v12 SCM is broken with RRTMGP

[mathomp4]

Earlier today, @atrayano discovered a bug in the v12 SCM setup (namely NCAR_NRDG was not detemplated). After fixing that locally (setting it to 16 as the SCM has a gwd_internal_rst), I tried running and:

 Integer*4 Resource Parameter: INTERPOLATE_OCEAN_ICE_CURRENTS:0, (default value)
 AGCM Date: 1997/06/18  Time: 23:45:00  Throughput(days/day)[Avg Tot Run]:         11.5         11.5        418.9  TimeRemaining(Est) 014:38:16   10.7% :  16.1% Mem Comm:Used
 Integer*4 Resource Parameter: RADSW_BINARY_CLOUDS:0, (default value)
 Character Resource Parameter: RRTMGP_GAS_SW:ExtData/g5gcm/radiation/RRTMGP/v1.8/rrtmgp-gas-sw-g112.nc
  A   1.000000000000000E-002  fractional increase of ptop was insufficient
  RRTMGP, GEOS-5 top (Pa)   1.00518357446300       0.967194020748138
pe=00000 FAIL at line=04979    GEOS_SolarGridComp.F90                   <RRTMGP Error: Model top too high for RRTMGP>
pe=00000 FAIL at line=03414    GEOS_SolarGridComp.F90                   <status=1>
pe=00000 FAIL at line=01862    MAPL_Generic.F90                         <status=1>
RADIATIONRun                                   810
PHYSICS::Run                                  2896
AGCMRun                                       2569
...

This is at least true in arm_97jul, but maybe all of them? I'll test and see.

[mathomp4]

The lines above refer to a pressure top kludge¹ via @dr0cloud:

      ! pmn: pressure KLUGE
      ! Because currently k_dist%press_ref_min ~ 1.005 > GEOS-5 ptop of 1.0 Pa.
      ! Find better solution, perhaps getting AER to add a higher top.
      press_ref_min = k_dist%get_press_min()
      ptop = minval(p_lev(:,1))
      if (press_ref_min > ptop) then
        ! allow a small increase of ptop
        if (press_ref_min - ptop <= ptop * ptop_increase_OK_fraction) then
          where (p_lev(:,1) < press_ref_min) p_lev(:,1) = press_ref_min
          ! make sure no pressure ordering issues were created
          _ASSERT(all(p_lev(:,1) < p_lay(:,1)), 'pressure kluge causes misordering')
        else
          write(*,*) ' A ', ptop_increase_OK_fraction, &
                       ' fractional increase of ptop was insufficient'
          write(*,*) ' RRTMGP, GEOS-5 top (Pa)', press_ref_min, ptop
          TEST_('Model top too high for RRTMGP')
        endif
      endif

I see nothing in the latest RC release notes from @RobertPincus to say anything has changed in RRTMGP, so this is most likely a GEOS issue.

I mean, @wmputman has run literally hundreds of thousands of columns with V12 at L181, so I mean, this should be good. To me it's saying the restarts we are using are bad now and should be resampled/regridded. I'll consult with @narnold1 and @bena-nasa on how we do this...

Footnotes

1. This will be an issue if we move forward with RRTMGP + L186 as that goes waaaaay up higher I think to 25 mPa (0.00025 hPa). ↩

[RobertPincus]

@mathomp4 Without looking at your code too carefully note that the restriction on pressure applies only to play not to plev since the latter is used only to compute the number of molecules per area in a layer.

You have to ask AER about extending to higher pressures. That's a data issue - the code is agnostic.