Hi Ryo and Olivier,
Thanks very much for the explanations. Having checked the dimensional axes of the example model output (ncar_ccsm_3_0) from the ipcc4 database, I found the variables are staggered on the B-grid, and the pre-cooked meridional overturning streamfunction (stfmmc) is located at the same latitudes as (u, v) and staggered in the vertical direction. Integrating from the bottom definitely gives a better matching (reproduce with the following script), but doesn't eliminate noisiness and discrepancy esp. in the upper ocean of my interest (which is intuitive since we integrate v from bottom up). Since it's an integration, the error can easily accumulate up.
It's good to use the pre-cooked meridional overturning streamfunctions (stfmmc) in the ipcc4 database, but only a limited number of runs have reported this particular variable. For example, ncar_ccsm3_0 doesn't have it for the scenario A1B, ukmo_hadcm3 doesn't report this variable at all. That's the reason why I was to calculate it from the meridional velocity. anyway...
D.
! the meridional velocity
use "http://esgcet.llnl.gov/cgi-bin/dap-cgi.py/ipcc4/20c3m/ncar_ccsm3_0/pcmdi.ipcc4.ncar_ccsm3_0.20c3m.run1.ocn.mo.xml?vo,time,depth,lat_395_a,lon_320"
! the pre-cooked meridional overturning streamfunctionuse/order=zxyt "http://esgcet.llnl.gov/cgi-bin/dap-cgi.py/ipcc4/20c3m/ncar_ccsm3_0/pcmdi.ipcc4.ncar_ccsm3_0.20c3m.run1.ocn.mo.xml?stfmmc,time,region,depth_41,lat_395"
let psi_precooked = stfmmc[d=2,i=4]/1e6let psi_calculated = vo[d=1,x=@din,z=@iin]/1e6 - vo[d=1,x=@din,z=@din]/1e6
ppl cross 1con/lev=10d/l=1 psi_precooked
con/lev=10d/l=1/over psi_calculated[gz=psi_precooked@asn,k=@shf:-1]On Fri, Jul 25, 2008 at 2:07 AM, Olivier Marti
<olivier.marti@xxxxxxxxxxxx> wrote:
Le 23 juil. 08 à 19:06, David Wang a écrit :
Hey Ferreters,
I calculated the meridional overturning streamfunction from the meridional
velocity field of an ipcc4 model using the simple syntax as:
vo[x=@din,z=@iin] ! vo is the variable for the meridional velocity
I compared the result with the pre-cooked meridional overturning
streamfunction from the same model output, and found the two have more or
less the similar patterns but the FERRET calculation is far more noisy. Now
it's unclear to me if the syntax I used is not accurate enough or something
else is responsible for the difference. I attach both the script and the
plot (the black contours are the pre-cooked, the red ones are computed by
FERRET).
David,
If I remember well, the CCSM ocean grid is not lat/lon. The 3D velocity fields in the IPCC database has been interpolated on a regular lat/lon grid. This interpolation is not likely to conserve all properties of the fields, like the non-divergence. Then you cannot get a correct overturning stream function from this field. On the contrary, CCSM probably has a method to compute correctly this function from the original grid. You should then use what they provide in the IPCC database.
Olivier
--
Olivier Marti mailto:olivier.marti[ad]lsce.ipsl.fr
Laboratoire des Sciences du Climat et de l'Environnement
Institut Pierre Simon Laplace
Laboratoire CEA-CNRS-UVSQ
Tel : +33 1 69 08 77 27 - Fax : +33 1 69 08 30 73
<http://www.ipsl.jussieu.fr/~omamce>
-----------------------------------------------
GEOSCIENTIFIC MODEL DEVELOPMENT
A NEW EGU JOURNAL FOR
DESCRIPTIONS OF MODELS OF THE EARTH SYSTEM
http://www.geoscientific-model-development.net
-----------------------------------------------