1. new CRM diagnostics dedicated to convective parameterization J-I Yano P. Bechtold, J.-P. Chaboureau, F. Guichard, J.-L. Redelsperger and J.-P. Lafore LMD, CNRM-GAME & Laboratoire d’Aérologie, Toulouse restitution of convective fluxes Yano et al. to appear potential energy convertibility (PEC)

2. CRMs-SCMs comparisons A lot of diagnostics have been produced domain average quantities mean profiles of T, q, Q1, Q2, cloud water convective fluxes, but not so much infact sub-domain quantities, more delicate often need to define criteria in CRMs (as in observations) in-cloud properties convective mass flux scheme

3. CRM simulations MesoNH, 3D,  x=2km comeNH, 2D and 3D,  x=500m to 2km (nz ~ 50) COARE cases : tropical convection over ocean ARM case : and midlatitude convection over land

4. « segmentally constant » classification knowing the mean properties of the different categories (up and down convective, up and down stratiform…) is not enough to recover the total convective fluxes (more complex than shallow convection) not negligeable

5. new model, new simulations, same conclusion (only the resolved part of the convective flux above) analysis in this « real » space: in terms of mass-flux based parameterizations, the CRM- derived estimates of mean properties within each category alone are not useful to evaluate parameterized convective fluxes

6. w’ < 0  ’ > 0 w’ < 0  ’ < 0 w’ > 0  ’ > 0 w’ > 0  ’ < 0 in the probability space very wide range of variability in  over both half-planes (w’> 0 & w’<0) see later (from snapshot at z~5km)

7. same but for the convective area The spread is not decreased much

8. total flux improvement with 4 quadrants but still too low

9. higher correlations for higher values (4 corners)

10. the effective values contributing to fluxes are much higher than simple means use weighted average rather than simple average We tried: with parameter « a », a = 0.25 hereafer

11. 2 half-planes 4 quadrants weighted 4 quadrants

12. rms error for the 3 methods + standard deviation (solid) same results for momentum flux latent heat flux sensible heat flux

13. the decomposition in the (w’,  ’) probability space does not guaranty that the mass fluxes are the same for all the variables ( , q, u, v, tracers) in terms of parameterization:  i w i for   i w i for q

14. combination of the 4 components into 2  i w i for  i w i for q positive drafts negative drafts

15. summary, discussion, in the context of today (GCSS) one-to-one comparison between CRMs and SCMs not always straightforward, e.g. convective fluxes the method proposed here (with weighted averages) enables to advance on this issue, by shifting to the probability space behind is the idea that the mass flux formulation stands more as an idealized framework rather than directly as a simplified picture of the reality (this view is coherent with recent LES analyses, e.g., showing the weaknesses of this formulation to retrieve the variances) … to be further explored: the meaning of entrainment/detrainment in this modified context

16. a generalization of CAPE with the potential energy convertibility (PEC) motivation: CAPE is a very useful convective parameter but CAPE has its own limitations in the context of parameterization e.g., extension of CAPE for entraining plumes CAPE variations not simply linked to convection e.g., diurnal cycle of deep convection over land, dry intrusion periods over the warm pool

17. PEC ~ rate of conversion of energy normalized by a measure of the vertical momentum

20. PEC far from perfect but provides a better measure of moist convective instability than CAPE from the case- studies Colors: 3 distinct COARE periods, from CRM runs

21. possible strategy: to modify entrainment rates so that they produce a buoyancy consistent with b* obtained from PEC work in progress