Problem initializing 5.1.27 w/ modal aerosols (CAM5) + chem -none

[zarzycki@umich_edu]

In the past we have run test cases (on aquaplanets) w/ CAM5 physics without specifying aerosol props/distributions by just using the chem -none flag during build. This worked at 5.1.09 (the last build we were using).

With the new switch over to 5.1.27, it appears chem -none switches over to the modal framework for aerosol initialization and gives me this error when running the SE dycore w/ CAM5 physics (and chem -none) on Bluefire with the aquaplanet use_case:

0: bulk_props_init: finished for SULFATE
0: (GETFIL): attempting to find local file ocpho_rrtmg_c101112.nc
0: (GETFIL): using /fs/cgd/csm/inputdata/atm/cam/physprops/ocpho_rrtmg_c101112.nc
1: Opened existing file /fs/cgd/csm/inputdata/atm/cam/physprops/ocpho_rrtmg_c101112.nc 62
0: bulk_props_init: finished for OCPHO
0: (GETFIL): attempting to find local file ocphi_rrtmg_c100508.nc
0: (GETFIL): using /fs/cgd/csm/inputdata/atm/cam/physprops/ocphi_rrtmg_c100508.nc
1: Opened existing file /fs/cgd/csm/inputdata/atm/cam/physprops/ocphi_rrtmg_c100508.nc 62
0: bulk_props_init: finished for OCPHI
0: (GETFIL): attempting to find local file bcpho_rrtmg_c100508.nc
0: (GETFIL): using /fs/cgd/csm/inputdata/atm/cam/physprops/bcpho_rrtmg_c100508.nc
1: Opened existing file /fs/cgd/csm/inputdata/atm/cam/physprops/bcpho_rrtmg_c100508.nc 62
0: bulk_props_init: finished for BCPHO
0: (GETFIL): attempting to find local file dust4_rrtmg_c090521.nc
0: (GETFIL): using /fs/cgd/csm/inputdata/atm/cam/physprops/dust4_rrtmg_c090521.nc
1: Opened existing file /fs/cgd/csm/inputdata/atm/cam/physprops/dust4_rrtmg_c090521.nc 62
0: bulk_props_init: finished for DUST4
0: (GETFIL): attempting to find local file ssam_rrtmg_c100508.nc
0: (GETFIL): using /fs/cgd/csm/inputdata/atm/cam/physprops/ssam_rrtmg_c100508.nc
1: Opened existing file /fs/cgd/csm/inputdata/atm/cam/physprops/ssam_rrtmg_c100508.nc 62
0: bulk_props_init: finished for SSAM
0:
0:rad_cnst_init: checking for radiative constituents
0: ENDRUN:init_modes ERROR: cannot find physics buffer field num_c1
2: ENDRUN:init_modes ERROR: cannot find physics buffer field num_c1
3: ENDRUN:init_modes ERROR: cannot find physics buffer field num_c1
1: ENDRUN:init_modes ERROR: cannot find physics buffer field num_c1
3:

The model builds/runs fine with CAM4 physics (bulk) and will also begin running if I edit build-namelist to force chem -none to -> bulk instead of modal (although this crashes ~2 days in with issues in the shallow convection package which are likely traced back to me initializing CAM5 w/o modal aerosols).

Is there any undocumented changes to the namelist flags that I should know about to eliminate the above run-time error? Are there just missing data files in the CESM input repo since the version is new? (/fs/cgd/csm/inputdata/)

Thanks.

 

[eaton]

In cam5_1_09 the default for "-chem none" was to use prescribed bulk aerosols. We have been adding functionality to eventually allow the use of prescribed modal aerosols, and at cam5_1_12 the default was switched to assume that for "-chem none", the prescribed aerosols are bulk if the cam4 physics is being used, or modal if the cam5 physics is used. However, the use of prescribed modal aerosols has not yet been fully implemented (so the switch was perhaps premature). The way to specify that you want the prescribed bulk aerosols is to add prescribed_aero_model='bulk' to the namelist.

 

[zarzycki@umich_edu]

In cam5_1_09 the default for "-chem none" was to use prescribed bulk aerosols. We have been adding functionality to eventually allow the use of prescribed modal aerosols, and at cam5_1_12 the default was switched to assume that for "-chem none", the prescribed aerosols are bulk if the cam4 physics is being used, or modal if the cam5 physics is used. However, the use of prescribed modal aerosols has not yet been fully implemented (so the switch was perhaps premature). The way to specify that you want the prescribed bulk aerosols is to add prescribed_aero_model='bulk' to the namelist.

Thanks, Brian, that's kind of what I figured.

I actually did this last week and successfully initialized the model but got a crash in the shallow convection routine a few days post-initialization -- but I'll have to do a little digging and see if I can figure out what the problem was there.

 

[eaton]

I've just become aware of a bug in this configuration which was introduced at cam5_1_14. The problem is in microp_aero_run which has ptend as an intent(out) argument. But when BAM aerosols are being used the ptend argument never gets initialized. The fix is to add a "call physics_ptend_init(ptend)" statement to the branch of the conditional that contains the call to ndrop_bam_run. (When modal aerosols are active ptend is set by the call to dropmixnuc.) Also the physics_ptend_init method will need to be added to the use association statement for physics_types.