Comparisons of model predictions with observations identify what is well-understood and, at the same time, reveal uncertainties or gaps in our understanding. This helps to set priorities for new research. Vigilant monitoring of the entire climate system—the atmosphere, oceans, land, and ice—is therefore critical, as the climate system may be full of surprises. Together, field and laboratory data and theoretical understanding are used to advance models of Earth’s climate system and to improve representation of key processes in them, especially those associated with clouds, aerosols, and transport of heat into the oceans. This is critical for accurately simulating climate change and associated changes in severe weather, especially at the regional and local scales important for policy decisions. Simulating how clouds will change with warming and in turn may themselves affect warming, remains one of the major challenges for global climate models, in part because many cloud processes occur on scales smaller than the current models can resolve. Greater computer power may enable some of these processes to be resolved in future-generation models. Dozens of groups and research institutions work on climate models, and scientists are now able to analyse results from essentially all of the world’s major Earth-System Models and compare them with each other and with observations. Such opportunities are of tremendous benefit in bringing out the strengths and weaknesses of various models and diagnosing the causes of differences among models, so that research can focus on the relevant processes. The differences among models allow estimates to be made of the uncertainties in projections of future climate change, and in understanding which aspects of these projections are robust. Studying how climate responded to major changes in the past is another way of checking that we understand how different processes work and that models are capable of performing under a wide range of conditions.