Evaluation and Improvement of the Turbulence Parameterization in Deep Convective Clouds.

Abstract:

Although turbulence processes have been extensively studied for the boundary layer, there are few studies that evaluate the turbulence parameterization inside convective clouds in atmospheric models. Yet, turbulence can be strong inside cumulus and cumulonimbus and can have an impact on the structure and dynamics of these clouds. This study aims at evaluating and improving the parameterization of subgrid turbulence in deep convective clouds simulated by numerical cloud resolving model at kilometer scale.

First, we have characterized the turbulence representation in deep convective clouds. For that, a Large-Eddy Simulation (LES) using simplified atmospheric conditions has been performed with the Meso-NH model to serve as a reference simulation of deep convection. This LES with a 50-m grid spacing is used to compute the turbulent fluxes at different coarser horizontal resolutions (500 m, 1 km, and 2 km). Vertical turbulent fluxes of liquid water potential temperature and non-precipitating total water mixing ratio have counter-gradient structures, indicative of nonlocal turbulence.

Second, a diagnostic assessment, from the reference fields, of the current turbulence parameterization of the Meso-NH model (subgrid scheme with a 1.5-order closure and diagnostic equations for the fluxes) at these coarser resolutions shows that turbulent kinetic energy is largely underestimated in the clouds, related to an underestimation of thermal production. The counter-gradient structures of vertical turbulent fluxes are not reproduced, indeed, the local K-gradient formulation is not suitable. Alternative parameterizations of some turbulent fluxes, proposed in the literature, are then tested. In particular, a parameterization based on horizontal gradients, gives a better representation of the thermal production of turbulence in the clouds, with a good representation of counter-gradient areas.

Third, the on-line evaluation from model runs with 2-km, 1-km, and 500-m horizontal grid spacing confirms the improvement when using the modified scheme, with an increase of subgrid turbulence and a significant decrease of vertical velocities in convective clouds.

In the next future, we plan to evaluate and to improve this new parameterization on real cases of deep convection from the two-month HyMeX field campaign.