[Wrf-users] SST and masking in Metgrid

Olafur Rognvaldsson or at belgingur.is
Wed Nov 25 15:26:42 MST 2009

Dear co-users, first of all let me apologies for the somewhat lengthy

We have been scratching our heads regarding the question of how Metgrid
handles masking and/or 2D interpolations and are hoping that some of you
might shed some light on this issue.

What we are trying to do is to scale down a future climate simulation,
made by the Arpege model, for a limited are using WRF. We soon found out
that "real.exe" was for some reason producing sea-ice in very strange
places (the Mediterranean sea for one:) This than caused numerical
instability and the WRF blew up after few weeks of simulation time.

We found out that it was the SKINTEMP variable that was the cause of
this sea-ice issue. The reason being that the coarse resolution SKINTEMP
variable from Arpege "overlapped" into the ocean at various coastal
areas such as south of the Alps, southeast of Vatnajokull glacier in
Iceland and the northern part of the Baltic sea.

To solve this we decided to create a new field from the Arpege data,
namely SST, by "masking" the SKINTEMP field by the LANDMASK variable. In
short, the SST is the same as SKINTEMP except over "Arpege" landpoints,
where it is given a default "NaN" value (just like "regular" SST field

Feeling pretty good about ourselves we now recreated the intermediate
files (FILE:2020-01-01..., etc. etc.) and reran "metgrid.exe" with some
slight modifications to METGRID.TBL. We re-defined SKINTEMP only over
land as to not create erraneous values over the ocean.

Now the strange things happens. We are simply unable to create a
plausible SST field where the SST values are correctly interpolated
towards the "true" coastline (i.e. as defined by the high resolution
geo_em_d0?.nc files).

The "SST-4pt.png" file is created by the following lines in METGRID.TBL

#        masked=land
#        interp_land_mask=LANDMASK(1)

The strange thing here is that "masked=land" and
"interp_land_mask=LANDMASK(1)" have nothing to say, i.e. the resulting
figure is the same whether this is commented out or not (i.e.
SST-4pt.png and SST-4pt-landmask1.png (not shown) look the same).

The next two figures, i.e. "SST-16pt.png" and "SST-16pt-landmask1.png"
are created in a similar way:

#        masked=land
#        interp_land_mask=LANDMASK(1)

What differs here is that when we use the 16pt interpolation scheme the
"masked=land" and "interp_land_mask=LANDMASK(1)" parameters in
METGRID.TBL start to kick in (which they did not when we only used the
4pt scheme). But still the Metgrid code is definately not doing what it
(according to us:) should be doing. One can see very strange values off
the coast of Iceland in both "SST-16pt.png" and "SST-16pt-landmask1.png".

This behavior is the same for versions 3.0.1 and 3.1.1 of the WPS package.

So, we were wondering if anyone knew what might be going on here. Are we
totally misunderstanding what Metgrid should be doing or are we perhaps
making an embarrassingly silly mistake somewhere along the road???

Best regards, Ólafur.
Ólafur Rögnvaldsson
Reiknistofa í veðurfræði - Belgingur
www.belgingur.is - www.riv.is
-------------- next part --------------
A non-text attachment was scrubbed...
Name: SST-4pt.png
Type: image/png
Size: 33346 bytes
Desc: not available
Url : http://mailman.ucar.edu/pipermail/wrf-users/attachments/20091125/f41ebfaf/attachment-0003.png 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: SST-16pt.png
Type: image/png
Size: 76746 bytes
Desc: not available
Url : http://mailman.ucar.edu/pipermail/wrf-users/attachments/20091125/f41ebfaf/attachment-0004.png 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: SST-16pt-landmask1.png
Type: image/png
Size: 70011 bytes
Desc: not available
Url : http://mailman.ucar.edu/pipermail/wrf-users/attachments/20091125/f41ebfaf/attachment-0005.png 

More information about the Wrf-users mailing list