Quoted from: https://doi.org/10.1175/1520-0442(2004)017%3C3666:TCCSMV%3E2.0.CO;2 

CAM2.0 includes a number of improvements to the physical parameterizations. The diagnostic cloud water scheme of CCM3 has been replaced with a prognostic scheme for total cloud condensate and is described in Rasch and Kristjansson (1998). Total water is predicted for a model grid box, and based on the grid box temperature, this condensate is partitioned between liquid and ice. Cloud fraction is still diagnosed similarly to CCM3. Given the increased vertical level structure, a new cloud overlap methodology was developed by Collins (2001). This approach provides a generalized scheme to treat the overlap of cloud layers. The CAM2.0 configuration assumes that clouds are maximally overlapped when in adjacent layers and randomly overlapped when there is a gap in between cloud layers.

CAM2.0 also employs the updated cloud water vapor emissivity/absorptivity scheme by Collins et al. (2002). This new treatment of the longwave properties of water vapor includes version 2.1 of the Clough–Kneizys–Davies (CKD) continuum developed by Clough et al. (1989). This improvement brings the modeled longwave fluxes and cooling rates into better agreement with detailed line-by-line calculations. It has a significant impact in enhancing longwave cooling in the upper troposphere, which in turn affects the behavior of the convective parameterization. Briefly, the change in convective activity led to a significant drying of the atmosphere, especially in the Tropics. To alleviate this degradation in the modeled hydrologic cycle, a change was made to include the evaporation of convective precipitation back to the atmosphere, that is, not all convective precipitation reaches the surface. A profile of the precipitate is produced by the moist convection process; the evaporation of this precipitate back into the atmosphere is directly proportional to the large-scale relative humidity in a given model layer. The change in the longwave scheme also addressed a longstanding bias in the polar regions where the original CCM3 clear-sky downward flux was too low. The CAM2.0 polar clear-sky longwave surface fluxes are now in very good agreement with observations.

Further enhancements to the uncoupled CAM2.0 include the inclusion of a new ozone dataset documented in Kiehl et al. (1999); the use of the thermodynamic component of the CCSM sea ice model, version 4 (CSIM4); the use of the new Community Land Model, version 2.0 (CLM2.0); and the implementation of realistic fractional land, ocean, and sea ice for grid boxes. The model also includes the capability of employing a reduced grid at high latitudes for computational efficiency.

Improvements in the climate simulation of the uncoupled CAM2.0 model compared to the climatology simulated by CCM3 include a more realistic distribution and amount of precipitable water, an improved clear longwave flux simulation compared to observations, an improved shortwave cloud forcing in the regions of cold sea surface temperatures (SSTs) at eastern ocean boundaries, and a reduction in central American convective activity with an associated improvement in surface wind stress in the subtropics. There are a number of areas where the CAM2.0 simulation has degraded the simulation compared to that of CCM3. These include a colder tropical tropopause, a significant warm bias over land at high latitudes, and a tendency to produce a double intertropical convergence zone (ITCZ) structure in the uncoupled mode, that is, with prescribed sea surface temperatures. These are significant biases in the model, and they increase in magnitude when CAM2.0 is coupled to the ocean, land, and sea ice components.