combine `history.nc` and `mpasout.nc` or only output `mpasout.nc` at forecast hour 1 (or and hour 2)

[guoqing-noaa]

Currently, in rrfs-workflow, MPAS-Model generates two set of output files:

1. history.nc (with accompanying diag.nc) for post-processing through MPASSIT, UPP
2. mpasout.nc (with accompanying invariant.nc) for data assimilation and reinitialization of next cycle MPAS-Model forecasts

For high resolution, and/or large domain, and/or long forecasts, this will take lots of unnecessary duplicate disk spaces.

There may be two routes to address this issue:

1. combine history.nc and mpasout.nc into one file (similar to the wrfout file in the WRF era)
   A preliminary comparison of fields in both files will be provided in the following post.

2. only output mpasout.nc at forecast hour 1 (or and hour 2) and output history.nc for all forecast hours
   It looks like at this moment we cannot achieve this. No such an option is available in the steam output_interval setting.
   Modifications to MPAS-Model source codes may be needed
   I was trying to find where we should make such a change. But I did not have a clear clue yet. I only found something so far like filename_interval = interval_out2 in framework/xml_stream_parser.c or the following part in core_atmosphere/physics/mpas_atmphys_manager.F

 if(trim(stream_interval) /= 'none') then
    call mpas_set_timeInterval(alarmTimeStep,timeString=stream_interval,ierr=ierr)

A few other thoughts:

• history.nc contains lots of static fields, if we can only output variant fields into history.nc and read static fields from invariant.nc or static.nc, it may save us a decent amount of disk spaces.

• not sure whether history.nc files are enough for JEDI to compute hofx (i.e. the model counter parts at the observation locations). If not, we may still need to generate long forecast mpasout.nc files

Tag @clark-evans @hu5970 @SamuelTrahanNOAA for awareness.

[guoqing-noaa]

history.nc has 121 fields (sorted) based on the output stream in Registry.xml:

Time,acrunoff,acsnow,angleEdge,areaCell,areaTriangle,cd,cda,cellsOnCell,cellsOnEdge,cellsOnVertex,ck,cka,cuprec,dcEdge,depv_dt_bl,depv_dt_cu,depv_dt_diab,depv_dt_diab_pv,depv_dt_fric,depv_dt_fric_pv,depv_dt_lw,depv_dt_mix,depv_dt_mp,depv_dt_sw,divergence,dtheta_dt_mp,dvEdge,dzs,edgesOnCell,edgesOnEdge,edgesOnVertex,ertel_pv,fzm,fzp,glw,grdflx,gsw,hfx,hpbl,iLev_DT,i_rainc,i_rainnc,indexToCellID,indexToEdgeID,indexToVertexID,initial_time,isice_lu,isltyp,iswater_lu,ivgtyp,ke,kiteAreasOnVertex,kpbl,lai,latCell,latEdge,latVertex,lh,lonCell,lonEdge,lonVertex,meshDensity,mminlu,nEdgesOnCell,nEdgesOnEdge,ozmixm,pin,precipw,pressure,q2,qfx,rainc,rainnc,relhum,rho,rhosnf,scalars,sfc_albedo,sh2o,skintemp,smois,snow,snowfallac,snowh,soilt1,sst,surface_pressure,t2m,th2m,theta,theta_pv,tslb,u,u10,uReconstructMeridional,uReconstructZonal,u_pv,v10,v_pv,vegfra,verticesOnCell,verticesOnEdge,vort_pv,vorticity,w,weightsOnEdge,xCell,xEdge,xVertex,xice,xland,xtime,yCell,yEdge,yVertex,zCell,zEdge,zVertex,zgrid,zz

mpasout.nc has 295 fields (sorted) based on the jedi_da stream in Registry.xml (note: there are efforts to remove some unnecessary fields in mpasout.nc. It is estimated that we may remove at least 1/3 of current list)

Time,aclwdnb,aclwdnbc,aclwdnt,aclwdntc,aclwupb,aclwupbc,aclwupt,aclwuptc,acrunoff,acsnom,acsnow,acswdnb,acswdnbc,acswdnt,acswdntc,acswupb,acswupbc,acswupt,acswuptc,albedo12m,br,canwat,cd,cda,ch,chklowq,chs,chs2,circulation,ck,cka,cldfrac,cldfrac_bl,cldfrac_cu,cov,cpm,cqs,cqs2,cubot,cuprec,cutop,delta,det_qv,det_thl,dew,diag_physics_noahmp,dtaux3d,dtauy3d,dusfcg,dvsfcg,dzs,edmf_a,edmf_ent,edmf_qc,edmf_qt,edmf_thl,edmf_w,el_pbl,evapprod,exner,exner_base,fh,flhc,flqc,fm,glw,graupelnc,graupelncv,grdflx,greenfrac,gsw,gz1oz0,h_oml,h_oml_initial,hfx,hpbl,hu_oml,hv_oml,i_aclwdnb,i_aclwdnbc,i_aclwdnt,i_aclwdntc,i_aclwupb,i_aclwupbc,i_aclwupt,i_aclwuptc,i_acswdnb,i_acswdnbc,i_acswdnt,i_acswdntc,i_acswupb,i_acswupbc,i_acswupt,i_acswuptc,i_rainc,i_rainnc,initial_time,isice_lu,isltyp,iswater_lu,ivgtyp,k22_shallow,kbcon_shallow,kpbl,ktop_deep,ktop_shallow,lai,lai12m,landmask,landusef,lh,mavail,mminlu,mol,mu_c,nca,ndays_accum,nifa2d,noahres,nsteps_accum,nt_c,nwfa2d,o3clim,o3vmr,potevp,precipfr,precipw,pressure_base,pressure_p,psih,psim,q2,qc_bl,qc_cu,qcg,qfx,qgh,qi_bl,qi_cu,qke,qke_adv,qsfc,qsg,qsq,qvg,rainc,raincv,rainnc,rainncv,rainprod,re_cloud,re_ice,re_snow,refl10cm,refl10cm_cu,refl10cm_max,regime,relhum,rho,rho_base,rho_p,rho_zz,rhosnf,rmol,rncblten,rncmpten,rniblten,rnifablten,rnifampten,rnimpten,rnrmpten,rnwfablten,rnwfampten,rqcblten,rqccuten,rqcmpten,rqgmpten,rqiblten,rqicuten,rqimpten,rqrcuten,rqrmpten,rqsblten,rqscuten,rqsmpten,rqvblten,rqvcuten,rqvdynten,rqvmpten,rt_diabatic_tend,rthblten,rthcuten,rthdynten,rtheta_base,rtheta_p,rthmpten,rthratenlw,rthratensw,ru,ru_p,rubldiff,rublten,rucuten,rvbldiff,rvblten,rvcuten,rw,rw_p,scalars,seaice,sfc_albbck,sfc_albedo,sfc_emibck,sfc_emiss,sfcrunoff,sh2o,sh3d,shdmax,shdmin,sigma_deep,skintemp,sm3d,smcrel,smfr3d,smois,smstav,smstot,snoalb,snopcx,snotime,snow,snowc,snowfallac,snowh,snownc,snowncv,soilf,soilt1,sr,sst,sstsk,sstsk_dtc,sstsk_dtw,sub3d_rqvcuten,sub3d_rthcuten,sub3d_rucuten,sub3d_rvcuten,sub_qv,sub_thl,surface_pressure,t2m,t_oml,t_oml_200m_initial,t_oml_initial,tday_accum,tend_sfc_pressure,ter,th2m,thc,theta,theta_base,theta_m,tke_pbl,tlag,tmn,tslb,tsnav,tsq,tyear_accum,tyear_mean,u,u10,uReconstructMeridional,uReconstructZonal,udrunoff,ust,ustm,v10,vcpool,vegfra,w,w0avg,wspd,wstar,xice,xicem,xland,xmb_shallow,xmb_total,xtime,z0,znt,zol,zs

There are 63 fields in history.nc but not in mpasout.nc: (sorted)

angleEdge,areaCell,areaTriangle,cellsOnCell,cellsOnEdge,cellsOnVertex,dcEdge,depv_dt_bl,depv_dt_cu,depv_dt_diab,depv_dt_diab_pv,depv_dt_fric,depv_dt_fric_pv,depv_dt_lw,depv_dt_mix,depv_dt_mp,depv_dt_sw,divergence,dtheta_dt_mp,dvEdge,edgesOnCell,edgesOnEdge,edgesOnVertex,ertel_pv,fzm,fzp,iLev_DT,indexToCellID,indexToEdgeID,indexToVertexID,ke,kiteAreasOnVertex,latCell,latEdge,latVertex,lonCell,lonEdge,lonVertex,meshDensity,nEdgesOnCell,nEdgesOnEdge,ozmixm,pin,pressure,theta_pv,u_pv,v_pv,verticesOnCell,verticesOnEdge,vort_pv,vorticity,weightsOnEdge,xCell,xEdge,xVertex,yCell,yEdge,yVertex,zCell,zEdge,zVertex,zgrid,zz

40 of the above 63 fields are in invariant.nc. so we have the following 23 fields left from history.nc:

depv_dt_bl,depv_dt_cu,depv_dt_diab,depv_dt_diab_pv,depv_dt_fric,depv_dt_fric_pv,depv_dt_lw,depv_dt_mix,depv_dt_mp,depv_dt_sw,divergence,dtheta_dt_mp,ertel_pv,iLev_DT,ke,ozmixm,pin,pressure,theta_pv,u_pv,v_pv,vort_pv,vorticity

17 of the above 23 fields are Ertel PV diagnositcs:

                        <!-- Begin Ertel PV diagnostics defined in diagnostics/Registry_pv.xml -->
                        <var name="ertel_pv"/>
                        <var name="u_pv"/>
                        <var name="v_pv"/>
                        <var name="theta_pv"/>
                        <var name="vort_pv"/>
                        <var name="iLev_DT"/>
#ifdef DO_PHYSICS
                        <var name="depv_dt_lw"/>
                        <var name="depv_dt_sw"/>
                        <var name="depv_dt_bl"/>
                        <var name="depv_dt_cu"/>
                        <var name="depv_dt_mix"/>
                        <var name="dtheta_dt_mp"/>
                        <var name="depv_dt_mp"/>
                        <var name="depv_dt_diab"/>
                        <var name="depv_dt_fric"/>
                        <var name="depv_dt_diab_pv"/>
                        <var name="depv_dt_fric_pv"/>
                        <!-- End Ertel PV diagnostics defined in diagnostics/Registry_pv.xml -->

@clark-evans I am curious about these 17 fields. Since they are diagnostics, should they be put in diag.nc instead of history.nc?

so in the end we have 6 fields which only exist inhistory.nc and we may either add them to diag.nc or mpasout.nc:

divergence,ke,ozmixm,pin,pressure,vorticity

[clark-evans]

Hi @guoqing-noaa -- regarding the PV diagnostics, for some reason they are defined in src/core_atmosphere/Registry.xml as being part of the output (i.e., history) stream. This is different from the isobaric and convective diagnostics, which are defined in src/core_atmosphere/Registry.xml as being part of the diagnostics stream.

All three sets of diagnostics have unique src/core_atmosphere/diagnostics/Registry_***.xml files to define the variable structure and associated variables. They all share the diag variable structure. Thus, I would think that the PV diagnostics listed in src/core_atmosphere/Registry.xml could be moved down into the diagnostics stream, though that would need testing. Alternatively, could we just suppress their output? As you note, they are only diagnostics (such that DA should not update them), and we do not need them for operations. Any developmental use is likely to be very limited.

[guoqing-noaa]

@clark-evans Thanks for the helpful information. I agree that lots of variables in the history stream looks like not needed for operations.

For us, mpassit only processes the following 44 variables into the final product grib2 files:
surface_pressure,xland,skintemp,snow,snowh,sst,hfx,lh,sfc_albedo,hpbl,ivgtyp,isltyp,vegfra,shdmin,shdmax,canwat,snowc,snownc,graupelnc,sfcrunoff,zgrid,w,theta,uReconstructZonal,uReconstructMeridional,qv,qc,qr,qi,qs,qg,nc,ni,nr,ng,volg,nwfa,nifa,pressure,rho,cldfrac_bl,tslb,smois,sh2o

[guoqing-noaa]

So in current situation, it looks like we only need to include pressure in mpasout.nc and then it can replace history.nc