...that is true, but I do not directly generate new knowledge about the world, I only generate numbers, and possibly knowledge about the model. It is an engineering challenge to create such a simulation, a ''machine'', that behaves somewhat analogous to the reality. How this is done? We have the advantage to have a few fundamental equations. We can write down those—fine. But when we want to implement these equations into a computer model, we have to decide which length scale we want to resolve. We have to set the resolution somehow and this leads to problems. When I look at the Navier-Stokes equations (the fundamental equations of hydrodynamics) and go to small length scales, I will get turbulence. But I cannot describe this turbulence explicitly in my GCM. Without turbulence, however, I will get wrong results. So the solution is to modify the equations. I do this by choosing which length scales I still resolve and which length scales I do not resolve any more. Then I have to investigate what effect the unresolved length scales will have on the resolved, the large length scales. I have to describe turbulence somehow implicitly, in order to get correct results for the large length scales, I am interested in. This technique is called parametrization and it has much of an art.
Was this article helpful?