Being a software engineer, I think I’m somewhat qualified to talk about computer models, at least moreso than my other musings about science in general.
In a converation on a climate site, I replied to someone inquiring about modeling. I said –
“Climate science is very difficult. Climate scientists don’t have a planet sitting nearby that they can use to run experiments with, so they’re limited to simulating things with computer models. I’m not going to argue GIGO (garbage in garbage out) because it’s not accurate. But say you are looking for a specific thing, like trying to guess how much CO2 it takes for the climate to do something wierd. Your model looks at that, but that’s because the model is made to look at that. It’s not made to discover new physical processes or climate influences. It can only use the influences assumed at the start.
You can therefore trust that scientists are working diligently and reporting to the best of their abilities. Nor are they misguided; they don’t know what they don’t know, so they are telling you what they DO know to the best of their ability.”
Now, to put this in some perspective, a recent paper from an astrophysicist (Svensmark) postulates that changes in the sun’s geomagnetic flux changes the amount of cosmic rays bombarding earth, and these rays play a fairly big role in cloud formation. Cloud formation changes the temperature. From what I can tell from reading the paper, this looks to be correct in essence, although the significance of the role is unknown until experiments are run (and they are scheduled to run soon at the CERN facility.)
Now, back to models. As I’d said, models do what models do and can’t discover processes; they have to have known processes as inputs so that they can simulate. Thus we have a case where ALL current models are ignoring a previously unknown physical process and one that a model can’t guess at; this reinforces the point that models don’t discover processes. More importantly, it says that models can only be trusted as far as how well the science of the inputs are understood.
Models are evolving. In the late 80’s models used by James hansen at NASA treated the oceans in a fairly primitive matter because the assumption was that this explained things well enough to model for CO2. Ahhh, but later it was discovered that this was inadequate, so more recent models do a better job of handling oceans and currents. The question is whether this is adequate now or merely “somewhat less inadequate” because important physical processes are being downplayed, aren’t known, or perhaps even wrong. (This happens.)
Overall the point is that models used by the climate guys are designed solely to simulate the effects of CO2, so when the conclusion reached of the interpreted data has something to do with CO2, are we supposed to be surprised?
