Exploratory Analysis Of The 3D Cloud Resolving Model Simulations of TOGA COARE: Preliminary Results

Global climate model studies suggest that cumulus momentum transport (CMT) in tropical oceanic convective cloud systems plays a significant role in the tropical mean circulation and transient variability. CMT is difficult to measure directly and can depend on the detailed structure and organization of the convection. Yet there have been comparatively few evaluations of CMT parameterizations and the assumptions underlying them using 3D cloud resolving model (CRM) simulations. We have analyzed CMT in a four month 3D 64x64x64 gridpoint CRM simulation of TOGA COARE with 1 km horizontal resolution. An additional 256x256x64 large-domain simulation was performed for a 10 day subperiod with strong convection combined with substantial mean vertical zonal wind shear, conditions favorably for strong CMT. Both simulations were identically forced with prescribed vertical motion, horizontal temperature and moisture advection, and relaxation of the domain-mean wind profile to observations on a one-hour timescale. Both were initialized with small amplitude white noise, but spun up realistic convection in less than a day. The domain-mean CMT in the small and large domain simulations for the 10-day common simulation period was compared. The two simulations showed remarkably similar CMT profiles on daily-mean timescales, suggesting that mesoscale contributions to CMT of scales greater than 64 km were small. The skill of a downgradient mixing-length parameterization CMT = Mc*L*DU/Dz was also tested. Here , Mc is convective mass flux, dU/dz is mean vertical shear, and L is a mixing length for updraft zonal velocity perturbations associated with entrainment and horizontal pressure gradient accelerations. This was done by regressing CMT at each height was regressed against Mc*DU/Dz at the same height across all 3D model snapshots over the 10 days. The correlation coefficient describes the accuracy of this downgradient parameterization, and L was calculated as the regression slope. In the upper troposphere (above 6 km), a good linear fit was obtained with L around 1.5 km, but in the mid-troposphere (3-5 km altitude). Binning methods to understand the relative contributions of unsaturated air, saturated updrafts and saturated downdrafts, and principal component analysis of the CMT profiles over the full 120-day record are in progress and will be also reported on.