DYCOMS II SCM results: Ping Zhu & Chris Bretherton

3 Oct 2003

Dear Martin et al.

Attached are a few results from 6 hour runs of the NCAR single column
model (the 'SCAM') applied to the GCSS WG1 case using various vertical
resolutions and timesteps. This is just to put some SCM results out on
the table for this case to stimulate discussion as to how to best do
this comparison. We'll try 48 hour runs later. All results are 0-6 hour
averages using the initial sounding and forcings we agreed on. We ran
the standard CAM PBL scheme (with and without the Hack shallow
convection scheme also used in the standard CAM) as well as the UW PBL
scheme implemented in CAM, with and without our shallow cumulus scheme
(shcu) and with and without the CAMs Klein-line parameterization (st)
for cloud fraction under strong inversions. The first figure shows
results for all these model configurations at various vertical
resolutions (26 levels is the CAM default) with the default CAM dynamics
timestep of 20 mins (actually this means a physics timestep of twice
this long). Fig 2 repeats this plot but with a 2 minute timestep; with
the UW scheme this sharpens and lifts the cloud layer. In either case,
the Hack scheme destroys the Sc in the default CAM, though not so much
with better vertical resolution, and the UW scheme tends to maintain the
correct inversion height better than the default scheme, especially at
high vertical resolution. Figs 3,4 repeat Figs. 1,2 but with Bjorn's
idealized radiation instead of interactive radiation; results don't seem
too sensitive to this.

We can get more discriminating results from a much longer, 6 day,
case from EPIC 2001Sc at 20S 85W which we have now developed into a good
SCM case (helped by ECMWF diagnoses of time-dependent vertical velocity
and other forcings.) I'll show that case and some results in the Friday
morning (31 Oct) session in Broomfield.

17 Oct 2003

Dear Martin:

Attached ASCII files are the results of my SCM simulations. "30l", "26l",
and "60l" indicate the model level that I configured. "20min" and "2min"
indicate the time step that I used in the simulations. "noHack" means that
I turned off Hack shallow convection scheme in the standard SCAM.
"UW_SCAM" indicates that I used UW PBL and shallow convection schemes.
"simple_radiation" indicates that I used simple radiative scheme suggested
by Bjorn. Let me know if anything's not clear.

Thanks,

ping

Results

short ascii dataset

Figure 1: (postscript)

Averaged profiles (7hrs) of 26 level, 30 level, and 60 level simulations with time step of 20 minutes. The simulations ues interactive radiative scheme. Symbols in the figure represent: cam(o): standard SCAM with Hack shallow cumulus scheme turned off; cam(o)+Hackshcu: standard SCAM with Hack shallow cumulus scheme turned on; cam(uw) st: UW boundary layer schemem with inversion cloud cover fraction (marine stratus) turned on; cam(uw) stno: UW boundary layer scheme with inversion cloud cover fraction turned off; cam(uw) stno+shcu: UW boundary layer scheme with UW shallow convection scheme.

Figure 2: (postscript)

The same as figure 1 but with time step 2 min.

Figure 3: (postscript)

The same as figure 1, but using simple radiative scheme suggested by Bjorn (see description of RF01 case).

Figure 4: (postscript)

The same as figure 3 but with time step 2 min.