Aqua Planet WACCM

[rddixon@wisc_edu]

Is it possible to run WACCM in aqua-planet mode?

I've set it up using:

configure -cam_bld $blddir
-cam_exedir $rundir
-cam_exe cam
-dyn fv
-hgrid 4x5
-phys cam4
-ocn aquaplanet
-ntasks 16
-chem waccm_mozart
-no-age_of_air_trc
-test -v -nosmp > &! configure out

and it builds properly, but when I try to create a namelist either it won't do it or it does but then crashes on the first timestep when the model is run. I'll keep doing some digging, and post some more detailed results, but was wondering if anyone else had tried this or has experience with it.

 

[rddixon@wisc_edu]

First I set up a namelist_options file with

stop_n = 1
stop_option = 'nmonths'
npr_yz = 8,2,2,8

Then I set up the namelist with

$cfgdir/build-namelist -s
-case $case
-runtype $runtype
-test
-config $blddir/config_cache.xml
-infile namelist_options > & namelist.out

which died at line 2572 and gave the following output in namelist.out

CAM build-namelist - ERROR: No default value found for ncdata
user defined attributes:
key=ape val=1
key=ic_ymd val=101

so I figured I had done a poor job of defining what ic_ymd was, so I tried two different corrections.

1) I added the tag -ignore_ic_date to the namelist, this allowed to the namelist to be built and the input data files to be found
2) I added the tag -use_case waccm_2000_cam4_lite, which is just like the waccm_2000_cam4 case except with the age of air tracers variables removed from the output

For the second case I ran the model with the namelists and it was able to initialize the model, it calculated some lightning NOx, successfully generated three history files, and then crashed:

****** ZM_CONV IENTROPY: Tmix did not converge ******

so somehow I was unable to invert the entropy equation, maybe the waccm physics is looking for the wrong boundary conditions at the surface and is incompatible with the aqua planet setting. is there another use_case that would be more appropriate for running a scenario like this?

 

[eaton]

The problem is most likely due to the initial state (from the initial file) being too far out of balance. build-namelist wasn't able to find a default for ncdata because there are no waccm initial files that have been spun up for an aquaplanet simulation. So you need to start from a waccm initial file that contains the effects of topography, and try to get it going in aquaplanet mode. Since build-namelist won't find a suitable dataset, probably the best thing is to force it to use the 4x5 file that's in the waccm_2000_cam4 use case file. Do this by setting it explicitly in your namelist_options file:

ncdata='/path/waccm_ic.nc'

The first thing to try to get the model running is to reduce the timestep. It may be necessary to try halving the timestep more than once.

Once the simulation has gotten on it's feet you can create a new initial file that can be used in future experiments, and return the timestep back to normal settings

 

[rddixon@wisc_edu]

Thanks for your help!

I was able to properly initialize the model by starting with the wa319_4x5_ref1.3.cam2.i.1953-01-01-00000.nc file and decreasing dtime to 400. It ran without error until the 20th day of the run, when it stopped because of an error with the QBO forcing:

qbo_timestep_init: Input qbo for date 0 sec 0 does not exceed model date 10120 sec 400 Stopping.

First off, I'm wondering if I should be forcing the QBO at all, since I am assuming the qbo forcings are for runs with normal topography. I spoke (briefly) to Dr. Gettelman at AGU last week and he raised concern that coupling WACCM with an aqua planet wouldn't handle the internal waves properly. Or perhaps I misunderstood his comment.

If I decided to continue with this run, my next thought is that I could change the qbo_forcing_file to qboseries_1953-2007_c081015.nc with hopes that it would match my initial file better and set qbo_cyclic = .false. Another possibility is to do away with QBO forcing completely by setting qbo_use_forcing = .false. but I'd worry that an important amount of physics would be lost in my run if the WACCM cannot handle the upward propagation of internal waves well enough to model the QBO.

Your thoughts are greatly appreciated.