So, think about how you'd develop your own climate model in broad strokes.
You have the core model, more CO2 leads to a larger greenhouse impact, that we can model with a high level of certainty. But, that single variable is pretty small. However, that slight rise impacts other things, albedo due to snow melting, changes in cloud cover (we know cloudy nights have a higher low T than clear nights in all cases), but clouds also reflect sunlight better than ground or ocean. Then there are other factors like permafrost melting and oceanic absorption of heat and CO2 and deforestation and ...
Hmmm, this gets complicated. But one could assemble a model, but how would you test it? It would have to be tested against historical data. We have a good record on CO2 levels over time. Do we have a good record of mean global temperature? Some say not so great. But let's stipulate that it is good enough.
So, we have a chart with rise of CO2 and another with rise of T globally. And we now fit parameters using the above variables to match that rise. Easy enough really. But if I do it, and you do it independently, we derive different equations. They back predict the same thing, but they may predict different slopes going forward.
So, it gets complicated quickly. Are the models right? Is one better than all the others? We can't know for some period of elapsed time, and good measurements.
