Even when I was a petroleum engineer 30 years ago, we used main frame
computers to model reservoir behaviour. Basically, we would generate a
mathematical model of the reservoir under the ground with known rock
parameters at each well location and see if we could get the computer to
predict historical behaviour. That is, we would withdraw the known
production from each well and see if the computer would accurately
predict the pressure in the reservoir, the elevation of the gas/oil
interface and the water/oil interface. If we could manipulate the
unknown parameters into getting the computer to accurately predict
historical measured pressures and gas/oil and oil/water interface
elevations, then we could use that same model to predict how the
reservoir would behave if we were to produce it in different ways, such
as by implementing a water flood or perhaps a miscible flood. In that
way, we maximize the amount of oil produced from large reservoirs.

Climate scientists do the same thing. If they can get their computer
models to predict historically accurate global temperature increases,
then they can use that model to predict what will happen with the global
climate in the future under different scenarios, such as the world
moving from gasoline powered automobiles to electric automobiles over
the next 40 years, or by changing from coal burning electric generating
stations to nuclear reactors, for example.


you hit the nail on the head regarding computer simulations.

when you create a simulation of a geological formation, you have a lot of infomration about the formation. you also have the benefit of testing the simulation on other formations and you can refine the simulation over time and experience.

electronics engineers use simulations to test circuits. the behavior of the components is mostly well known. as soon as you get into an area where the behavior of a component is not known exactly, the simulation fails.

the Earths climate system is chaotic (this is not just a verbal description, it has a technical mathematical meaning) and the behavior of the componets are not well known. Also they do not have the benefit of testing the simlation model. And yet they want to to use the results of these models as the basis for important economc and social policy decisions.

I would like to create a challange for the climtae modellers. Create a small system that has some complex chaotic behaivor that can be accuratly measured. Lets see if the models can predict the future behior of this small system. Oh say, predict what number a rouleete wheel ball will land on. Think you can do that? Creating a model of the Earth's climate is a similar problem.

Mark