Climate feedbacks represent mechanisms that alter the sensitivity of the earth climate system. It has been suggested that the current spread in climate model sensitivity to a CO₂ forcing is a result of different treatments of climate feedbacks. The determination of the climate system sensitivity is critical to understanding how the system will respond to a CO₂ radiative forcing. The strength of a climate feedback is defined in terms of annual, global mean top of atmosphere (TOA) radiative perturbation. However, contributions to the global, annual mean feedbacks can originate from different geographical regions and vertical layers within the atmosphere. In addition, the contributions to the annual mean TOA radiative perturbation can be disproportionately distributed throughout the annual cycle. This study performs offline, partial radiative perturbation-style, radiative calculations to determine the geographical, vertical, and seasonal distributions of the major climate feedbacks contributing to the TOA radiative energy budget: clouds, water vapor, temperature, and surface albedo. These feedback strengths are diagnosed from NCAR CCSM3.0 model output for the SRESA1B emission scenario simulated for the IPCC AR4. It is found that the tropics and sub-tropical climate responses drive the sign and strength of the water vapor and cloud feedbacks. In addition, a significant annual cycle of the SW cloud and surface albedo feedbacks is found. The inter-seasonal variations of the SW cloud and surface albedo feedbacks found here show a different pattern than previously published results. The radiative perturbations are then used as input into the newly developed Coupled Feedback Response Analysis Method (CFRAM), which uses a total energy based method to isolate partial temperature changes due to individual feedbacks in the atmosphere and at the surface.

Many authors have calculated climate feedback radiative perturbations in different manners using seasonal mean, monthly mean, daily mean, and every time step model output. Monthly mean model output is used in this study. A comparison of the global mean clear sky TOA net flux calculation using monthly mean model output with the monthly mean model output TOA net flux reveals a global mean bias in the offline radiation calculations compared to the model simulated TOA net flux of +3.95 Wm^-2 with a standard deviation of 3.78 Wm^-2. In order to handle complexities associated with cloud overlap, the Monte Carlo Independent Column Approximation (MCICA) technique is uniquely adapted for use in the context of this study. This technique relies on a stochastic cloud generator using a maximum-random overlap rule to sample the monthly mean cloud frequency profile. It is shown that the global mean bias in the calculation of the TOA net flux compared to NCAR CCSM3.0 model output is +1.74 Wm^-2 and a standard deviation of 6.71 Wm ^-2 using this technique. However, the results suggest that the technique provides a very good estimate of all feedback sensitivity parameters despite bias associated with using monthly mean model output.