First of all, it is important to realize that we always make models for something and not of something. They are made for a certain purpose but they are not models of something per se. Take the maximally simple models—they are made to describe a fundamental connection in a simple way. You cannot expect that they reproduce the evolution of the climate system in detail, but their purpose is to generate understanding of key mechanisms. One example is the model of the geostrophic wind: This is a wind that is characterized by a flow direction parallel to the isobars (and not from high pressure regions towards lower pressure regions, as one might naively expect). Here we have a simple model already. With the simple models, I can generate understanding but rarely (meaningful) accurate numbers. But it is these types of models that we have in mind, when we say ''I understand a mechanism''.
In the GCMs, on the other hand, I try to include as many processes as possible-of course preferably those ones that I expect to be crucial for the problem I am investigating. That are not only first order processes but one also tries to include higher order processes—second order, maybe even third order. The limits are only set by the available computer-power. But these models do not generate understanding, they provide an experimental platform, where we can change or adjust parameters. For example, I can try out what happens to the climate when I remove Australia from my model world. But this does not imply that I understand why things are changing then, for this I again need the simple models. They can help me to understand the mechanisms.
Was this article helpful?