ERA-Interim and regrid2 v. xdef/ydef

[Michael Fiorino]

dear matt,

the pdef option is generally used for grids where lat/lon are *not*
univariate or the # of points in longitude varies with say latitude
(e.g., reduced gaussian grid).  it is possible to force grads to treat
the uniform 1.5 deg lat/lon grid as preprojected by creating a pdef file
with the interpolation coefficients.  see:

http://grads.iges.org/grads/gadoc/pdef.html#file

this is clearly more complicated and NOT easier than using a regridding
function.

i have to say i'm a bit concerned that a reason you wouldn't want to use
regrid2 is because you're 'computer illiterate'.  that strikes me as a
bit irresponsible; that you would blindly accept the result from an
internal grads interpolation?

you'll have to do a little work to either get a version of grads with
regrid2 already built in; build it yourself; and then understand what
the code is actually doing if you want to understand the nature of the
regridded era data...  for example, if you want to regrid the 1.5 deg
era data to higher resolution, then you should use either bilinear or
3rd order bessel interpolation if the field is continuous (e.g.,
pressure), but for
less-continuous fields like precipitation, box averaging would be more
appropriate.  all these methods are implemented in regrid2, but you have
to be careful in which method you use.

fortunately, the technical issue of using regrid2 should be less
burdensome as i have converted my regrid2 'user defined function' into
an opengrads 'user defined extension'.  the extension now works in both
opengrads 1.10 and opengrads 2.0.  the opengrads version is called 're2'
and works like the original regrid2 except for a few bugs i found in the
original regrid2.f code so that re2 is actually more correct than
regrid2....

i'm finishing up some other extensions (e.g., mfhilo - for finding
max/min in a 2-d field) and i'm now checking that they work in both
opengrads 1.10 and 2.0.  i expect to get these into the opengrads
distributions 1.10 and 2.0 shortly, but in the meantime re2 is available
the opengrads2.0 bundles at:

http://sourceforge.net/project/showfiles.php?group_id=161773&package_id=305032

the re2 in these bundles is not 100% the same as the one i'm working
now, but the algorithms are correct so that you can use with
confidence that it's working the way i intended.

hope this helps, but remember it's always important to state precisely
which version of grads you're using...

regards, mike


Matt Jeglum wrote:
>  I have been using the ERA-Interim dataset, and after running it
through grib2ctl I get control files
>  that work, but do not include any sort of PDEF specification.  I am
ultimately interested in
>  interpolating the 1.5 degree grid on the ERA-Interim to a higher
resolution.  I have heard of two
>  methods to do this.  One is to use the regrid2 function, but since I
am computer illiterate and the
>  other option is quite easy, I am avoiding that.  It seems possible to
change the xdef and ydef
>  specifications in the ctl file to interpolate to the higher resolution
to begin with.  All you would need
>  to change is the # of grid points and the increment.  This has worked
for others but not for me, and I
>  suspect it may be due to the lack of PDEF.  Is this method legitimate?
If so, anyone know the ERA-
>  Interim projection specs for PDEF?