If you are incorporating a new process, you first must be able to reproduce the current climate. I have noticed that a lot of people think that we inject data into a model. We don't. The starting point is the governing equations: Equations that govern the motion of the fluids; equations for the exchange of momentum, heat, fresh water, between various components of the climate system; equations that govern the uptake and release of various gases; equations that govern radiation, formation of clouds, evaporation etc. So you have all these governing equations that you have to link together under conservation of various properties, such as the energy. The next step is to drive all these equations by energy from the sun. So you turn on the sun and then you have to get a climate that looks like a real climate. And then you also have to observe its seasonal variation and its internal variability.
If you can reproduce the current state of the climate, you have to go back and look at the temporal evolution of the climate. So you are going to start at, say, 1850 and run a transient simulation through to where we are today. When you get comfortable with the comparison between the transient evolution of the model and the 20th century climate, you can look at small perturbations around the present climate. You go back to some paleo-examples; maybe the last glacial maximum. And now you observe what happens to the model if you change the way the earth is tilted etc. Perhaps you change the level of CO2 and you look what happens then and you compare with some paleo proxy records.
I like to think of it as a three-step evaluation process: one is the contemporary climate; two is the transient climate; three is paleoclimate. And only then, if you feel comfortable with reproducing what we know, can you start undertaking projections with the model.
Was this article helpful?