significance of subgrid-scale parametrization for cloud resolving modelling

1. significance of subgrid-scale parametrization for cloud resolving modelling Françoise Guichard (thanks to) F. Couvreux, J.-L. Redelsperger, J.-P. Lafore, M. Tomasini

2. cloud-resolving simulations • which « object » ? (space and time scales) • mature squall line phase, one week of COARE-IFA, day-time convection... • for which purpose? • phenomenology, process study, larger-scale parametrizations • numerics • resolution, domain size, filters, sponge layer, advection schemes... • parametrizations (of subgrid-scale processes) • subgrid-scale fluxes (turbulence), microphysics, radiation • subgrid coupling between microphysics and turbulence, radiation... • boundary conditions • initial conditions, boundary conditions (surface & « large-scale forcing ») different sets of choices (with intricate dependencies) these various ingredients play more or less important roles depending on the object/purpose CRMs well suited to explicitely simulate precipitating deep convection (not developed to simulate boundary layer clouds)

3. focus restricted to broad considerations: turbulence (or subgrid-scale fluxes) subgrid coupling between microphysics and turbulence (what sensitivity studies focussing on resolution tell us) no discussion about: microphysics importance of stratiform cirrus (maintenance/dissipation) significance of evaporation of precipitating hydrometeors radiation cloud-radiation interactions

4. a few words about resolution Bryan et al. (2003) for MCSs a « turbulent » point of view (structure, budget,spectra) buoyancy flux qe (x,z) in simulated squall line Dx = 1 km Dx = 125 m bulk features

5. w spectrum at 5 km AGL : 6.Dx (~filter order here) Bryan et al. (2003) TKE (shaded) & [TKE/( resolvedKE+TKE)] Dx = 1 km 0.1 Dx = 500 m Dx = 250 m not strictly an LES-type simulation (turbulenty speaking;subgrid flux not negligeable, l/D) note: among other things, no consideration of cold nor subgrid microphysics (i.e. simulation probably using fixed thresholds for microphysic processes independently of Dx...) Dx = 125 m

6. formulation of subgrid-scale turbulence often based on Smagorinsky, prognostic TKE introduced in many schemes, but still often length scale l = D (grid), different ways to treat the impact of stability (strong numerical diffusion in some CRMs) subgrid-scale microphysics often ignored [introduction of an artificial cutoff] interactions between subgrid-scale motions and microphysics most often neglected sensitivity to horiz. resolution sensitivity to subgrid scale param. sensitivity to horiz. resolution convective storm Dx = 800 m to 2 km Dx = 800 m to 2 km with no subgrid precip with best subgrid param. (103kg.s-1) surface rainrate Redelsperger and Sommeria (1986)

7. subgrid-scale vertical water flux change sign when subgrid-scale interactions between microphysics and subgrid fluxestaken into account! Redelsperger and Sommeria (1986)

8. specific features associated with boundary layer treatment w(x,y) at 0.6 zi(convective boundary layer, clear sky) 30 km max | w | < 1cm.s-1 Dx = 100 m Dx = 2 km Dx = 10 km 1D turbulence scheme 1D turbulence scheme classical organization (open cells) development of spurious organizations expected behaviour with this resolution adapted from Couvreux (2001)

9. potential temperature Dx = 100 m Dx = 2 km Dx = 10 km 3 2 height (km) 1 10 12 14 16 8 lst (h) Wyngaard (2004) buoyancy flux Couvreux (2001) total flux subgrid flux resolved flux 3.5 2.5 here resolved motions react to (compensate for) too weak subgrid transport, in their own way (cf. organisations) (km) terra incognita 1.5 0.5 -0.5 0 0.5 1. 1.5 (K..m.s-1) intricacy of interactions between resolved & subgrid processes

10. from clear sky to cloudy boundary layers ... Fiedler and Kong (2003) mesoscale modelling , Dx = 2 km cold air outbreak AVHRR image liquid water path (kg.m-2) mesoscale cellular convection solid stratiform cloud deck : change from 1 to 0.3 of a coeff. involved in turbu. diffusitivity

11. ... to daytime precipitating convection Dx = 2 km rainwater mixing ratio (g.kg-1) 5 4 3 height (km ASL) 2 1 5 4 without subgrid condensation 3 height (km ASL) 2 1 5 with mixing length l= D in turbulence scheme [D= (Dx.Dy.Dz)1/3] 4 3 height (km ASL) 2 1 6 9 12 15 18 21 24 3 local solar time (h) highly sensitive, through complex interactions with simulated BL structure

12. summary developement of CRMs began in the 80 ’s ; since then, they have been increasingly used (useful) and validated ; they constitute very valuable tools this occurs ... even though a proper treatment of subgrid-scale fluxes in CRMs requires more than given by LES-based turbulence schemes probably because the quality of a CRM simulation does not rely on its turbulent (dry) scheme alone + subgrid-scale processes also include subgrid-scale microphysics + subgrid-scale motions and subgrid-scale microphysics interact specific issues regarding parametrisation in CRMs arise from the scales at which boundary layer and moist convection interact computing power available now allows advancing on these issues good timing as CRMs are sollicited to simulate more complex objects and to answer more delicate questions

13. simulation of trade wind cumuli –Stevens et al. (2002) cloud liquid water horizontal cross-section Dx = 80 m Dx = 40 m Dx = 10 m Dx = 20 m (g.kg-1)