Quoted from: https://ncar.github.io/CAM/doc/build/html/cam5_scientific_guide/introduction.html#cam3 

The CAM3 was the fifth generation of the NCAR atmospheric GCM. The name of the model series was changed from Community Climate Model to Community Atmosphere Model to reflect the role of CAM3 in the fully coupled climate system. In contrast to previous generations of the atmospheric model, CAM3 had been designed through a collaborative process with users and developers in the Atmospheric Model Working Group (AMWG). The AMWG includes scientists from NCAR, the university community, and government laboratories. For CAM3, the consensus of the AMWG was to retain the spectral Eulerian dynamical core for the first official release although the code includes the option to run with semi-Lagrange dynamics or with finite-volume dynamics (FV). The addition of FV was a major extension to the model provided through a collaboration between NCAR and NASA Goddard’s Data Assimilation Office (DAO). The major changes in the physics included treatment of cloud condensed water using a prognostic formulation with a bulk microphysical component following [RKristjansson98] and a macroscale component following [ZLB+03]. The [ZM95] parameterization for deep convection was retained from CCM3.

A new treatment of geometrical cloud overlap in the radiation calculations computed the shortwave and longwave fluxes and heating rates for random overlap, maximum overlap, or an arbitrary combination of maximum and random overlap. The calculation was completely separated from the radiative parameterizations. The introduction of the generalized overlap assumptions permitted more realistic treatments of cloud-radiative interactions. The methodology was designed and validated against calculations based upon the independent column approximation (ICA). A new parameterization for the longwave absorptivity and emissivity of water vapor preserved the formulation of the radiative transfer equations using the absorptivity/emissivity method. The components of the method related to water vapor were replaced with new terms calculated with the General Line-by-line Atmospheric Transmittance and Radiance Model (GENLN3). The mean absolute errors in the surface and top-of-atmosphere clear-sky longwave fluxes for standard atmospheres were reduced to less than 1 W/m. The near-infrared absorption by water vapor was also updated to a parameterization based upon the HITRAN2k line database [RBB+03] that incorporated the CKD 2.4 prescription for the continuum. The magnitude of errors in flux divergences and heating rates relative to modern LBL calculations were reduced by approximately seven times compared to the previous CCM3 parameterization. The uniform background aerosol was replaced with a present-day climatology of sulfate, sea-salt, carbonaceous, and soil-dust aerosols. The climatology was obtained from a chemical transport model forced with meteorological analysis and constrained by assimilation of satellite aerosol retrievals. These aerosols affect the shortwave energy budget of the atmosphere. CAM3 also included a mechanism for treating the shortwave and longwave effects of volcanic aerosols. Evaporation of convective precipitation following [Sun88] was implemented and enhancement of atmospheric moisture through this mechanism was offset by drying introduced by changes in the longwave absorptivity and emissivity. A careful formulation of vertical diffusion of dry static energy was also implemented.

Additional capabilities included a new thermodynamic package for sea ice in order to mimic the major non-dynamical aspects of CSIM; including snow depth, brine pockets, internal shortwave radiative transfer, surface albedo, ice-atmosphere drag, and surface exchange fluxes. CAM3 also allowed for an explicit representation of fractional land and sea-ice coverage that gave a much more accurate representation of flux exchanges from coastal boundaries, island regions, and ice edges. This fractional specification provided a mechanism to account for flux differences due to sub-grid inhomogeneity of surface types. A new, extensible climatological and time-mean sea-surface temperature boundary data was made available from a blended product using the global HadISST OI dataset prior to 1981 and the Smith/Reynolds EOF dataset post-1981. Coupling was upgraded in order to couple the dynamical core with the parameterization suite in a purely time split or process split manner. The distinction is that in the process split approximation the physics and dynamics are both calculated from the same past state, while in the time split approximations the dynamics and physics are calculated sequentially, each based on the state produced by the other.