Editing Principles of Climate Modeling (section)

=== The Challenge of Clouds ===
Cloud formation is the most prominent and notoriously difficult process to parameterize. A model cannot simulate an individual cumulus cloud. Instead, the model looks at the average temperature and humidity of a 100-kilometer grid box. If the humidity reaches a certain threshold, the parameterization algorithm assumes that a certain percentage of that grid box is covered by clouds, and it mathematically alters the albedo and rainfall of the box accordingly.

Because clouds can both cool the planet (by reflecting shortwave solar radiation) and warm the planet (by trapping longwave thermal radiation), slight differences in how different modeling teams parameterize cloud physics remain the largest source of uncertainty in projecting exactly how sensitive the Earth's climate is to doubling CO2 concentrations.