The recent developments carried out at CNRM/GMAP concerning the "CYCORA" package were tested for the TOGA-COARE and CLEOPATRA (the squall line of July 21, 1992, over southern Germany) cases. For the turbulent diffusion parameterization new values of the critical Richardson number and of the exponent controlling its vertical profile have been proposed: USURIC=1 (USURIC is the inverse of the critical Richardson number) and USURICE=0.5, insuring the theoretical expected values of the critical Richardson number, 1 for laminar turbulence and 1/4 in three- dimensional turbulent regime. These values together with the parameter GCCSV=1, increasing the occurrence of the cases when the shallow convection correction is applied, led to better results as one can see from the evolution of the convective mass-flux for the TOGA- COARE case (see Fig. 1, upper part - "CYCORA" simulation, bottom part - "CYCORA" + new tuning): the convective activity was enhanced in the last part of integration, when the squall line should reach a quasi-stationary state.
Regarding the deep convection parameterization, tests were run concerning the "ensembling entrainment", developed by J. -F. Geleyn and J. M. Piriou in order to get higher convective clouds tops. The idea is to introduce a relaxation of the moist static energy of the cloud towards the moist static energy of a fictitious "non entraining ascent". The relaxation coefficient is locally modulated by the complement to 1 of the factor of variation of entrainment rate between its maximum and minimum in order to have the maximum effect at the top. By using the "ensembling entrainment" the simulation of the TOGA-COARE squall line was significantly improved:
- the clouds reached higher levels (see Fig. 2 : mass flux evolution, upper part- experiment "gcvff");
- the convective activity was enhanced in the last part of the of the simulation; (see again Fig. 2)
- the precipitation slightly increased but the evolution became more spiky (see Figure 3, experiment "gcvff").
The presented result was obtained by using the new tuning for the turbulent diffusion and the proposed values for the "ensembling entrainment": relaxation coefficient GCVFF=2.E-05, critical thickness of the precipitable clouds ECMNP=3000, the coefficient controlling the dependence of the entrainment rate on the integral cloud buoyancy GCVALFA=5.E-05. The experiments have showed that convective activity could be enhanced by the following set-up: GCVALFA=4.E-05 and the minimum and maximum value of the "ceiling" entrainment rate at the top at the basis of the clouds, TENTR=3.5E-06 and TENTRX=7E-05.
The "ensembling entrainment" was tested also for the squall line of July 21, 1992. The sensitivity to this modification was smaller in this case (simulations have been carried for a 7.3- km resolution) and no improvement was noticed: the displacement speed was still to fast and the vertical velocity nuclei were even smaller.
Another tested modification (developed by J.F.Geleyn) concerned the smoothing of the humidity turbulent diffusion flux for the deep convection computation: this time the humidity flux was scaled by the saturation humidity. The experiments realised for the TOGA- COARE case showed a smoother evolution for the convective mass flux (see Fig. 2, bottom part, experiment "test") and also for precipitation (experiment "test" from Fig. 3). Even if the convective activity was diminished the amount of precipitation at the end of the integration was close to the result obtained with the CNRM 3D cloud resolving model (labelled "rsi3d" in Fig. 3) due to the increase of the stratiform precipitation. It should be mentioned that for these experiments the smoothing of the humidity turbulent flux (with a smoothing coefficient GCVPSI=0.5) was used together with the "ensembling entrainment".
The research on the vertical acceleration in the environment of a hydrostatic atmospheric model has significantly advanced during the last period. The acceleration has been derived from the expression for the advecting vertical velocity which is based on the continuity equation. It has required second order space derivatives of the prognostic variables to be obtained. This is a relatively easy but memory and time consuming procedure. Additionally, a prognostic equation for the divergence had to be simulated. It has been derived from the momentum equation and is only adiabatic. Some additional simplification has been allowed in the framework of a high resolution limited area model such as no curvature terms and no meridional derivative of the Coriolis parameter.
The uniformly accelerating motion trajectory scheme, proposed previously, has been tested with the new vertical acceleration. It has been compared to other trajectory schemes as well. Two extreme cases has been used for testing - the 'Baltic Jet' case and the 'Cleopatra' one. The first one is particular with a dominating horizontal advection and the second one is characterized by a strong vertical motion within a convection plume. The proposed trajectory scheme was the only one to assure stable integration without noise in the 'Baltic Jet' case as well as sufficiently well developed convection in the 'Cleopatra' case.
Some perspectives for the 2TLSL method have been foreseen. For the hydrostatic model the predictor-corrector method may be an alternative solution for the present deficiencies of the operational scheme. The vertical SL trajectory problem is very interesting in the non-hydrostatic model where the vertical velocity is an independent prognostic variable.
The first version of the thesis is ready and an article has been submitted to QJRMS and accepted with very minor revisions. The defense will take place before the end of 2000 and you should know everything about it in the next Newsletter.
The new horizontal diffusion scheme has been introduced into the ALADIN model. The scheme, using the damping ability of different semi-Lagrangian interpolators, acts like an gridpoint diffusion operator. The advantage of this new scheme compared to a real gridpoint horizontal diffusion is mainly the computational efficiency. It is very cheap and it allows for the same computational cost to use a large variety of diffusion orders. On the other hand the e-folding time (diffusivity) of the new diffusion is a function of its diffusion order (selectivity) and thus cannot be chosen independently.
The new scheme was tested in a parallel suite and compared to the current operational version of Aladin/LACE with satisfying results. Some case studies also showed an improvement of the forecast (the 97's Christmas storm in Ireland and the 99's Christmas storm in France).
The advantages (+) and disadvantages (-) of the new diffusion scheme when compared to the current spectral horizontal diffusion scheme are the following (and consisten)t:
An article and of course the whole thesis is being written with, hopefully, a defense at the beginning of 2001. More news with, may be, the defense schedule or result, in the next Newsletter.
|
Home |