Using PISM for flow-line modeling

As described in sections Computational box and Spatial grid, PISM is a three-dimensional model. Moreover, parameters Mx and My have to be greater than or equal to three, so it is not possible to turn PISM into a 2D (flow-line) model by setting Mx or My to 1.

There is a way around this, though: by using the -periodicity option to tell PISM to make the computational grid \(y\)-periodic and providing initial and boundary conditions that are functions of \(x\) only one can ensure that there is no flow in the \(y\)-direction. (Option -periodicity takes an argument specifying the direction: none, x, y and xy — for “periodic in both X- and Y-directions”.)

In this case Mx can be any number; we want to avoid unnecessary computations, though, so “-Mx 3” is the obvious choice.

One remaining problem is that PISM still expects input files to contain both x and y dimensions. To help with this, PISM comes with a Python script that turns NetCDF files with \(N\) grid points along a flow line into files with 2D fields containing \(N\times3\) grid points.[1]

Here’s an example which uses the script util/ to create a minimal, and obviously unrealistic, dataset. A file is created by util/, but it is not ready to use with PISM. Proceed as follows, after checking that util/ is on your path:                         # creates with only an x-direction -o --expand -d y

produces a PISM-ready Specifically, “expands” its input file in the y-direction. Now we can “bootstrap” from

mpiexec -n 2 pismr -surface given -i -bootstrap -periodicity y \
        -Mx 201 -My 3 -Lx 1000 -Ly 4 -Lz 2000 -Mz 11 -y 10000 -o

To make it easier to visualize data in the file created by PISM, “collapse” it: -o --collapse -d y


[1]This script requires the numpy and netCDF4 Python modules. Run --help for a full list of options.

Previous Up Next