Quote:
Originally Posted by PKFFW
1: I assume by "fail" you mean it comes up with a result that just seems totally wrong or is complete gibberish? Or is there some other type of failure like the program freezes or something?(honest questions as I thought a model simply told you what was supposed to happen so if the model say XYZ will happen then how can that be a "fail" unless we already know what will happen and then why do we need the model?)
|
More the former. For example, perhaps weather prediction models systematically fail to develop a certain kind of low pressure system rapidly enough. Scientists could then develop an observational program to intensively observe these events and then use the resulting data to improve the representation of the relevant processes in the model.
The latter is also a valid area of research. Models approximate the continuous natural world discretely, i.e., in finite time steps and at discrete grid points. As a simple example, if the time step is made too large, then the approximation may be poor, the model may be unstable, and it may never converge to a reasonable solution -- it "blows up". Improving the numerical stability of models, for example maintaining the same level of accuracy with a longer time step so that the model can be run faster, is definitely an active area of research.
Quote:
|
2: These targeted and improved observations, are they of the physical processes of the climate or are they of the model? I mean, when the model fails do you just go over the model and tweak it until it doesn't fail anymore by adjusting values and such(and then you have to ask if the model was correct and only failed because it didn't do what we wanted it to or if the model is actually broke, see question 1 above) or do you go out and look at the climate some more and try to work out what is going on in the climate that isn't accounted for in the model and that is why the model failed?
|
Both, kind of... see above.
Quote:
|
I guess my point goes more towards the fact that modelling is, by its very nature, basically an extrapolation of our understanding and assumptions combined and not empirical data. If that understanding and those assumptions are faulty in any way to begin with, then that will influence the results of the modelling.
|
Our understanding and assumptions are, however, informed by empirical observations. When this fails to be the case, then problems can, as you've shown, definitely arise.
I would say that the earth is more heavily observed now, and with a much more diverse set of observing platforms, than ever before, which is very encouraging and, from a scientific perspective, exciting.