The combined CloudSat and CALIPSO satellite observations provide the first simultaneous measurements of cloud and precipitation vertical structure and are used to examine the representation of tropical clouds and precipitation in the Community Atmosphere Model version 3 (CAM3). A simulator package utilizing a model‐to‐satellite approach facilitates comparison of model simulations to observations, and a revised clustering method is used to sort the subgrid‐scale patterns of clouds and precipitation into principal cloud regimes. Results from weather forecasts performed with CAM3 suggest that the model underestimates the horizontal extent of low‐level and midlevel clouds in subsidence regions but overestimates that of high clouds in ascending regions. CAM3 strongly overestimates the frequency of occurrence of the deep convection with heavy precipitation regime but underestimates the horizontal extent of clouds and precipitation at low and middle levels when this regime occurs. This suggests that the model overestimates convective precipitation and underestimates stratiform precipitation consistent with a previous study that used only precipitation observations. Tropical cloud regimes are also evaluated in a different version of the model, CAM3.5, which uses a highly entraining plume in the parameterization of deep convection. While the frequency of occurrence of the deep convection with heavy precipitation regime from CAM3.5 forecasts decreases, the incidence of the low clouds with precipitation and congestus regimes increases. As a result, the parameterization change does not reduce the frequency of precipitating convection, which is far too high relative to observations. For both versions of CAM, clouds and precipitation are overly reflective at the frequency of the CloudSat radar and thin clouds that could be detected by the lidar only are underestimated.