Here are some further instructions.

vertical resolution.The LES models have been run on a very high vertical resolution, typically with a grid spacing of 40 meters. Such a high vertical resolution (100 levels in lowest 4 km of the atmosphere) cannot be afforded in Atmospheric General Circulation Models. On the other hand, at present most AGCMs run at a rather ``low'' vertical resolution, with typically only 10-15 levels in the lower 4 km of the atmosphere and sometimes less. For the present case, this means at best 2 or 3 levels in the (for mass flux schemes the active, positive buoyant) cloud. Initial experiments with our model indicated that the results degrade very rapidly when the active cloud is resolved by less than 4 grid points. We expect this sensitivity to be similar in other SCMs because most models use a similar (``advanced'') combined massflux/turbulent-diffusion approach (and it will be interesting to see if this is true).

Although we certainly agree that the models should be run at the resolution used in GCMs, we also feel that analysis of the results is almost impossible on a low vertical resolution. One should note that the lower resolution might benefit from the high resolution results, as coding errors, numerical instabilities, and ``fawlty'' physics become more apparent on high resolution. In addition, a gradual increase in vertical resolution might be expected in the coming years.

Our approach can be summarized as follows: "first understand the physics of the SCM on a rather high resolution (with small numerical errors) and then investigate the degrading on lower resolution.'' We therefore request to run your model on (at least) two resolutions.

``standard (low) 19 level resolution'' : 60 level ECMWF resolution

This is actually not such a low resolution for a GCM. It has about 20 levels in the lower 4 km of the atmosphere, but most of the levels are concentrated in the lowest 1 km. In the cloud the grid spacing is approx. 300 m, equivalent to 3 or 4 levels in the cloud. This resolution should be considered as the minimum resolution needed to take advantage of the ``advanced'' physics most SCMs use.

``high 40 level resolution''

40 levels in the lowest 4 km. In the cloud the resolution is 100-120 meters. The cloud layer is now rather well resolved by 6-8 levels.

Initial profiles at these resolutions and the grid can be found here. One may also choose to replace the latter high-resolution by any other higher resolution (but you will have to do the interpolation yourself; please check carefully). The suggested resolution is the lower limit of what I call high resolution.

if possible also run the model on your present operational resolution. For this resolution one only needs to sent in setA (mean profiles) and setE (time series).

Some model might not be able to run with half an atmosphere (extending to only 4 km). If that is true the profiles should be extrapolated. It is perhaps easiest to take a very stable, very dry atmosphere above 4 km to rule out any interference.

time resolution. Most GCMs run with a large time step (order 5 minutes - 30 minutes). However, if possible, please run you model with a relatively small time step of order 1 min - 5 minutes. The reason is that numerical errors caused by the time descretization are very unpredictable, thus making an analysis of the results almost impossible. For your own interest you may also run with larger time steps.

requested output. The original requested output was focussed on the LES models. We defined a somewhat different set of output parameters more suitable for the SCMs. In particular, more output is requested with respect to the turbulence scheme and the massflux scheme. The list of output variables are here. Please adhere the formatting requirements.

surface fluxes. From previous SCM intercomparisons it appears that in some models it is not easy to prescribe the surface fluxes. In many models this means that the surface routine has to be modified because the surface flux is implicitely calculated in the turbulence scheme. Some small errors in the surface fluxes are therefore acceptable, as long as these stay within 5 % of the actual flux (so at most 25 W/m2 for the latent heat flux during day time).

specific humidity or mixing ratio.In the original case prescription there was some confusion between specific humidity qv and mixing ratio rv. The initial profiles in the original case description are in rv. We give both qv and rv. But don't worry too much about this because differences are only of order 1%, as qv = rv / (1 + rv) ,with rv order 0.01 kg/kg (10 g/kg). For the forcing tendencies (advection) the difference between qv and rv can be neglected.

radiation. Radiation (both longwave and shortwave) should be switched off completely.