PISM 2.3.0 is out

Published: Apr 1, 2026 by The PISM Authors

We are pleased to announce the release of PISM v2.3.0.

Notable changes since v2.2.0:

• Rename configuration parameters and command-line options related to reporting diagnostics (see issue #509).

  Configuration parameters

  • Everything with the prefix output.extra. to output.spatial.
  • Everything with the prefix output.timeseries. to output.scalar., except output.timeseries.filename to output.scalar.file
  • time_stepping.hit_extra_times to time_stepping.hit_spatial_times
  • time_stepping.hit_ts_times to time_stepping.hit_scalar_times
  • output.ISMIP6_extra_variables to ISMIP6_spatial_variables
  • output.ISMIP6_ts_variables to output.ISMIP6_scalar_variables

  Command-line options:

  • -extra_(append|file|times|vars|stop_missing|split) to -spatial_(append|file|times|vars|stop_missing|split)
  • -ts_(append|file|times|vars) to -scalar_(append|file|times|vars)
  • -extra_force_output_times to -spatial_force_output_times

  Using an old command-line option will make PISM stop with an error message.

• Implement asynchronous output to “snapshot” files and “spatial diagnostic” files (-save_file ... and -spatial_file ...). Requires PISM built with YAC, PROJ and Python packages netCDF4, yac (Python bindings for YAC), mpi4py and NumPy.
• Implemented a more efficient algorithm for the solution of the Eikonal equation in PICO.
• Add the ability to read in inputs (e.g. climate forcing) on longitude,latitude grids, including rotated pole (rlon,rlat) grids.
• Automatically build and upload Docker containers for each new PISM release to the GitHub Container Registry (uses Intel’s oneAPI compilers and Intel MPI; see ghcr.io/pism/pism).
• Remove dependency on the Python package nose because it is old and unmaintained. Now we use Python’s built-in unittest to set up Python-based regression tests.
• Spatially-variable diagnostic diffusivity: at a grid point (i,j) report max(D(i,j,0), D(i, j, 1)), i.e. maximum over staggered grid locations just to the east and just to the north of the grid point (i,j). Previous versions averaged the SIA diffusivity from all adjacent icy locations, which makes it impossible to see the value of the maximum SIA diffusivity used to choose the time step length.
• Add configuration parameter output.experiment_id: if set to a string, PISM adds an extra dimension (length 1) to output files. This makes it possible to open several output files (e.g. different ensemble members) at once using xarray’s xarray.open_mfdataset(). (Note: this feature needs further testing to ensure that PISM’s way of doing this is compatible with other tools. Implementation details are likely to change in the near future.)
• Save the JSON string encoding PISM’s configuration parameters to the pism_config variable in output files. This makes it easier to look up configuration parameter values in output files opened using xarray.open_mfdataset().
• Write run info (wall_clock_time, step_counter, model_years_per_processor_hour) to time-dependent variables instead of using attributes of the run_stats variable.
• Fix a bug in the steady state hydrology model: we need to save the time of the last update of the estimate of the water flux Q to a time-dependent variable.
• Fix issue #568 (crash when reading isochrone deposition times)
• Fix issue #569 (improper handling of calving front B.C. at domain boundaries)
• Require CMake 3.20 or newer.
• Require PETSc 3.15 or newer.
• Require NetCDF 4.7 or newer.
• Require YAC 3.14.0 or newer.