CISM exercise II: run diagnostic test cases
In this exercise, we will run a few of CISM's higher-order test cases, which span a wide range of flow regimes. From the top level directory where you build the code, change to into the higher-order tests directory:
You will see the following list of subdirectories
dome/ # parabolic shaped dome with simple boundary conditions ismip-hom/ # ISMIP-HOM test suite ross/ # Ross ice shelf test case shelf/ # ice shelf test cases on simplified domains
along with a few other files (note that the comments after the hash marks above have been added here). While you may want to look over the tests/higher-order/README file at some point, most of the necessary information from that file is contained on this page. While you are welcome to explore any of the test cases on your own (most of them can be run by simply following the instructions in the README files within each subdirectory), for this exercise we will pick a few representative examples that can be run relatively quickly.
All of the test cases use the simple_glide executable, which is built from the simple_glide.F90 driver in example-drivers/simple_glide/src/. To run any of the tests, you will need to change into the respective subdirectory and either copy the simple_glide executable to that directory
cp ../../../example-drivers/simple_glide/src/simple_gide ./
or make a virtual link to the actual executable file
ln -s ../../../example-drivers/simple_glide/src/simple_glide ./
The Dome test case
This is a very simple test case, simulating the three-dimensional flow field within an isothermal, parabolic shaped dome with no-slip basal boundary conditions and zero flux lateral boundary conditions. You can execute the test with
The call to the python script first builds an input netCDF file in the output/ subdirectory and executes simple_glide. In general, whenever simple_glide is executed, it expects to be followed by the name of a text file with the ".config" extension. If such a file is not specified, you will usually see something like
Enter name of GLIDE configuration file to be read
Here, the python script knows to pass the included "dome.config" script to simple_glide on its own. While there are numerous default settings in the code, in general it will need a configuration file of some sort to specify various things like grid size and spacing, various solver options, boundary conditions, etc. A good way to get a feel for what these options are and what parts of the code they trigger is to look in the ".config" file, find an options (e.g. "evolution = 3"), and then "grep" for that option in the file glide_types.F90 in the libglide/ subdirectory.
As the code runs, you will see some output to the screen like
(dH/dt using incremental remapping) time = 0.0000000 Running Payne/Price higher-order dynamics solver iter # resid (L2 norm) target resid 1 223.257 0.100000E-03 2 223.150 0.100000E-03 3 216.909 0.100000E-03 4 203.843 0.100000E-03 5 180.486 0.100000E-03 6 149.276 0.100000E-03 7 116.333 0.100000E-03 ... 39 0.341224E-03 0.100000E-03 40 0.226718E-03 0.100000E-03 41 0.150639E-03 0.100000E-03 42 0.100090E-03 0.100000E-03 43 0.665039E-04 0.100000E-03
The output you see here is fairly standard. It tells us the following information
- Which solver we are using to evolve the ice thickness (if at all). Here, we see that we are using incremental remapping (which we will discuss further later on). However, looking in the ".config" file we see that the start and end times are identical, so no geometric evolution will take place; we are simply after a diagnostic solution here.
- The current time step we are solving for.
- The dynamics scheme we are using (here, the Blatter-Pattyn equations as formulated and solved by Payne and Price).
- the non-linear iteration number, the residual (the L2 norm of the vector r = Ax - b), and the target residual
Note that when the residual is less than or equal to the target residual, the nonlinear iterations are halted and we have a converged solution. Here it took 43 iterations to arrive at a converged solution.
You can look at the model output using any convenient netCDF file viewer. A python-based netCDF file viewer, viewNetCDF.py is included in top level of the tests/higher-order subdirectory. Another common viewer installed on many machines is NCVIEW. To examine the output file using NCVEW, type
Your output for the variable velnorm at level "0" (that is, sigma coordinate level 0, which is the upper surface of the ice sheet), should look something like what is shown in the figure labeled Dome test case. Admittedly, this is not a very exciting test case. However, as a sanity check it should confirm whether or not the model is working as expected and shows that, for a very "shallow-ice" like test case, the model indeed reproduces something that looks very much like shallow-ice flow.
The Confined Shelf test case
Now we will go all the way to the other end of the spectrum and demonstrate that the exact same model can also accurately reproduce ice shelf flow. In the follow idealized ice shelf test case, there has been no change at all in the governing equations of the model. The only thing that has changed is the geometry and the boundary conditions of the test problem. Instead of a parabolic dome with no basal slip we now have a flat, floating slab of ice with free-basal slip, zero-flux boundary conditions on three sides, and open-ocean (ice shelf) boundary conditions on the fourth side.
To run the test case, change into the tests/higher-order/shelf/ subdirectory. There are two idealized ice shelf tests cases here, confined-shelf.py and circular-shelf.py. To run the confined shelf test case, proceed with a similar set of steps as when running the dome test case. First, cp or link to the simple_glide executable, then run the test,
As in the previous test case, you should see gradually decreasing residuals as screen output. When the test completes, examine the output with
A color contour plot of CISM output (made in Matlab) is shown in the figure labeled Confined shelf test case. The black and white contour plot shows output for the same experiment using an SSA (ice shelf) model. It is from experiment 3 (page 7) of the EISMINT (European Ice Sheet Model InTercomparison) ice shelf intercomparison project documentation. We will return to this experiment and add some additional complexity to it in a later exercise.
If there is time, you may also want to try running the "circular-shelf" experiment, which demonstrates that CISM can also implement an accurate ice shelf boundary condition for an ice shelf front with a non-trivial shape in map view (i.e. one for which the shelf-front normal vectors are not parallel to coordinate directions).
The ISMIP-HOM test cases
The last set of diagnostic problems we will look at are from the ISMIP-HOM test suite. While the test suite includes a total of 6 tests we will look only at the tests for diagnostic solutions on idealized, three-dimensional domains. Each of these tests (A and C) includes a subset of 6 tests and, in the interest of time, we will look at only 3 of these.
Both tests consist of a uniformly sloping slab of ice with periodic lateral velocities in the x and y directions (i.e. in map plane). For test A, the basal topography varies periodically in x and y directions and the there is a no-slip basal boundary condition. For test C, the basal traction coefficient varies periodically in x and y and the thickness is uniform throughout the domain. While the amplitude of the variations (topography in A and traction coefficient in C) is the same for all tests, the wavelength, λ, is decreased by a factor of two for each successive test. For λ=160 km, the velocity solutions are essentially equal to those from a 0-order shallow ice model. Halving λ to 80 km, then to 40, 20, 10, and finally 5 km, the higher-order components of the stress balance become successively more important to the velocity solution. Figures 1 and 2 below shows relevant input data for each of the two experiments for λ = 80km. Here, in the interest of time, we will only run tests for the first three wavelengths in the series (160, 80, and 40 km).
To run the experiments, we will use some python scripts developed by colleagues at the University of Montana (also, see this link). As with the other test cases, the python script set up the necessary netCDF input files. In addition, the python scripts simplify things by allowing you to run and plot the results from multiple tests and multiple domain wavelengths sequentially. In addition, they plot CISM output relative to the model means and standard deviations from the actual benchmark study of Pattyn et. al (2008). This is a great convenience, as anyone who has ever done this on their own will attest to. First, move into the tests/higher-order/ismip-hom subdirectory. To execute test A, for λ=160, 80, and 40 km, type
python runISMIPHOM.py --exp=a --size=160,80,40
or for shorthand
python runISMIPHOM.py -e a -s 160,80,40
As with the other test cases above, you should see some screen output showing model residuals decreasing as the nonlinear iterations proceed (did you remember to link to or copy the simple_glide executable to the tests/higher-order/ismip-hom/ subdirectory?).
Plotting model output
To compare CISM output from your model runs with that from the ISMIP-HOM benchmark study of Pattyn et. al (2008), we will execute the python plotting scripts in a similar manner. For test A, type
python plotISMIPHOM.py --exp=a --size=160,80,40
python plotISMIPHOM.py -e a -s 160,80,40
Your output figure will have a ".png" extension and will be placed in the output/ subdirectory. It should look something like the figure here labeled ISMIP-HOM A Output.
Now go through the same set of steps for test case C (again, with wavelengths of 160, 80, and 40 km). You should get a figure that looks something like Figure 4.
- Try adjusting the horizontal and vertical grid spacing to see how it affects the results and/or model performance. For example, for the 80km tests, decrease the number of horizontal grid cells by a factor of two and increase the grid spacing by a factor of two,
[grid] upn = 11 ewn = 26 nsn = 26 dew = 3200 dns = 3200
How much faster does the model converge on a solution? Does the output still fall within the standard deviation given by the benchmarks? How What happens if the vertical resolution is doubled?
- Compare higher-order and 0-order solutions for test A with the 80 km domain length. To do this, in ishom.a.config, set the diagnostic_run flag to 0 instead of 1, rebuild the ishom.a.nc file using the python script (as done above), and re-run the model. When the model has finished running, examine ishom.a.out.nc using NCVIEW. Click on the variable uvelhom to make a colormap of the higher-order x component of velocity at time 1 (as shown in figure below).
Figure 4: Using NCVIEW to plot output of "ishom.a.out.nc" to compare higher-order and SIA solutions. Note that the value of current time is 2001, not 2000.
Click somewhere on the image to get a 2d velocity profile (choose x0 under Xaxis). Next, pick the variable uvel (the velocity from the SIA model) and do the same thing. When comparing the two profiles you should see something like in Figure 5.
Figure 5: Comparison of higher-order (top) and SIA (bottom) velocity profiles. For the same model domain, the HO velocities are ~25% slower due to the influence of horizontal-stress gradients, which the SIA model does not "feel" at all.
- Do the same for ISMIP-HOM test A for the 40 km domain. You should notice that, as the magnitude of the higher-order velocities continues to decrease with decreasing domain length, those for the SIA model do not. Why is this?
- Compare the values of the variable vvel (the across-flow velocity calculated from the SIA model) and vvelhom (the across-flow velocity calculated from the higher-order model) at time 200001. Can you explain the differences?