Update calculation of time-averaged radiation variables #874
Conversation
…moother curves between radiation calls and diagnostic resets.
LarissaReames-NOAA
changed the title
|
@LarissaReames-NOAA Will this have any impact on exported SW fields ( |
This should only apply to history file diagnostic fields with time_avg_kind = "rad_sw", "rad_lw", or "rad_swlw_min" in ccpp/driver/GFS_diagnostics.F90 |
|
@junwang-noaa Who could we have review this PR? |
|
|
||
| if (time_int == dt_atmos ) then | ||
| var2(i,j) = Diag(idx)%data(nb)%var2(ix)*lcnvfac | ||
| hist%rad_accum(idx,ii,jj) = Diag(idx)%data(nb)%var2(ix) |
There was a problem hiding this comment.
The computation of accumulated radiation or physics stays in CCPP physics (fluxr), I suggest making the accumulation related change in ccpp, not in the IO file (fv3atm_history_io.F90). We only apply time scale or mask update in the fv3atm_history_io.F90.
There was a problem hiding this comment.
I looked the CCPP code, and see that the fluxr is accumulated every radiation hour (3600s) (multiplied by coszdg(i) / coszen(i)) in the code:
physics/physics/Interstitials/UFS_SCM_NEPTUNE/GFS_rrtmg_post.F90
so we need modify the bucket weighting in order to calculate the dswrf_ave and ulwrf correctly
Description
This PR fixes incorrect diagnostic output for fhzero-averaged radiation output variables. Presently, there is a significant "sawtooth" appearance to these fields (such as dswrf_ave and ulwrf_avetoa) when we expect smoother lines with jumps only at fhzero diagnostic reset times. Here is an example of the present (OLD) result and the result from the current PR (FIX) from a 24-hour S2S regression test forecast.
Issue(s) addressed
Testing
Ran full regression test suite on Hera. See log files here: /scratch1/NAGAPE/hpc-wof1/lreames/ufs-weather-model-38a29a6/tests/logs/log_hera
All regression tests fail as sfcf*.nc files are not identical. New baselines will be necessary.
Dependencies
N/A. All changes are contained in fv3atm.
Requirements before merging