DYCOMS II SCM results: Adrian Lock, RPN, Canada

15 Oct 2003

Dear Martin,

I've succeeded in running the DYCOMS simulation with the Canadian
SCM. Unfortunately this is something of an off-line model, in that it
only calls the boundary layer scheme and not their convection or
microphysics schemes - so no chance to test interactions between
parametrizations or drizzle. It's also the first time someone has tried
to do SCM simulations with cloud and I'm finding their operational
statistical cloud scheme (after Bechtold and Siebesma) is giving rather
strange q_l profiles at the moment. Because this may well be something
to do with the code interface, I'm only submitting now some tests I've
done with just an all-or-nothing condensation scheme.

So, it's a 1.5 order turbulence model with Bougeault/Lacarerre
mixing and dissipation lengths. I've run at two resolutions, uniform
50m and the operational resolution (12 levels below 1800m). The 50m
resolution simulation has a timestep of 300s and I've used 300s, 450s
and 900s at the operational resolution. I've run for 48 hours using
Bjorn's step 3 set-up (full idealised radiation scheme, subsidence,
specified surface fluxes) and the profile plots are averages over 0-6
hours (solid lines), 42-48 hours (dotted lines) and 3.5-4 hours
(dashed). The colours are the same in all figures.
I've also done plots for Bjorn's full set of time series but
haven't worked out how to do netcdf yet, so only the ascii datasets
you asked for at this point - I hope they're self explanatory (they
have a one line description at the top if not). Let me know if there's
anything else I could plot for you, or if you'd like to have the full
output in ascii, or anything else.

Basically the results seem OK to me, particularly if you stick to
the first few hours! It's always interesting how sensitive the cloud
is, though. There's very little difference in the bl mean TKE or
buoyancy flux between all these simulations but it's enough to give a
large variation in LWP on the time scale of a day - all caused by
variations in the entrainment rate, of course.

I won't try and interpret these results any further yet - a lot of the
spikes may be to do with the simple cloud scheme I'm using - but I thought
I'd let you know I was making some progress! Unfortunately, the Met
Office SCM is not proving very portable so I'm not very optimistic I'll
have any results from that in time for the meeting.

All the best,
Adrian

16 Oct 2003

corrected results:

Dear Martin
I've just discovered an error in my model - some parts were defining
Exner in terms of p_surf and others in terms of 1000hPa. Correcting it
makes some differences, not least of which is an instability that
appears in the TKE equation for timesteps of 5 minutes or greater at 50m
resolution - I don't know why this wasn't there before!

So, I've attached a corrected set of figures and new datasets (with
T instead of theta_l), everything else is the same as before. Note that
there now appears to be a coherent signal in terms of entrainment rate:
longer timestep equals greater entrainment rate at both resolutions. I
hope to get some results with the Bechtold and Siebesma cloud scheme
soon, not least so that I can ask them about it at the meeting!

All the best,
Adrian

17 Oct 2003

Dear Martin'

As promised I've got some more results for you from the Canadian SCM
but this time using the Bechtold, Siebesma et al statistical cloud
scheme (instead of an all-or-nothing scheme). Same format as before -
high and operational resolutions and two different timesteps for each.
Results are now really very consistent under changes to timestep and
resolution, with only a hint that shorter timesteps give more
entrainment. Note that I'm running step 3 with fixed surface fluxes and
so the near surface bl actually just gets warmer than Bjorn's 295K used
to diagnose zi. I'll try and switch to fixed SST to match your
simulation.

The only thing I'm not too sure about now is the rather strange ql
profiles. It doesn't look so bad for RF01 because the cloud is quite
shallow but, particularly at 50m resolution, you can still see a
tendency for a large peak in ql at cloud-top (greater than the
supersaturation) with almost uniform (and rather small) ql throughout
the rest of the cloud layer. Something for me to look into, at least!

All the best,
Adrian

Results using all-or-nothing cloud scheme

(short ascii data set)

Figure1: (postscript)

Figure 2: (postscript)

Results using Bechtold & Siebesma statistical cloud scheme

(short ascii data set)

Figure1: (postscript)

Figure 2: (postscript)