Climate modelers are the first to point out that the models are imperfect, so it's reasonable to ask what the uncertainties are and whether, given those, we can trust models to provide useful predictions. Their main shortcomings are:

• Limited observations
• Incomplete understanding of climate processes and their interrelationships
• Imperfect model representations of the processes
• Relatively large grid boxes that don't represent smaller terrain features

To determine whether climate models are useful and where problems might exist, scientists run tests to see how well the results match actual past temperature changes.

If the two don't match, then scientists have to figure out if there's a problem with the observations or with some component of the models. For example, a few years ago some of the satellite data did not show parts of the atmosphere warming at the same rate as models did. Eventually an error in the satellite data calculations was found, and now the two are in agreement.