search

NOAA-EMC / NCEPLIBS-ip

Fortran 90 subprograms to be used for interpolating between nearly all grids used at NCEP.
Other
5 stars 9 forks source link

add support for rotated lat/lon in ip2 code to support wgrib2 #211

Closed edwardhartnett closed 7 months ago

edwardhartnett commented 9 months ago

wgrib2 has it's own internal copy of the ip2 code, with come changes to support rotated lat/lon grids.

These need to be supported in NCEPLIBS-ip.

When @webisu gets back from vacation in January he can provide details.

Somehow the lack of rotated lat/lon is causing two different versions of wgrib2 to be built, one with, and one without. See https://github.com/NOAA-EMC/wgrib2/issues/70.

This is needed for the wgrib2-3.3.0 release, but first we have to do the wgrib2-3.2.0 release. So this will be needed sometime in Q1 of 2024 (as a new release of NCEPLIBS-ip which can handle rotated lat/lon and can be used by wgrib2.)

edwardhartnett commented 7 months ago

@AlexanderRichert-NOAA we need this to be fixed to get the new wgrib2 build working.

AlexanderRichert-NOAA commented 7 months ago

Okay, looking into this now

AlexanderRichert-NOAA commented 7 months ago

@webisu @edwardhartnett would you be able to point me to the missing routines? I seem to recall Kyle's merge of ip2 into ip was fairly comprehensive, but I may be missing something.

webisu commented 7 months ago

Alex, If my old discussions with George Gayno, ip2lib did not support parameters 32768 and 32769

https://www.nco.ncep.noaa.gov/pmb/docs/grib2/grib2_doc/grib2_table3-1.shtml

When interpolating NAM and RFFS rotated lat-lon grids, the grib2 files will use a grid template value of 32768 or 32769. Wgrib2 would convert these values to a grib1 gds of 203 or 205. Then it would call the grib1 version of iplib

You can search for 203 and 205 in the grib1 version of the library. That may or may not help you to find the codes that need to be fixed in the grib2 ipolates library,

The grib2 iplib supports rotated lat-lon grids, so supporting 32768 and 32769 means you have to map the NCEP definition of the rotation to the the WMO definition of the rotation. Probably straight forward if you knowledgeable. However, it is outside of my expertise.

Wesley

On Fri, Mar 1, 2024 at 12:39 PM Alex Richert @.***> wrote:

@webisu https://github.com/webisu @edwardhartnett https://github.com/edwardhartnett would you be able to point me to the missing routines? I seem to recall Kyle's merge of ip2 into ip was fairly comprehensive, but I may be missing something.

— Reply to this email directly, view it on GitHub https://github.com/NOAA-EMC/NCEPLIBS-ip/issues/211#issuecomment-1973611546, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEIB7ZTDFDFDKCX5KAA7DP3YWC4LVAVCNFSM6AAAAABAZ3GUD2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNZTGYYTCNJUGY . You are receiving this because you were mentioned.Message ID: @.***>

AlexanderRichert-NOAA commented 7 months ago

Is there a file (or couple files if there are two grid numbers to test) and a wgrib2 command+expected result that I can test with? I'm assuming this will involve the -new_grid rot-ll... option?

I'll have to refresh myself on grib standards/grid definitions/etc. but I'm hoping the consolidated ip code will help here.

AlexanderRichert-NOAA commented 7 months ago

@webisu if I've understood the requirement correctly, the idea is to be able to provide the ip library directly with GDT value 32768 or 32769. If that is the case, please see #231 and see if that looks reasonable. In short, the current ip library supports those grids for grib2, but uses GDT number 1 for both (which I'm not sure I understand), then infers E vs. B based on some offsets in the GDT data. The PR would allow it take the 3276x values directly. If that seems like the right idea I can add unit testing and push the changes.

webisu commented 7 months ago

Alex, You have the requirement right. When you have the GDT == 1, then you are supporting the WMO definition of the rotated lat-lon grid. The grib2 version of iplibs supports rotated lat-lon grids from the regional FV3, Canadian regional, and some of the European regional models. (I am not aware of any that isn't supported but I didn't do a search.) I don't know whether the models are providing a full or staggered grid.

The NAM, RRFS and perhaps the hurricane model are using GDT 32768 or

32769. I really don't pay attention to the regional models. Dusan added the code to support these models. He is a regional guy and understands what a "rotate latitude/longitude (Arakawa Non-E Staggered grid)" means.

Wesley

On Fri, Mar 1, 2024 at 6:53 PM Alex Richert @.***> wrote:

@webisu https://github.com/webisu if I've understood the requirement correctly, the idea is to be able to provide the ip library directly with GDT value 32768 or 32769. If that is the case, please see #231 https://github.com/NOAA-EMC/NCEPLIBS-ip/pull/231 and see if that looks reasonable. In short, the current ip library supports those grids for grib2, but uses GDT number 1 for both (which I'm not sure I understand), then infers E vs. B based on some offsets in the GDT data. The PR would allow it take the 3276x values directly. If that seems like the right idea I can add unit testing and push the changes.

— Reply to this email directly, view it on GitHub https://github.com/NOAA-EMC/NCEPLIBS-ip/issues/211#issuecomment-1974095566, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEIB7ZWJMAHCZZQ3NOJJLQ3YWEIF3AVCNFSM6AAAAABAZ3GUD2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNZUGA4TKNJWGY . You are receiving this because you were mentioned.Message ID: @.***>