Using PISM for flow-line modeling¶

As described in sections Computational box and Spatial grid, PISM is a three-dimensional model. Moreover, parameters grid.Mx and grid.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 grid.Mx or grid.My to \(1\).

There is a way around this, though: by using grid.periodicity 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.

In this case grid.My can be any number; we want to avoid unnecessary computations, though, so grid.My of \(3\) 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 pism_flowline 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 examples/preprocessing/flowlineslab.py to create a minimal, and obviously unrealistic, dataset. The file slab.nc created by this script contains all the required information but is not ready to use with PISM. Proceed as follows:

examples/preprocessing/flowlineslab.py # creates slab.nc with only an x-direction
pism_flowline -o slab-in.nc --expand -d y slab.nc

This produces a PISM-ready slab-in.nc. Specifically, pism_flowline “expands” its input file in the \(y\)-direction. Now we can “bootstrap” from slab-in.nc:

mpiexec -n 2 pism \
        -surface given \
        -bootstrap -i slab-in.nc \
        -Mx 201 -Lx 1000 \
        -My 3 -Ly 4 -periodicity y \
        -Lz 2000 -Mz 11 \
        -y 10000 -o pism-out.nc

To make it easier to visualize data in the file created by PISM, “collapse” it using NCO:

ncks -O -d y,1 pism-out.nc slab-out.nc
ncwa -O -a time,y slab-out.nc slab-out.nc

Footnotes