Revised calculation of wind GUST for UFS applications

[wzzheng90]

This PR is to address the issue "Gust wind weaker than mean wind in GFS" #636 (#636).

[WenMeng-NOAA]

@wzzheng90 I would suggest to add the following changes:

• In CALGUST.f: add change log with your name after line 18

• In CALGUST.f : line 78-79
  change
  CALL EXCH(UH(1,jsta_2l,L))
  CALL EXCH(VH(1,jsta_2l,L))
  into
  CALL EXCH(UH(ista_2l,jsta_2l,L))
  CALL EXCH(VH(ista_2l,jsta_2l,L))
  These changes are for 2D decomposition update and will have no impact on results you finished.

• In MDLFLD.f: line 3777
  Change
  IF(MODELNAME == 'GFS')THEN
  into
  IF(MODELNAME == 'GFS' .OR. MODELNAME == 'FV3R')THEN
  Also add a change log in this subroutine.

[wzzheng90]

@wzzheng90 I would suggest to add the following changes:

* In CALGUST.f: add change log with your name after line 18

* In CALGUST.f :  line 78-79
  change
  CALL EXCH(UH(1,jsta_2l,L))
  CALL EXCH(VH(1,jsta_2l,L))
  into
  CALL EXCH(UH(ista_2l,jsta_2l,L))
  CALL EXCH(VH(ista_2l,jsta_2l,L))
  These changes are for 2D decomposition update and will have no impact on results you finished.

* In MDLFLD.f: line 3777
  Change
  IF(MODELNAME  ==  'GFS')THEN
  into
  IF(MODELNAME == 'GFS' .OR. MODELNAME == 'FV3R')THEN
  Also add a change log in this subroutine.

I added your suggestions for 1) and 2). Thanks!!
However, I am reluctant to add the 3rd change. In fact, in "MDLFLD.f", Line 3777, it needs to compute PBL height based on Richardson number. GFS does use Richardson number to calculate the PBL height (PBLRI) so we can directly use the PBL height from GFS moel. But for other models such as HAFS or RRFS, it needs to re-calculate PBLRI. I did a test and the result is quite different if using the PBL height from HAFS or RRFS.

[WenMeng-NOAA]

@wzzheng90

However, I am reluctant to add the 3rd change. In fact, in "MDLFLD.f", Line 3777, it needs to compute PBL height based on Richardson number. GFS does use Richardson number to calculate the PBL height (PBLRI) so we can directly use the PBL height from GFS moel. But for other models such as HAFS or RRFS, it needs to re-calculate PBLRI. I did a test and the result is quite different if using the PBL height from HAFS or RRFS.

GFS output PBL height as "hpbl" in netcdf. The UPP read it as pblh (see here) The line 3777 is for assigning gfs's pblh as pblri. I checked rrfs and hafs model history files. The hpbl is included. So I might think the modelname "FV3R" should be added in line 3777 to keep all ufs models consistent.

[wzzheng90]

@wzzheng90

However, I am reluctant to add the 3rd change. In fact, in "MDLFLD.f", Line 3777, it needs to compute PBL height based on Richardson number. GFS does use Richardson number to calculate the PBL height (PBLRI) so we can directly use the PBL height from GFS moel. But for other models such as HAFS or RRFS, it needs to re-calculate PBLRI. I did a test and the result is quite different if using the PBL height from HAFS or RRFS.

GFS output PBL height as "hpbl" in netcdf. The UPP read it as pblh (see here) The line 3777 is for assigning gfs's pblh as pblri. I checked rrfs and hafs model history files. The hpbl is included. So I might think the modelname "FV3R" should be added in line 3777 to keep all ufs models consistent.

Maybe I did not explain it clearly. My point is that the PBL height from RRFS or HAFS model would not be derived based on Richardson number (different PBL schemes), so here it needs to re-calculate. I am not clear why here it needs Ri-based PBL height.
So this would be discussed later. Anyway, let's include "FV3R" at present. Thanks.

[WenMeng-NOAA]

@wzzheng90

However, I am reluctant to add the 3rd change. In fact, in "MDLFLD.f", Line 3777, it needs to compute PBL height based on Richardson number. GFS does use Richardson number to calculate the PBL height (PBLRI) so we can directly use the PBL height from GFS moel. But for other models such as HAFS or RRFS, it needs to re-calculate PBLRI. I did a test and the result is quite different if using the PBL height from HAFS or RRFS.

GFS output PBL height as "hpbl" in netcdf. The UPP read it as pblh (see here) The line 3777 is for assigning gfs's pblh as pblri. I checked rrfs and hafs model history files. The hpbl is included. So I might think the modelname "FV3R" should be added in line 3777 to keep all ufs models consistent.

Maybe I did not explain it clearly. My point is that the PBL height from RRFS or HAFS model would not be derived based on Richardson number (different PBL schemes), so here it needs to re-calculate. I am not clear why here it needs Ri-based PBL height. So this would be discussed later. Anyway, let's include "FV3R" at present. Thanks.

@wzzheng90 Are you meaning at model side, global FV3 and regional FV3 use different schemes to calculate PBL height? In that case, there is no need to add change in line 3777.

[wzzheng90]

@WenMeng-NOAA yes, using different PBL schemes. However, it may need to evaluate the results against the Obs to see which one is better in the future.

[WenMeng-NOAA]

The UPP RT tests were completed on WCOSS2, Hera and Orion. The baselines for gfs and hafs and rrfs will be recreated due to gust calculation update.

gfs: 
    master--
   14:21345016:GUST:surface:rpn_corr=0.996106:rpn_rms=0.706175
hafs: 
    HURPRS--
   3:2977237:GUST:surface:rpn_corr=0.994846:rpn_rms=0.556539
rrfs:  
    NATLEV--
   1286:5456532867:GUST:surface:rpn_corr=0.99558:rpn_rms=0.698341
    PRSLEV--
    51:293497035:GUST:surface:rpn_corr=0.99558:rpn_rms=0.698341   

[WenMeng-NOAA]

I will proceed to merging this PR and skip the CI failure from intel build. That failure have no impact on UPP performance on WCOSS2 and other HPC platforms. The fix could be implemented later.