[Ncep.list.emc.fv3gfs_tickets] FV3GFS Ticket #1: Update global_chgres program for FV3 core

[FV3GFS Trac Ticket]

#1: Update global_chgres program for FV3 core
-----------------------------+----------------------------
  Reporter:  george.gayno@…  |      Owner:  fanglin.yang@…
      Type:  task            |     Status:  assigned
  Priority:  major           |  Milestone:
 Component:  component1      |    Version:
Resolution:                  |   Keywords:  global_chgres
-----------------------------+----------------------------

Comment (by george.gayno@…):

 Replying to [comment:44 george.gayno@…]:
 > Replying to [comment:43 george.gayno@…]:
 > > As Fanglin noted in comment:42, the output from the updated
 `global_chgres` developed under this ticket is not bit identical to that
 from the `global_chgres` as modified by GFDL. There are several reasons
 for this:
 > >
 > > * GFDL modified the May 2016 GFS version (13.0.5) of `global_chgres`
 for FV3.  The new `global_chgres` developed under this ticket added GFDL's
 changes to the
 [https://svnemc.ncep.noaa.gov/trac/gfs/browser/branches/gfs_q3fy17?rev=84343&order=date&desc=1
 Q3FY17 version] (v14.1.0).  Numerous changes were made to global_chgres
 for v14.1.0.
 > >
 > > * The
 [https://svnemc.ncep.noaa.gov/trac/fv3gfs/browser/branches/chgres_fv3_2/global_shared.v15.0.0/sorc/global_chgres.gfdl.fd?rev=87746
 GFDL version] interpolated winds to the FV3 grid using
 [https://svnemc.ncep.noaa.gov/trac/fv3gfs/browser/branches/chgres_fv3_2/global_shared.v15.0.0/sorc/global_chgres.gfdl.fd/chgres.f?rev=87746#L4327
 IPOLATES], which is for scalar fields.  The updated version uses
 [https://svnemc.ncep.noaa.gov/trac/fv3gfs/browser/branches/chgres_fv3_2/global_shared.v15.0.0/sorc/global_chgres.fd/chgres_utils.f90?rev=87955#L122
 IPOLATEV], which is for vectors. IPOLATEV moves the vectors along a great
 circle route to the target point. This results in differences near the
 poles and very minor differences elsewhere.
 > >
 > > * The GFDL version did not set soil type correctly.  It looked like it
 was interpolated using the bi-linear method.  That is not correct for a
 categorical field.  The new `global_chgres` uses nearest neighbor.
 >
 > I attached a plot of the soil type:
 https://svnemc.ncep.noaa.gov/trac/fv3gfs/attachment/ticket/1/soil.gif

 I know why the soil type was not correct.  I was pointing to the wrong
 global soil type file.  Recently, the grib 1 parameter number was changed
 from 230 to 224.  The GFDL code expected the old number.  sfcsub.f sets
 the interpolation method based on the parameter number.
 >
 > >
 > > * The
 [https://svnemc.ncep.noaa.gov/trac/fv3gfs/browser/branches/chgres_fv3_2/global_shared.v15.0.0/sorc/global_chgres.fd/sfcsub.f?rev=87955
 sfcsub.f] routine, used to set several static fields on the FV3 grid, is
 not the same between the
 [https://svnemc.ncep.noaa.gov/trac/fv3gfs/browser/branches/chgres_fv3_2/global_shared.v15.0.0/sorc/global_chgres.gfdl.fd/sfcsub.f?rev=87746
 GFDL] and updated versions.  For example, the z0 for vegetation type 12 is
 3.5 cm in the GFDL version and 12.5 cm in the updated version.
 > >
 > > * GFDL added their
 [https://svnemc.ncep.noaa.gov/trac/fv3gfs/browser/branches/chgres_fv3_2/global_shared.v15.0.0/sorc/global_chgres.gfdl.fd/surface_chgres.f?rev=87746#L3096
 own logic to search] for valid data at FV3 water points not located near
 input GFS water points (i.e., isolated lakes).  The updated version uses
 the search logic in IPOLATES.  The GFDL search radius is much larger,
 which results in differences in SST and ice along coastlines and inland
 water bodies.
 > >
 > > In summary, there are more differences in the surface files than the
 sigma files.

--
Ticket URL: <https://svnemc.ncep.noaa.gov/trac/fv3gfs/ticket/1#comment:45>
fv3gfs <https://svnemc.ncep.noaa.gov/trac/fv3gfs>
NGGPS FV3GFS Development