Could single-column model be run through CESM?

[shanshans@uchicago_edu]

I am wondering if CAM single-column model could be run through CESM (e.g., CESM 1.0.3, CESM1.1.1). In the User's Guide to CAM5.1 (http://www.cesm.ucar.edu/models/cesm1.0/cam/docs/ug5_1/ug.html#config_scam), it is mentioned that "SCAM configuration" is available. Does this option is also available in CESM? Thank you!Shanshan 

 

[jet]

test

 

[jet]

SCAM is supported in CESM under the latest development tags. Included is one sample compset for theARM IOP case.  For the latest cesm tags you can get a listing of compsets by typing./create_newcase -list compsets | grep -i scamalias: FARM95C4                  longname: AR95_CAM4%SCAM_CLM40%SP_CICE%PRES_DOCN%DOM_RTM_SGLC_SWAVTo build the model on yellowstone you would type ./create_newcase -case FARM95C4.01 -mach yellowstone -res T42_T42 -compset FARM95C4 cd FARM95C4.01./cesm_setup./*.build./*.runIn earlier tags (like cesm1_1_1) for instance the compset name might be different and you might have to append-pts_lat=36.6 -pts_lon=262.5 to the create_newcase command. In the cesm1_1_1 tags the compset was called FARM95C5 and for that tagtyping create_newcase --list would have printed a list of all valid compsetsThe scam command for cesm1_1_1 should look like:./create_newcase -case FARM95C5 -mach yellowstone -res T42_T42 -compset FARM95C5 -pts_lat=36.6 -pts_lon=262.5./cesm_setup./*.build./*.run
For most cases running scam on a supercomputer is overkill.  It was developed to berun on smaller platforms like your laptop or local linux box.  If you have a supported compilerand the free netcdf libraries installed you should be able to run scam on that platform. You canmodify the cesm scripts or there is a sample run-scam.csh script that lists the steps needed torun CAM/SCAM without using the CESM scripts. jt

 

[shanshans@uchicago_edu]

Thank you very much! I will go ahead to take a try now.  Shanshan 

 

[shanshans@uchicago_edu]

Hi jet,The single-column model works well throguh create_newcase in CESM, but only for a short period of time. If I need to run longer (e.g., for 10 years), should I generate the required forcing data like these in the inputdata folder /inputdata/atm/cam/scam/iop ?  I found" -camiop Configure CAM to generate an IOP file that can be used to drive SCAM. This switch only works with the Eulerian dycore." Do you know where I can find more introduction to "-camiop"? Could "-camiop" be used through create_newcase in CESM or this should be run in CAM5.1 only through *.csh? 
Best regards,Shanshan 

 

[jet]

Hi Shanshan:

SCAM is setup to read data off of the IOP when the model date and time are greater than or equal to the date and time of an IOP time slice of data.  For standard IOP's like ARM, this forcing data is not provided on the model time step but every three hours.  Thus for standar IOP's SCAM only updates its large scale forcing every three hours.  It hold the forcing constant for all SCAM time steps that fall in between the 3hr boundaries.  Standard IOP files provide 3 hr forcing values for periods of days to weeks.  If you would like to run for longer periods you have a few options depending on the experiment you are running.  If for instance you are running the model to some type of equilibrium and may want to use a constant forcing over the length of your experiment.  In this case you could create an IOP file with just two time steps of forcing data on it that are separated by a period of time greater than what you will be running the model for.  If you want to run for at least 10 years with constant forcing you would create an IOP file with two identical slices of forcing data and set the date and times for those slices to be something more than 10 years apart (maybe 100 years).  The model will read in the first IOP time slice and then hold that forcing constant for the length of your 10 year run.  You can create the new IOP file using any number of netcdf routines that create and modify netcdf data like CDO (climate data operators) or Charlie Zender's NCO utilities or NCAR NCL.

You can also have CAM generate an IOP file by passing the -camiop option when configuring the model. When build-namelist is called after the model is configured and compiled it will see that CAM is configured for generating an IOP and will set up the namelist such that the h1 history file that is produced can be used as an IOP file. The CAM IOP file is just a normal CAM history file that has some special IOP fields on it.  You can change the averaging and output options for the IOP history file just as you would for any CAM history file.  By default the h1 history file is a global 3d file.  Writing a single h1 IOP file for a number of days or even weeks would create a file with a reasonable size but tryiing to output a single file with 10 years of global data on would generate a huge file that would be very difficult to work with.  Most likely you will have to set up the CAM namelist to generate a number of h1 files with a reasonable set of time slices on each file.  You will also need to decide what type of averaging will be needed for you experiment. (Does your experiment require 10 years of itime step data or will 3hr averages work, or maybe daily averages or monthly averages would do.) 

I haven't tried this with the CESM scripts but it shouldn't be too hard to accomplish. You could for example create a case using a Eulerian resolution and compset that would provide the type of forcing you would like to use and then modify the env_build.xml  (called env_conf.xml in newer versions) file to pass the -camiop option to configure and then add your special averaging and output frequency flags to a user_nl_cam file in your case directory.
SCAM is designed to fill in a background state with data taken from the initial condition and then overlay whatever forcing is available via the IOP file.  If its run with poor or no large scale advective forcing we usually see the column dry out pretty quickly and move to a state that isn't very realistic. jt

 

[shanshans@uchicago_edu]

Hi Jet,Thank you so much!Shanshan