Problems with copygb on small areas

[SRUDY_78]

Hello,

I got a problem with converting UPP output (grib 1 data) to lat lon grid using copygb. I use the copygb_hwrf.txt file out of UPP to feed the copygb.exe. However, for bigger areas I got no problems, but I tried a high resolution run on a small area and got no values except on the edges of the output grib.

Does anyone know what that can be?

I use UPPV4.1 for postprocessing. Output is ok. The command after unipost has finished is

./copygb.exe -xg "255 0 152 152 34784 139108 128 35467 139944 5 4 64" WRFPRS_d03_2019-08-03_01:00:00 wrfprs_d03_2019-08-03_01:00:00

The WRF run got the following resolutions: 7500,1500,500 meters. Output after copygb for distances 7500 and 1500 are ok, the last one not. I allready extended e_we and e_sn to a bigger size from 109 points to 154 points, but it doesn't work.

Thanks for help in advance.
Best regards
Sebastian

Additionals:

I found on the ncep noaa page this:

Recently discovered bug in IPOLATES. To avoid the bug, all grids should have a size of greater than 1 degree longitude. The IPOLATES library wanted to handle the case of a global grid going from xE to (x+360)E which is be encoded as going from xE to xE. To allow for rounding errors, a grid from xE to yE (y-x ≤ 1) was condsidered to be a grid from xE to (360+x)E. This causes a problem when the grids are ≤ than 1 degree in longitude. This bug will occur occur for both the input and output grids.

I have seen that IPOLATES is used by copygb. Can anyone confirm that this is a bug in copygb, too?

 

[hertneky]

Unfortunately, UPP no longer supports GRIB1 and copygb, so we can't look into it further, but this certainly sounds like it could be the issue. I see in your grid definition, that your longitude goes from 139.108 to 139.944 which does not encompass a distance of 1 degree longitude. Did you try increasing the grid to at least 1 degree. Did that work? If so, then the bug you pointed out is likely the issue. Also did you try GRIB2 output from UPP and using wgrib2 new_grid to regrid ? Did that give the same issue? Since I see the bug statement on the wgrib2 website, I would think you would see the same response with GRIB2 format using wgrib2, but curious if you tried it.

The interpolation routine is part of the NCEP libraries: https://github.com/NOAA-EMC/NCEPLIBS. You could possibly open an issue about the interpolation bug there for grid domains <1 deg longitude.