| A fundamental test of the representation of clouds in models is evaluating the simulation of present-day climate against available observations. Satellite retrievals of cloud properties provide an attractive baseline for this evaluation because they can provide near global coverage and long records. However, comparisons of modeled and satellite-retrieved cloud properties are difficult because the quantities that can be represented by a model and those that can be observed from space are fundamentally different. Satellite simulators have emerged in recent decades as a means to account for these differences by producing pseudo-retrievals of cloud properties from model diagnosed descriptions of the atmosphere, but these simulators are subject to their own uncertainties as well that have not been well-quantified in the existing literature. In addition to uncertainties regarding the simulation of satellite retrievals themselves, a more fundamental source of uncertainty exists in connecting the different spatial scales between satellite retrievals and large-scale models. Systematic errors arising due to assumptions about the unresolved cloud and precipitation condensate distributions are identified here. Simulated satellite retrievals are shown in this study to be particularly sensitive to the treatment of cloud and precipitation occurrence overlap as well as to unresolved condensate variability. To correct for these errors, an improved treatment of unresolved clouds and precipitation is implemented for use with the simulator framework and is shown to substantially reduce the identified errors. |
