Axel's ROPP8.0 wishlist

[Ian Culverwell]

Axel von Engeln (EUMETSAT) undertook some beta testing of ROPP8.0. Here are his findings:

Hi Ian,

attached the first version of my modifications (sorry, directly the
file, no patch - should be simple to figure out). Also one, rather
difficult (draft) file we have, which I have used for testing so
far.


On my observations/review of ROPP-8 so far, all minor issues:

- Installation:
 
  - I wanted to install the latest version first into my local
environment, but use all the libraries available in the general
environment (e.g. netCDF, HDF, bufr, ...). This seems to be not
possible

    with the setup, so I had to tell it to look into the local
installation directory (PREFIX) first, and then search also in our
general environment. I think that could be improved. Is there an 
additional library path variable available? Did I miss it?

  - ropp_pp installation with extra aclocal, autoconf run is a bit
odd. Can that be improved?

  - 1DVar test not working (I know you are on it).

- Default Output Levels:

  - For a comparison of refractivities, I wanted to quickly run
ROPP on more than the 200m default resolution from 200m to 60km.
This is rather difficult if one does not have a full level 1B 
file available (including some level 1a variables it seems). So I
just changed the code and recompiled for another default grid. But
it would be nice if one can also give that on the command line.


- ropp_fm_bg2ro_1d

  - My first runs that did not set the attribute institution seem
to result in the bg_source being put there. Is that possible?

  - If I don't provide a lat_tp entry it seems that some elements
are set to 0 and the rest to missing (WRONG: Just re-checked,
ROPP-8 is doing it correctly now, ROPP-7 had the issue ;-) )

  - I think the compressibility, improved interpolation should be
the default if they provide better results

- other issues:

  - I think the netcdf_getatt does not provide an error message
when the attribute is not found (also with -d option). When the
directory to an attribute is not found it seems to be working.

  - the global attribute software_version seems to allow only a
number, but we process it also with Yaros, so I need either another
field, or one that accepts strings plus numbers

  - the global attributes seem to be converted to capital letters,
but only the first word

  - the global attribute thin_method seems to always attach a
[v3.1], even if I set it to a method. This is likely some ROPP
default, but we use another thinner.

  - the attached new GRAS 1B file has a time offset of some 20ms,
but that seems to be set to missing later on. I agree, this is 
rather lousy file, but I checked that it is correctly done in my
code segments. So this must happen later in the code.

As said, this is the first version of our new files, for you to
start looking into issues. We will update the format further, but
likely not the segments that are used by eum2ropp. I will provide
more test files once we are done.

I also did little tests on the option to write out level 1a data. I
only checked that the resolution thinned and high_res is working
for level 1b.

We'll work thought them.

[Ian Culverwell]

ropp_io_read_ncdf_get.f90

[Ian Culverwell]

GRAS_1B_M02_20120909000057Z_20120909000118Z_N_T_20141021103321Z_G15_NN_0001.nc

[Ian Culverwell]

First set of fixes committed at r4361:

Updates to EUM lev1a reader in response to ROPP8.0 beta testing: (1) revised ropp_read_ncdf_get.f90; (2) introduction of new global attribute 'processing_software', to hold info about YAROS etc; (3) remove annoying [v3.1] from thin_method if not from ROPP; (4) preserve case of global attributes; (5) strengthen checking of creation_date.

[Ian Culverwell]

Also included options to allow user to specify {min output GPH, max output GPH, number of uniformly spaced levels between them} as input options to ropp_fm_bg2ro_1d at r4356.

[Ian Culverwell]

Extension of valid range of time_offset from (0, 240) to (-10, 240) were made at r4362. (Slightly negative, but legitimate, values were being set to ropp_MDFV, as Axel found.)

[Ian Culverwell]

That's as much as I can do now, so closing the ticket.