Note, the following instructions are only valid for the 2018 CESM2 release. If you wish to run the dynamical core configuration prior to the CESM2 release, please contact Isla Simpson [ firstname.lastname@example.org ]
The following describes the Dry Adiabatic Baroclinic Instability (DABI) test case for the CAM dynamical core. This test case is documented in Polvani, Scott and Thomas (2004), MWR, 132, 2539--2552. It aims to produce a converged solution of the dry, adiabatic, primitive equations by running a short integration (12 days) of a slightly perturbed, baroclinically unstable, midlatitude jet. The initial conditions for this test case consist of a simple zonal flow and a localized temperature perturbation in the mid-latitudes, each of which are specified analytically.
The initial condition files for this test case are provided for the Eulerian spectral-transform dynamical core at three default resolutions within the CESM release (T42L30, T85L30 and T85L60). The T42L30 case is faster to run, but does not produce a result that is quite at the converged solution; this case may be desirable for those who wish to run a quick test to ensure the dynamical core is working well. The other two resolutions take slightly longer to run, but produce the converged solution. These initial condition datasets were generated with NCL codes that can be downloaded here. Users who wish to move on to generate their own initial conditions may find it helpful to use these as a starting point to ensure they have the correct initial condition file format.
Running the DABI test case
After downloading the latest CESM release, users may perform this test by following the procedures outlined below (Note, this configuration is only available from CESM2.0 onward). See the CESM user's guide for more infomation on creating and running new cases. The following describes how to run the DABI test case using a CESM release that is located in the directory $CESM.
Step 1: Create the DABI test case
This can be done using the create_newcase script located in the directory $CESM/cime/scripts/ e.g., for the T42L30 resolution
./create_newcase --compset FDABIP04 --res T42z30_T42_mg17 --case $CASEDIR
where the case directory ($CASEDIR) is specified by the user. To run the T85L30 or T85L60 resolutions, replace T42z30_T42_mg17 with T85z30_T85_mg17 or T85z60_T85_mg17 respectively.
Step 2: Configure the DABI test case
The FDABIP04 compset ensures that the diffusion settings for this test case are set up automatically. This includes the second-order diffusion mechanism with the resolution-independent diffusion coefficient 7x10^5m^2/s as specified in Polvani et al (2004). Note that the FDABIP04 test case only runs out of the box with the second-order diffusion for the CAM Eulerian spectral-transform dynamical core. In addition, the Eulerian transform dynamical core uses an Asselin time filter with the default coefficient 0.06 to ensure numerical stability and a total energy fixer is activated at each time step. The default length of the simulation is set to 5 days, so in order to perform the 12 day test case, the following command must be invoked
./xmlchange --file env_run.xml --id STOP_N --val 12
Step 3: Set up and build the DABI test case
From within $CASEDIR run
Step 4: Run the DABI test case
Step 5: Validate the DABI test case output
A number of fields can be used to check that the simulation has produced the converged solution. Two such fields are shown below for T85L30 (see also Figures 4 and 9 of Polvani, Scott and Thomas (2004)). Figure 1 shows the relative vorticity on the 0.975 sigma level and Figure 2 shows the pressure vertical velocity at sigma=0.5 and 45N, both at midnight of day 12 of the integration. These plots can be reproduced using an ncl script that can be downloaded here. If the dynamical core is set up correctly, these plots should be reproduced to a high degree of accuracy. Note that the FDABIP04 test case only runs out of the box with the second order diffusion specified by Polvani et al (2004) for the Eulerian spectral-transform dynamical core. As such, the exact solution below will not be obtained with the other dynamical cores unless the diffusion of Polvani et al (2004) is specified by the user.