ROCKMAN wrote:Plant -As you point out models are difficult. Especially since there is no such thing as an UNBIASED model. Every complex model requires many dozens (if not hundreds) of assumptions built into the model. And those assumptions always run in a range. And often a very wide range. So the modeler has to chose those values. And all to often the modeler has an objective. Human nature being what it is assumptions are chosen leading to the goal. Which isn't to say they are incorrect. But also not to say they are correct. And typically those assumptions rarely have any absolute measure of validity.
Thus one is really saying they agree with the assumptions made when they agree with the model. Which is often just agreeing with the modeler's OPINIONS.
I agree with you that the built-in assumptions to every model really complicate things, and some of those built-in assumptions are pretty subtle.
One of the primary tools used in meteorology for forecasting weather, hurricanes etc., is the 'ensemble forecast', where you take a bunch of models, put them together, and see where they agree and disagree.
You can also run your model, change the initial conditions slightly, run it again, change the initial conditions again, run it again, etc. You wind up with a bunch of models each with slightly different starting conditions. When you put them together you can see where the slightly different initial conditions don't make much of a difference to the final output.
With meteorology and climatology this is important as so many factors are not linear, and a minor change in one initial parameter can mean major changes as you move forward in time.
The other problem is that even with supercomputers, this is very time consuming and there often isn't enough time to run too many models covering too big an area. This does work for hurricanes though and is common .
Some of the modeller's I know frequently do 'back forecasts' and 'model verification', checking this week to see how last weeks forecast conformed to reality, and where the model was wrong, try to figure out whether they screwed something up with the initial conditions, or if there were other factors that came into play that they didn't expect.
The best modeller's take it very seriously. But there are a lot of junk models out there.