Modeling climate means taking a lot of of factors, studying their individual impacts, their interactive impacts, with ultimately developing some kind of an equation that has the ability not only to predict what will happen, but what has happened. So you need data and that includes historical data. The thing with historical data is that the further back you go, the fuzzier it gets. Record keeping wasn't anywhere as encompassing as it is today. Some information needs to be inferred and back calculated, adding uncertainty which needs to become part of the model.
Then a model is released, refined but then we find it's unable to explain this pause we seem to be having with global temperatures having stagnated. For how long, who knows? Will it pick up with more ferocity or diminish, who knows? But we do know the model needs to be revised. Somehow something(s) were missed. It may be a term, a factor that may not act independently but something that onteracts with the other factors.
A concern I have is that there can be an inherent danger in forcing the equation to fit the observation. Let me illustrate by example. Back in the day I ran a chemical analytical lab supervising chemists and techs. Among other things we would be required to determine the concentrations of elements and metals. upper management became sold on a software package that would be capable of data analysis, report generation, and all the good things that ran on 386 and 486 computers.
One chemist, Wayne, had been charged with determining the concentrations of various solvents by gas chromatography. So, freed from having to use a scientific calculator, Wayne ran his calibration curves, ran his sample, and used the new software.
I looked over all reports, examined the data, and questioned when appropriate. So Wayne comes back with the results and I asked what the error was? None was the reply. Ummmmm...really? How was the calibration curve? Perfect, I used the new software program. Keep in mind, Wayne is a degreed chemist. What he'd done is taken his calibration data, amount vs. response, dumped it into the program and selected a best fit which spit out a 5th order polynomial. The thing is, the detector, within its limits, is a first order response. Wayne just wasn't thinking and his data, which wasn't that good looking for a first order fit, suggested there was a problem somewhere. I forget if it was standard preparation or something else but after addressing the usual suspects he then had a decent first order response that could be mathematically fitted. Not perfect but entirely reasonable.
I recount the above because IMO, this climate change mathematical modeling still has a way to go. This doesn't mean I don't believe there are issues going around climate wise. Sea levels have been creeping up and some islands are going to disappear, coast lines are going to be reshaped and so forth. Carbon credits, taxes or whatever punitive measures can be thought of don't strike me as smart or effective. Planning does. Changing where we grow certain crops sounds reasonable. Better mpg and other efficiencies sounds like a good idea. Growing government maybe not so much. Politicians by and large are whores looking to divide, segment, and exploit what's trendy.