal places less accurate than he was talking about.