1. © Crown copyright Met Office Towards understanding the mechanisms responsible for different cloud-climate responses in GCMs. Mark Webb, Adrian Lock (Met Office), Tomoo Ogura (NIES) 4th PAN-GCSS meeting: Advances in modelling and observing clouds and convection. Toulouse, June 2008

2. © Crown copyright Met Office Acknowledgements Sandrine Bony, Chris Bretherton, Gill Martin, William Ingram, Bjorn Stevens, Joao Teixeira, George Tselioudis, Mark Ringer, Keith Williams, Rob Wood

3. © Crown copyright Met Office CFMIP-GCSS activities for better understanding of cloud-climate feedback processes Process studies based on the analysis of high-frequency climate model output e.g. GPCI, ARM sites Sensitivity tests (in GCMs and SCMs) Idealised low cloud feedback studies with SCM/CRM/LES

4. © Crown copyright Met Office Standard deviation of cloud radiative responses to CO 2 doubling (W/m-2)

5. © Crown copyright Met Office SW cloud radiative response Wm -2 TWP GPCI SE Pacific Wm -2 Triangles show ensemble standard deviations of local values Squares show correlation of local and global values across ensemble 1

6. © Crown copyright Met Office LW cloud radiative response Wm -2 TWP GPCI SE Pacific Wm -2 Triangles show ensemble standard deviations of local values Squares show correlation of local and global values across ensemble 1

7. © Crown copyright Met Office HadGEM3 development version containing: PC2 cloud scheme with prognostic cloud fraction, liquid and ice Lock boundary layer scheme Standard experiments (STD) a/ 10 year atmos-only forced with SST/sea ice obs climatology b/ CO 2 forcing: fixed SST + 2xCO 2 (Hansen method) c/ Climate change: control + CMIP 1% patterned SST composite d/ Climate change: control + 2K uniform SST CFMIP-2 Pilot experiments

8. © Crown copyright Met Office

9. Control 2CO 2 Patterned SST Uniform SST GPCI

10. © Crown copyright Met Office Control 2CO 2 Patterned SST Uniform SST SETP

11. © Crown copyright Met Office Lock scheme uses following BL types: 1. Stable 2. Sc over stable 3. Well mixed 4. Decoupled Sc not over Cu decreasing cloud fraction/water 5. Decoupled Sc over Cu higher cloud top 6. Cumulus capped positive SW response? If so we would expect to see consistent changes in frequency of BL types and the SW CRF response Could the SW responses be mainly due to transitions between boundary layer types?

12. © Crown copyright Met Office Control 2CO 2 Patterned SST Uniform SST SETP

13. © Crown copyright Met Office Control 2CO 2 Patterned SST Uniform SST SETP

14. © Crown copyright Met Office This idea is not supported for 2CO 2, but is for the SST forcings: 3. Well mixed Sc 4. Decoupled Sc not over Cu 5. Decoupled Sc over Cu 6. Cumulus capped SW CRF response Pat +2K +ve +ve -ve Transitions between BL types are larger in the SST forced runs, so these may be dominating smaller changes in the SW caused by cloud changes within the BL regimes Are the SW responses mainly due to transitions between BL types?

15. © Crown copyright Met Office In PC2 the largest source terms for clouds in this region are: - condensation due to LW cooling (mainly at cloud top) - condensate detrained from shallow convection We can repeat the experiments removing these terms: 1/ liquid condensate formed through LW cooling falls out (No CTC) 2/ liquid condensate from convective detrainment falls out (No CD) We might expect LW cooling to be more important nearer the coast, and convective detrainment to be more important in the centre of the section How can we trace these responses back to modelling assumptions?

16. © Crown copyright Met Office Control 2CO 2 Patterned SST Uniform SST

17. © Crown copyright Met Office Control 2CO 2 Patterned SST Uniform SST

18. © Crown copyright Met Office Control 2CO 2 Patterned SST Uniform SST

19. © Crown copyright Met Office Control 2CO 2 Patterned SST Uniform SST

20. © Crown copyright Met Office In HadGEM3/PC2, SETP SW responses to SST forcing are largely consistent with transitions between BL types Responses near the coast coincide with transitions between non-convective BL types, and require condensation from LW cooling but not convective detrainment. Responses further west coincide mainly with transitions between decoupled Sc over Cu and trade Cu, in opposite directions for uniform and patterned SST perturbations. These require convective detrainment, but are weaker without LW cooling Summary

21. © Crown copyright Met Office Caldwell and Bretherton (submitted) argue that reductions in LW cooling due to increases in CO 2 or water vapour can lead to reduced cloud top entrainment and a shallower BL We can change or remove the dependence of entrainment on LW cooling to see if this mechanism explains the reductions in cloud near BL top in our results Why does shallow convective detrainment reduce? Does reduced LW cooling of the BL force a reduction in latent heat release from shallow convection? We could test this by discarding the latent heating from shallow convection. Future work

22. © Crown copyright Met Office The South East Tropical Pacific is a convenient area to study shallow cloud feedbacks in GCMs Sensitivity tests in CFMIP-2 could help us to understand cloud response mechanisms at play in other GCMs We will be able to do many more such sensitivity tests if we can develop relevant SCM forcing cases Capturing the important shallow cloud feedbacks across models may well require cases for well mixed, convective and transition boundary layers Concluding remarks