tREADME.md - pism - [fork] customized build of PISM, the parallel ice sheet model (tillflux branch)
(HTM) git clone git://src.adamsgaard.dk/pism
(DIR) Log
(DIR) Files
(DIR) Refs
(DIR) LICENSE
---
tREADME.md (6091B)
---
1 MISMIP in PISM
2 ==============
3
4 This directory contains scripts that can be used to run MISMIP experiments using PISM. To understand the intent of these experiments, please see the MISMIP website at http://homepages.ulb.ac.be/~fpattyn/mismip/, and download the intercomparison description PDF from that site.
5
6 Older PISM versions included C++ code managing MISMIP experiments. With the addition of more sophisticated reporting code that old code became unnecessary. Here all MISMIP-specific code is in Python scripts.
7
8 Step by step instructions
9 -------------------------
10
11 First of all, you will need to copy or symlink `util/PISMNC.py` into the
12 current directory to make sure that Python will be able to import it.
13
14 The script `run.py` is used to generate a `bash` scripts performing MISMIP
15 experiments. Running `run.py --help` produces the following:
16
17 Usage: run.py [options]
18
19 Creates a script running MISMIP experiments.
20
21 Options:
22 -h, --help show this help message and exit
23 --initials=INITIALS Initials (3 letters)
24 -e EXPERIMENT, --experiment=EXPERIMENT
25 MISMIP experiments (one of '1a', '1b', '2a', '2b',
26 '3a', '3b')
27 -s STEP, --step=STEP MISMIP step number
28 -u, --uniform_thickness
29 Use uniform 10 m ice thickness
30 -a, --all Run all experiments
31 -m MODE, --mode=MODE MISMIP grid mode
32 --Mx=MX Custom grid size; use with --mode=3
33 --model=MODEL Models: 1 - SSA only; 2 - SIA+SSA
34 --executable=EXECUTABLE
35 Executable to run, e.g. 'mpiexec -n 4 pismr'
36
37 For example, to set up MISMIP experiment `1a` using grid mode 1, a 12 km grid, run
38
39 ./run.py -e 1a --mode=1 > experiment-1a-mode-1.sh
40
41 This will create `experiment-1a-mode-1.sh` as well as the bootstrapping file
42 `MISMIP_boot_1a_M1_A1.nc` and configuration files corresponding to each "step"
43 in the experiment.
44
45 Run in the backround with 2 cores and saving output to a text file this way:
46
47 bash experiment-1a-mode-1.sh 2 >& out.1a-mode-1 &
48
49 You can also copy the script (along with
50 `MISMIP_boot_1a_M1_A1.nc` and `MISMIP_conf_1a_A*.nc`) to a supercomputer to
51 do the run later. For such application, the script helpfully uses environment variables `PISM_DO`,
52 `PISM_BIN` and `PISM_MPIDO`. For example, on some Cray machines you might do
53
54 PISM_MPIDO="aprun -n " bash experiment-1a-mode-1.sh 32
55
56 will use `aprun` on 32 cores. Alternatively, you can use
57
58 ./run.py -e 1a --mode=1 --executable="aprun -n 32 pismr"
59
60 or similar to skip the "preamble" handling environment variables and get "raw"
61 commands.
62
63
64 Refined grid runs
65 -----------------
66
67 The above "grid mode 1" runs use 150 grid spaces in the MISMIP modeling domain,
68 which is 301 grid points in PISM's (doubled) domain. The domain is doubled because
69 PISM is easiest configure as a whole ice sheet model with ice free ocean at the
70 edge of the computation domain. (Compare the example in `examples/jako/`, however.)
71
72 To run a higher resolution 3 km grid, with somewhat-improved grounding line
73 performance, ask the `run.py` script to put option `-Mx 1201` into the bash
74 script:
75
76 ./run.py -e 1a --mode=3 --Mx=1201 > experiment-1a-mode-3.sh
77
78 Then this is a 4 core run:
79
80 bash experiment-1a-mode-3.sh 4 >& out.1a-mode-3 &
81
82
83 Technical details
84 -----------------
85
86 The script `MISMIP.py` contains MISMIP parameters and the code needed to
87 compute the semi-analytic grounding line location and the corresponding
88 thickness profile for each experiment.
89
90 The script `prepare.py` contains functions using `MISMIP.py` to generate
91 PISM-readable NetCDF files with semi-analytic ice thickness profiles, and
92 the prescribed accumulation map. It can be imported as a module or run
93 as a script to generate PISM bootstrapping files.
94
95 The script `run.py` generates `bash` scripts performing MISMIP runs using
96 `MISMIP.py` and `prepare.py`.
97
98 Implementation details
99 ----------------------
100
101 We can turn PISM's default sliding law into MISMIP's power law by setting the
102 threshold speed to 1 meter per second, which will make it inactive.
103
104 The `-pseudo_plastic_uthreshold` command-line option takes an argument in meters per year, so we use `-pseudo_plastic_uthreshold 3.15569259747e7`, where `3.15569259747e7` is the number of seconds in a year.
105
106 The MISMIP parameter C corresponds to `tauc` in PISM. It can be set using `-yield_stress constant -tauc C`.
107
108 The MISMIP power law exponent `m` corresponds to `-pseudo_plastic_q` in PISM.
109
110 We use the `-config_override` option to set other MISMIP-specific parameters, such as ice softness, ice density and others.
111
112 Note that PISM does not at this time implement the stopping criteria described in the MISMIP specification. Instead we use the maximum run lengths that are provided as an alternative. On the other hand, PISM's output files contain all the information necessary to compute the rate of change of the grounding line position and the thickness rate of change during post-processing.
113
114 Post-processing
115 ---------------
116
117 Converting PISM output files to ASCII files following MISMIP
118 specifications is left as an exercise.
119
120 However, we do provide the `plot.py` script for visualization.
121
122 It plots modeled ice flux as a function of the distance from the divide and the geometry profile.
123
124 We see a discontinuity in the flux at the grounding
125 line. This is an issue in PISM that needs to be addressed to improve its
126 handling of the grounding line motion. For example, try
127
128 ./plot.py ABC1_1a_M1_A1.nc -f -o flux.png
129
130 Run `plot.py --help` for a list of command-line options. You can also produce plots for several PISM output files at once. For example,
131
132 ./plot.py ABC*.nc
133
134 will create geometry profile *and* ice flux plots for all matched files.
135
136 Also, note that the variable `ice_area_glacierized_grounded` in `ts_ABC1_1a_M1_A1.nc` and similar
137 allows one to see time-dependent changes in the grounding line location
138 because grounded ice area is proportional to the distance from the divide to the
139 grounding line.