Response to Reviewer B Thank you for your helpful review. In response to your suggestions and those of the other reviewer, we have split the paper into two, closely related parts. The first (referred to as Part I below) contains the analysis of SP-CAM cloud changes and has undergone minor revisions relative to the same content in the original submission. The second (Part II) describes the column analogue and sensitivity of the results of the column analogue to different assumptions, including: (i) resolution (expanded since original submission), (ii) the effect of diurnally-varying versus diurnally-averaged insolation, (iii) the use of differing assumptions about the momentum damping timescale (vertically-uniform versus vertically-varying), and (iv) the use of a tropics-mean temperature sounding as the reference state in the column analogue. The material in Part II has expanded since the original submission, so that we could more fully explain the column analogue and address some of the issues raised in the reviews. Despite this, we have still left some questions to the future, including the application of this framework to other cloud regimes. Overview: Since we have split the manuscript into two parts, the reviewer's main point has been addressed, as each part is now more digestible. In light of the split we are able to retain material that the reviewer suggests moving to a supplement. Detailed Comments: We have added 25th/75th percentile boundaries to plots of cloud fraction and net cloud forcing in figure 4 (Part I) for SP-CAM, to give some sense of the substantial variability within bins as well as some brief discussion about the issue (Part I). We have included a comparison with ISCCP simulator data as a new panel in Figure 3 (Part I). This shows the discrepancy between model and observation is somewhat reduced. In Figure 4e (Part I), the strong artifact near 100hPa is due to tropopause deepening in the warmer climate. A mention of this has been added to the text. The only grid mismatch is in the upper part of the stratosphere. Elsewhere the grids are co-located. The text describing the grids has been revised slightly. The font sizes were enlarged in Figure 4 and Figure 9 (Part I). The short discussion on specific humidity versus relative humidity on p. 21 is an important one because it helps show a key aspect of the physical mechanism, so we would like to retain it (Part I). The suggestions for moving section 6 and parts of section 7 (material now in Part II) have been addressed by the new paper configuration. The page 34 typo (now in Appendix A. of Part II) has been corrected. The wording regarding the time-averaging period has been clarified (now in Part II) It is possible that the low cloud feedback mechanism could be found in nature. Since we are limited in this study to investigating this effect using model output, we'd rather not assert that the mechanism is at work in nature without some observational evidence to back it up. We did add to the discussion a mention of preliminary looks we made at other GCM's low cloud changes and associated changes in radiative cooling. The default setup of the CRM and LES runs in the revised paper includes a diurnal cycle of insolation. The effect of including the diurnal cycle is discussed in section 5 of Part II. In the runs in the revised Part II, we have not included transient forcings. We have performed some initial experiments with time-varying modulation of omega and the large-scale horizontal advection. However, we would like to be careful that the transients are faithful to those in the parent model, so that we will pursue that in future work. In the revised part II, we have included a sensitivity study in which a tropical mean temperature sounding is used as the reference state. This extension of the framework could be amenable for such initiatives as the cloud feedbacks model intercomparision project (CFMIP). The effects of coarse resolution on the simulation of these clouds is characterized more fully in the revised part II, with an expanded resolution sensitivity study. In addition, we do suggest a physical explanation for the over-prediction of cloud fraction in SP-CAM. Our results do suggest that SP-CAM does over-predict cloud fraction and cloud forcing in these LTS bins. However, we cannot assert this with certainty, as the modification of the diabatic forcing in these regions could affect the large-scale flow and subsequently drive cloud changes in these bins. The cloud changes in higher resolution simulations occur mainly just below the inversion. This suggests that the strengthening of the inversion in the +2K runs may play a larger role in the cloud changes than the radiative cooling mechanism. We did not try an 8x8 configuration of the CRM. However, we did perform short 3D LES simulations to gain some understanding of the role of dimensionality in our results. We do intend to apply the column framework to other boundary layer cloud deciles, including both stratocumulus and deep convection bins. The consistency of the results across the three finest grids in the resolution study in Part II suggests that our resolution is sufficient for this trade cumulus cloud regime and that, indeed, the cloud fraction decreases with finer resolutions. We would speculate that finer resolution would lead to larger cloud fractions for stratocumulus clouds. In Stevens, Ackerman & Bretherton (2003), simulations of an ATEX-like case had increasing stratocumulus cloud fractions at the inversion with increasingly finer resolutions. Note that their cumulus-under-stratocumulus case differs from ours in that their subsidence was much stronger, and their simulations had no insolation. Also, Stevens et al (2005, the DYCOMS RF01 LES intercomparison paper) showed increasing LWP with decreasing Dz down to 1m. The combination of our results with these papers would suggest that SP-CAM is likely to over-estimate the trade cumulus cloud fraction and under-estimate the stratocumulus cloud fraction due to under-resolution. We do intend to study SP-CAM's representation of stratocumulus clouds using the column modeling framework in the future. (Note that much of this discussion is included in the conclusions of the revised Part II.) We feel that this framework will be useful for exploring the cloud response in other GCMs. However, this is work that we hope to pursue in the future. As mentioned in Part II, our results suggest that the SP-CAM boundary layer cloud response in these LTS bins is too strong and that a finer resolution embedded CRM might produce both less cloud and a more weakly negative change in SWCF.