Modern climate models are highly complex and run on supercomputers. But our trust in climate predictions comes not from the complexity and expensiveness of these models, it comes from our ability to conceptualize and understand their behavior (and that of the real world) in fundamental physical terms.
How does the radiative energy balance change as the temperature changes? How do properties of clouds (which also strongly influence the energy balance) depend on the atmospheric state? How do small-scale processes such as in the turbulent boundary layer depend on the large-scale state and feed back on it?
We study these physical aspects of climate and in doing so contribute to better understand and predict it.