February 2010 update: THIS PAGE IS DEPRECATED.
PISM stable0.3 will appear soon and this page will be replaced with a new, and much more complete, version of SeaRISE-Greenland, as well as a link to this old page. For now, please see the poster presented at the Dec 2009 AGU Fall Meeting by Aschwanden et al.
Tools
You'll need PISM stable0.2 (or development branch) and two of the "NetCDF Operators", namely ncrename and ncatted.
Getting the dataset
The Greenland master dataset for SeaRISE is described at this page:websrv.cs.umt.edu/isis/index.php/Present_Day_Greenland.
(The particular file used in the script below is version 3 of the master dataset.)
Adjusting the metadata
The following two commands rename variables and dimensions and fix the units; these commands are in the spin-up script below:
ncrename -O -v x1,x -v y1,y -d x1,x -d
y1,y -v mapping,polar_stereographic \
-v presprcp,snowaccum Greenland_5km_v3.nc Greenland_5km_pism.nc
ncatted -O -a units,snowaccum,m,c,"m a-1" Greenland_5km_pism.nc
Greenland_5km_pism.nc
The resulting file Greenland_5km_pism.nc can be used with PISM.
Spin-up and run
A reference "spun-up" NetCDF file from CISM/GLIMMER will be used
once available.
This short shell script "psearise_v3.sh" shows a "demo" spin-up and run with PISM using above data. No source code modification are necessary to load all the data and treat it in a reasonable way for a trial run. The PISM command that loads and runs from the master dataset file (with metadata changes above) is:
pgrn -no_EI_delete -boot_from
Greenland_5km_pism.nc \
-Mx 151 -My 281 -Lz 4000 -Mz 51 -Mbz 51 -quadZ -skip 5 -y 100 -o
g10km_y100.nc
As demonstrated in the script, the spinup continues on from this. It regrids (interpolates) to a 5km grid and adds physics (described below).
Demo results
The demo script does a spin-up process using the nonsliding SIA and a total of 110Ka model years on 10km and then 5km (at -10Ka, by a "regrid" option). Then it produces a 5km result from a 1000 model year run with sliding and membrane stress balance, using a steady climate: g5km_ssa1000.nc.gz (~73Mb).
The animated GIFs at left came from this run. They were produced by IDV 2.7a1, which loads PISM NetCDF output files directly. Click on the GIFs for higher resolution MPEG files, also from IDV. These movies show 50 model year changes per frame, with a half-second delay per frame, from the final 1000 year run described in the script above.
The script includes a pcctest command to view the climate inputs, the time- and space-dependent modeled snow temperature, and the mass-balance that comes from the positive degree day model in PISM. The climate parameterization (lapse rates) are the PISM default values, those from the EISMINT-Greenland intercomparison. The pcctest "movie" is here: pddmovie.nc.gz
Discussion: thoughts about these "demo results" (Bueler)
The point of these first results is to demonstrate that "off-the-shelf" PISM can be used to work on the SeaRISE assessment process. Actual assessment is, naturally, much more work.
The movies at left show velocities in meters per year on a logarithmic color scale (base 10). The upper movie shows the modeled surface speed coloring the modeled surface elevation. The lower movie shows the basal sliding speed. The basal sliding speed is only significant where the till yield stress is sufficiently small, and this is controlled by thermodynamics as well as time-independent till friction angle. As demonstrated in the script, the till friction angle is a function of bed elevation, so sediments with a marine history have lower strength than upland bed. We believe this is an effective heuristic in the absense of inverse model information about bed strength. (See fields bwat,topg,phi,tauc in g5km_ssa1000.nc.gz.) The dynamical model is the one described in The shallow shelf approximation as a "sliding law" in a thermomechanically coupled ice sheet model.
The surface temperature and mass-balance parameterizations called for at the assessment description site are different from the PISM defaults used here. The PISM defaults are those from EISMINT-Greenland. This can and will be fixed, but it requires modifying source code, naturally.
Many parameter choices must be considered, now. In some cases the PISM defaults work fine, in others not. For example, the run yielded a volume growth rate of +144 km3/year. A primary reason for this growth rate is clearly that the mass balance parameterization used here (EISMINT-Greenland) calls for adding ice all around the periphery. Reconsideration of the mass balance parameterization is called for (including using additional information in version 5+ of the master dataset).
Also, the Humboldt glacier in NW Greenland slides strongly into the strait between Greenland and Ellesmere Island. But at present this glacier is relatively inactive, though it has a marine calving front and bed elevations significantly below sea level upstream from the grounding line. The speed of this glacier is controlled by basal resistance, with the till friction angle being the controllable parameter in PISM.
The NE Greenland ice stream does not start sufficiently far upstream. Note that the basal heat flux in the data set is missing and constant geothermal flux is used. We hypothesize that a geothermal hot-spot is exactly needed to initiate the stream far in the interior. (Version 5 of the master dataset adds Shapiro-Ritzwoller geothermal flux. But still there is no concentration of heat near the source of NE Greenland ice stream, despite observations suggesting such.)
The mass continuity time steps were on the order of 0.007 model years, with conservation of energy (temperature) and SSA stress balance updates every 40 such short steps, so on the order of a quarter-year (on the 5km grid of the data). The total run time for psearise_v3.sh was under 400 processor hours, run over a weekend on an 8 core machine. PISM can certainly scale up to 200+ processors for this run, without communication waste.