In fact, that's not the case. The basics of climate change are very straightforward - CO2 increases the warming of the atmosphere, which has been known since 1896, and that in turn will increase the amount of water vapour, which will also increase the temperature of the atmosphere. The effect of those feedbacks on other feedbacks are all known through perfectly standard scientific investigation.
What you're actually referring to are the predictions of specific effect, such as "temperature will increase by such and such an amount for such and such an addition of CO2". Even those, however, are unlike economic models, because economic models deal with intelligent actors who can (and do) apply the knowledge of the models to the system being modelled, and whose reaction to a change in economic circumstances is inherently unpredictable as a result. CO2 doesn't seek arbitrage opportunities, and provides, for a certain concentration, a known and invariable amount of forcing.
Such models are routinely used in virtually every branch of science, from chemistry through astrophysics to population biology. It's always possible to say that the state of knowledge is insufficient to permit modelling to be worthwhile, but the proof of the pudding would seem to be in whether the models can reproduce observations:
Given that the models do reproduce observations rather well - as they do - it seems to me that you need to fall back on the view that either the models or the data are 'fixed' to allow the reproduction of observations, or that it's essentially the result of lucky guesswork. The latter we should rationally discard on the basis that you're talking about literally tens of thousands of runs of different models - the former requires an enormous global conspiracy of silence. Which are you going for?