Partial Radiative Perturbation Cloud Radiative Forcing Method

1. Partial Radiative Perturbation Cloud Radiative Forcing Method Advantages: •Explicitly measures differential behavior of radiative fluxes in response to imposed climate change scenarios •Isolates each feedback effect Disadvantages: •Computationally expensive •Results cannot be compared with observations •Uses SST perturbations to induce TOA fluxes, and clear-sky and total-sky fluxes are used to determine the sensitivity Advantages: •Straightforward to implement •Requires little computational overhead •Definition is consistent with observable quantities Disadvantage: •Does not account for potential differences in T and H2O vapor distributions between clear and cloudy atmospheres

2. Model and Experimental Design •Geophysical Fluid Dynamics Laboratory (GFDL) Atmospheric Model 2 (AM2) •Several different model versions are used in this study (such that models with different climate sensitivities can be compared) •To induce climate changes, 2K SST perturbations are used Given radiative forcing G, climate system restores equilibrium by inducing a change in surface temperature: G = [(F-Q)] (F=OLR, Q=absorbed SW radiation, both at TOA) Climate Sensitivity, Λ = TS / [(F-Q)] Cloud Radiative Forcing, CRF: CRF = (Fclear – F ) – (Qclear – Q )

3. Partial Radiative Perturbations To compute cloud feedback, input values for profiles of temperature, water vapor, and surface albedo are used from the –2K simulation, while cloud amount and cloud water paths are taken from the +2K simulation: R=F-Q; T, C, r: profiles of temperature, clouds and H2O vapor S: surface albedo Also, feedback parameter for each variable can be written: ΔTs=-G/λ, Where λ=λT+λC+λr+λαs

4. How is this different from the CRF approach? The prime quantities represent the altered properties in the perturbed climate, i.e. T’=T+ΔT

5. Important Radiative Feedbacks Total Sky Feedback – Clear Sky Feedback

6. Let’s now assess the change in net cloud forcing (due to cloud and non-cloud feedbacks) Now, assume a scenario with no cloud feedback (λC=0) This represents the effects of noncloud feedbacks, and is inherently incorporated into calculations of cloud feedbacks using the ‘CRF’ method

7. And, if we further partition λ*CRF into LW and SW components, then: λ*CRF-LW = (λT-λT(CLR))+(λr-λr(CLR)) = -0.24 Wm-2K-1 λ*CRF-SW = (λαS-λαS(CLR)) = -0.05 Wm-2K-1 Thus, we could expect the CRF method to yield LW and SW cloud feedbacks that are about 0.24 Wm-2K-1 and 0.05 Wm-2K-1 smaller respectively than those calculated by the PRP method …

8. We see the right trend at least for the LW feedback (as the CRF LW feedback are all smaller than the PRP LW feedback), but not for the SW feedback The net cloud feedback somehow seems reasonable, as the CRF method yields cloud feedbacks that are smaller by ~0.3-0.4Wm-2K-1 than the PRP method

9. Final Thoughts ●Though there is a possible problem with the SW figure, the main point of the paper was to highlight subtle, yet important differences between the PRP and CRF methods for calculating cloud feedbacks ●Reductions in cloud forcing can be associated with a positive cloud feedback (as noncloud effects are potentially very important) ●It is argued that most of the models used in the Cess. et al. (1996) would actually have positive cloud feedback if the PRP method were utilized (even though ΔCRF<0 in nearly half of the models)