Hello,
First let me say that there are *no* "problems of spherical
harmonic filter"
Background:
I know that Song Feng has calculated Eliassen-Palm (E-P)
fluxes from assorted model runs. Also, I think from
NCEP Reanalysis. The E-P fluxes provide
a measure of the local acceleration [ du/dt ]
as a result of eddy heat and momentum fluxes.
There are several variants of these EP fluxes.
In the variant Song Feng is using, three terms are involved.
Song Feng notes:
" ... my code work reasonable well for 4 variables. For the
other two variables (dudt, divk) the results are also good
except for 925-700hPa levels."
Ummm, if something were wrong with the spherical harmonic
functions they would have been noted at all levels not
just with the 925-700 layers. What is unique to these levels?
Double Ummmm! well, mountains, of course. In the real world
there are no continuous fields at these levels. TAntarctica
continent, Andes, Rockies, Himalays, etc all prevent this.
For convenience (mostly graphical) grid points affected by
mountains are artifically filled with interpolated or
extrapolated values. Make no mistake about these numbers
they are *all bogus*. In the case of, say, temperature the
fields might look pretty 'reasonable., perhaps, because a standard
lapse rate has been used. (Also, for geopotential
height.) However, for the most part
these values should *not* be treated as realistic and should not
be used in any equations.
In the case of EP-fluxes, all terms in the EP equation were
*extrapolated* to the 925-700 levels. In Song Feng's case
only vertical extrapolation was used.
They might *look* ok because the mountainous areas are local
and relatively small scale. However, when a global spherical
harmonic operator is applied, the small scale stuff affects
all wave numbers. The results are contaminated.
Bottom line: one should not apply a global operator to
fields that affected by mountains.
good luck
This archive was generated by hypermail 2b29 : Tue Feb 19 2002 - 13:21:25 MST