🚀 go-pugleaf

In article <Bx7J15.190@mentor.cc.purdue.edu>, I wrote:

> I was wondering if anyone knows of a "generic" regional/mesoscale
> forecast model that has been run on (or can be easily ported to) a
> UNIX workstation (in our case, a 25 MIP RISC-based DECstation 5000).
> From the point of view of memory requirements, it seems like a model
> with a 100 X 100 X 20 grid should easily fit on a workstation with ~24
> MB; the main issue would simply be computation time, which is also (it
> seems to me) relatively unimportant if the physics is not too
> elaborate and you're willing to let the model run for a week or so.
> 
> Our idea is to play around with *short-term* (e.g., 6-12 hours)
> simulations of evolution of midlatitude oceanic cyclones after
> initialization with a combination of ECMWF analyses and (possibly)
> satellite-derived moisture data.  At least initially, this would be
> strictly for our own education; that's why we're not yet prepared to
> beg for time using someone else's full-blown model with all the bells
> and whistles and running on a mainframe somewhere...
> 
> Characteristics we're looking for include:
> 	1) hydrostatic, incompressible flow
> 	2) simple bulk parameterization of surface fluxes
> 	3) simple parameterization of condensation/precipitation
> 	4) no radiation, or at least very simply parameterized radiation
> 	5) externally specified lateral boundary values (e.g., from 
> 		interpolated ECMWF analyses)
> 	6) grid resolution ~50-100 km
> 	7) model domain ~1500 km or so
> 
> stuff deleted...
>

I received a number of responses; the most interesting of which are
condensed below.

----------------------------------------------------
From: rmg3@grebyn.com (Robert Grumbine)

  I believe that the Penn State/NCAR mesoscale model has been used
on workstations.  Tom Warner at Penn State would know, as would 
Richard Anthes at NCAR.  A revised version of the model (MM5
as opposed to MM4) is to be released fairly soon.

  The NMC's ETA model is reputedly designed to be portable, but
I don't know if it has actually been tried on any workstations.
Ralph Peterson is the head of that group.

  Colorado State U. has a mesoscale model as well, but I know little
about the model or who is working on/with it.

-------------------------------------------
From zaphiri@watson.ibm.com:

I am familiar with RAMS of Colorado State University, that runs on any
Unix platform. I would strongly reccomend it for the things you are
planning to do.

Pretty soon it will be running on a cluster of RISC and on other parallel
systems. If you need any help with it, contact Roger Pielke, Bill Cotton
or Craig Trembak/Bill Walko.

Zaphiris Christidis

-------------------------------------------
From hbrooks@uiatma.atmos.uiuc.edu:

The best mesoscale model around now is the The Penn State/NCAR model,
MM4.  It beats the life out of the second most common model, CSU-RAMS.
It runs on IBM RISC-6000 stations of a variety of flavors.  I believe
Nelson Seaman at TPSU is the correct contact person.

Harold Brooks                    hbrooks@uiatma.atmos.uiuc.edu
National Severe Storms Laboratory/CIMMS (Norman, OK)

<
****** Note: Nelson Seaman put me in touch with Annette Lario
(lario@nwp01meteo.psu.edu) who allowed me to ftp the source code for
MM4 - Grant Petty

>

-------------------------------------------
From stvjas@meteor.wisc.edu:

Here at Wisconsin, Greg Tripoli is working on (among other things) setting up 
his mesoscale model for real-time runs on a local RISC IBM-6000. Presently, 
someone is working on the analysis scheme and trying to run the Dec 15, 1987 
case study (the 10+mb gravity wave in a blizzard). I don't know what kind of 
resolution will be needed to run a 12 hour forecast fast enough to be useful 
ahead of time. If all goes well, we will eventually be able to (in some 
relatively primative way) assimilate profiler data into the forecast cycle, 
though that isn't even being worked on yet. The model is described in a recent
Mon. Wea. Rev. article (July 1992).

Stephen Jascourt   stvjas@meteor.wisc.edu

<followup message:>

This model runs slower and is much more sophisticated than what you want, and
is a research model for Tripoli's research group rather than a community
model having free access. I posted my message just for comparison, not
intending to suggest you use that particular model-- so you and others can
see who is doing what of a similar nature. Just sending this to make sure
you didn't get the wrong impression.

Good luck! Finally the usenet is being used for the kind of communication it
was intended for-- for people doing similar kinds of things to communicate
about it to each other and their peers.

Stephen 

---------------------------------------------
From: cats@knmi.nl  (Gerard Cats)

Hello Grant,

My name is Gerard Cats, my attention to your request was drawn by Toon
Moene, my office mate at KNMI (Royal Netherlands Meteorological
Institute).  I am the system manager of the HIRLAM project (HIgh
Resolution Limited Area Model), a joint project by Denmark, Finland,
Iceland, Ireland, Netherlands, Norway, Sweden and cooperation with
France and Spain.  The HIRLAM system consists of a complete
forecasting system, with analysis, forecast model and pre- and
postprocessing.  It is a full-blown 1 to 2 days forecast system, it
thus will not necessarily meet your request for simple physics....  It
runs on Cray and Convex.  The forecast model component has also been
ported to DEC stations, MASPAR and even NeXt (by Toon).  It takes some
time indeed on a DEC 5000, in the order of a minute per timestep for a
64*64*16 area.  The timestep for a 50 km resolution is 5 or 6 minutes,
it takes several hours for a 36 hours forecast, a Convex 210 is say 5
times faster.  To get at the code, you will have to contact the
project leader, Peter Lonnberg, (Finland, email (I guess)
fnp@ecmwf.co.uk, tel.+358019291, address Finnish Meteorological
Institute, PO Box 503, SF-00101 Helsinki, Finland).  There are
agreements on code distribution, I guess free for research.  People
with most experience porting the model to DEC are S Thorsteinsson,
Vedurstofa Islands (Iceland),fax +354125411/+354128121 Nils
Gustafsson, SMHI (Sweden) fax +4611170207 Lex wolters, Leiden
University (The Netherlands), email: llexx@rulwi.leidenuniv.nl Nils
and Lex are mainly interested for porting the code to MASPAR (1 to 16k
processors), Nils for the spectral model version, Lex for the
gridpoint model (the latter is now the default HIRLAM system). To
avoid complications I add that none of these 3 is allowed to provide
you with the code.

----------------------------------------------------
From: amos shapir <amos@CS.HUJI.AC.IL>

I have seen exactly such a system on an IBM RS/6000 workstation, in use
by the Air Force.  Unfortunately, details are classified of course.  All
I can say is they used standard X-windows libraries, and it didn't take
long to develop - maybe it's even based on some PD package (I can ask
next time I'm there, which may happen in a month or two).

-- 
	Amos Shapir		Net: amos@cs.huji.ac.il
Paper: The Hebrew Univ. of Jerusalem, Dept. of Comp. Science.
       Givat-Ram, Jerusalem 91904, Israel
------------------------------------------------------

-- 
Grant W. Petty                                gpetty@rain.atms.purdue.edu
Assistant Prof. of Atmospheric Science        (317) 494-2544
Dept. of Earth & Atmospheric Sciences
Purdue University, West Lafayette IN  47907-1397