"Scientists that evaluate climate models, develop physical process parameterizations, and utilize climate model results are convinced (at least to some degree) of the usefulness of climate models for their research. They are convinced because of the model’s relation to theory and physical understanding of the processes involved, consistency of the simulated responses among different models and different model versions, and the ability of the model and model components to simulate historical observations."
Anyone can create a model that behaves consistently when it is trained on historical data to predict historical data.
How do they overcome the problem of overfitting and why should anyone have any confidence in it until it shows accuracy going forward?
I think the key is in the last quoted sentence: there isn't one model. There are many, and they are generated in different ways with differing emphasis on different data sets. They gain confidence when these seem to agree with each other and the acceptable physical knowledge already understood. They also check the sanity of the model to make sure that it is consistent with what has already been observed. In that way anyone may gain confidence that multiple approaches are converging on similar results.
That, and there are metric crap-tonnes of data to work with of all sorts. Just ungodly amounts of it.
Anyone can create a model that behaves consistently when it is trained on historical data to predict historical data.
How do they overcome the problem of overfitting and why should anyone have any confidence in it until it shows accuracy going forward?