I haven't run, or even seen, such a model, but I presume they would do both real world or empirical data, at different times, depending on what they're trying to analyze.
Pretty clearly, models are used for more than just to analyze the next 5-10-15-20 years.
They would be verified using past data, to determine whether they would produce the climate results that we've already seen occur.
Then they'd be used with more current data, to look at the near future.
And other times, they'd be used with hypothetical data, to see if there are tipping points beyond which recovery is not possible.
Yes they do. Ideally the model would be run multiple times with slightly different initial conditions as a way of determining the sensitivity to the initial conditions. Averaging multiple runs would remove some of that sensitivity. What they would use as initial conditions would vary depending on the model, but typically they would use reanalysis, which is numerical weather prediction model output run with all available data. That essentially serves as a "best guess" at the state of the atmosphere at a particular moment and is available on a global grid. There are several different reanalyses available but typically the ones used are from ECMWF or NCEP.
Yes it matters. The initial values determine the output of the model, especially the parameters of various physical processes.
I would have thought so, though wouldn't you choose actual values for the start of a model run?
My understanding (from my participation here) is that real world data is used to calibrate the models. Empirical data is real world data; isn't it?
From what I have read, I don't think it matters a great deal. All that really matters is that warming spits out the other end. If it doesn't, the modeler takes out his wrench and hammer and monkies with it until it does.
Only Jimmy Hansen would know for sure and he is off doing his communist activism thing.
No because it's all B.s anyhow.
