tmismip.rst - 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
---
tmismip.rst (6300B)
---
1 .. include:: ../../global.txt
2
3 .. _sec-MISMIP:
4
5 MISMIP
6 ------
7
8 This intercomparison addresses grounding line dynamics by considering an idealized
9 one-dimensional stream-shelf system. In summary, a flowline ice stream and ice shelf
10 system is modeled, the reversibility of grounding line movement under changes in the ice
11 softness is tested, different sliding laws are tested, and the behavior of grounding lines
12 on reverse-slope beds is tested. The intercomparison process is described at the website
13
14 |mismip-url|
15
16 Find a full text description there, along with the published report on the results
17 :cite:`MISMIP2012`; that paper includes results from PISM version 0.1. These documents are
18 essential reading for understanding MISMIP results generally, and for appreciating the
19 brief discussion in this subsection.
20
21 PISM's version of MISMIP includes an attached ice shelf even though modeling the shelf is
22 theoretically unnecessary in the flow line case. The analysis in :cite:`SchoofMarine1` shows
23 that the only effect of an ice shelf, in the flow line case, is to transfer the force
24 imbalance at the calving front directly to the ice column at the grounding line. Such an
25 analysis does not apply to ice shelves with two horizontal dimensions; real ice shelves
26 have "buttressing" and "side drag" and other forces not present in the flow line
27 :cite:`Goldbergetal2009`. See the next subsection on MISMIP3d and the Ross ice shelf example in
28 section :ref:`sec-ross`, among other examples.
29
30 We must adapt the usual 3D PISM model to two horizontal dimensions, i.e. to do flow-line
31 problems (see section :ref:`sec-flowline-modeling`). The flow direction for MISMIP is
32 taken to be "`x`". We periodize the cross-flow direction "`y`", and use the minimum number
33 of points in the `y`-direction. This number turns out to be "``-My 3``"; fewer points than
34 this in the cross-flow direction confuses the finite difference scheme.
35
36 PISM can do MISMIP experiments with either of two applicable ice dynamics models. Model 1
37 is a pure SSA model; "category 2" in the MISMIP classification. Model 2 combines SIA and
38 SSA velocities as described in :cite:`Winkelmannetal2011`; "category 3" because it resolves
39 "vertical" shear (i.e. using SIA flow).
40
41 There are many runs for a complete MISMIP intercomparison submission. Specifically, for a
42 given model there are `62` runs for each grid choice, and three (suggested) grid choices,
43 so a full suite is `3 \times 62 = 186` runs.
44
45 The coarsest grid ("mode 1") has 12 km spacing. The finest grid, "mode 2" with 1.2 km
46 spacing, accounts for all the compute time, however; in the MISMIP description it is 1500
47 grid spaces in the flow line direction (= 3001 grid *points* in PISM's doubled
48 computational domain). In between is "mode 3", a mode interpretable by the intercomparison
49 participant, and here we just use a 6 km grid.
50
51 The implementation of MISMIP in PISM conforms to the intercomparison description, but that
52 document specifies
53
54 ... we require that the rate of change of grounding line position be `0.1` m/a or
55 less, while the rate of change of ice thickness at each grid point at which ice
56 thickness is defined must be less than `10^{-4}` m/a...
57
58 as a standard for "steady state". The scripts here do not implement this stopping
59 criterion. However, we report enough information, in PISM output files with scalar and
60 spatially-variable time-series, to compute a grounding line rate or the time at which the
61 thickness rate of change drops below `10^{-4}` m/a.
62
63 See
64
65 .. code-block:: none
66
67 examples/mismip/mismip2d/README.md
68
69 for usage of the scripts that run MISMIP experiments in PISM. For example, as described in
70 this ``README.md``, the commands
71
72 .. code-block:: none
73
74 ./run.py -e 1a --mode=1 > experiment-1a-mode-1.sh
75 bash experiment-1a-mode-1.sh 2 >& out.1a-mode-1 &
76 ./plot.py ABC1_1a_M1_A7.nc -p -o profileA7.png
77
78 first generate a bash script, then use it to do a run which takes about 20 minutes, and
79 then generate an image in ``.png`` format. Note that step 7 is in the middle of the
80 experiment. It is shown in :numref:`fig-MISMIPmodel1exper1aA7` (left).
81
82
83 .. figure:: figures/mismip-resolution.png
84 :name: fig-MISMIPmodel1exper1aA7
85
86 A marine ice sheet profile in the MISMIP intercomparison; PISM model 1, experiment 1a,
87 at step 7. Left: grid mode 1 (12 km grid). Right: grid mode 3 (6 km grid).
88
89 .. figure:: figures/SM-1a-A1.png
90 :name: fig-SMexper1aM1A1
91
92 Analytical profile for steady state of experiment 1a, step 1, from theory in
93 :cite:`SchoofMarine1`. This is a boundary layer asymptotic matching result, but not the
94 exact solution to the equations.
95
96 The script ``MISMIP.py`` in ``examples/mismip/mismip2d`` has the ability to compute the
97 profile from the Schoof's :cite:`SchoofMarine1` asymptotic-matching boundary layer theory. This
98 script is a Python translation, using ``scipy`` and ``pylab``, of the `provided MATLAB
99 codes <mismip-code_>`_. For example,
100
101 .. code-block:: none
102
103 python MISMIP.py -o mismip_analytic.png
104
105 produces a ``.png`` image file with :numref:`fig-SMexper1aM1A1`. By default
106 ``run.py`` uses the asymptotic-matching thickness result from the :cite:`SchoofMarine1` theory
107 to initialize the initial ice thickness, as allowed by the MISMIP specification.
108
109 .. figure:: figures/profileA7-M2.png
110 :name: fig-MISMIPmode2results
111
112 Results from MISMIP grid mode 2, with 1.2 km spacing, for steady state of experiment
113 1a: profile at step 7 (compare :numref:`fig-MISMIPmodel1exper1aA7`).
114
115 Generally the PISM result does not put the grounding line in the same location as Schoof's
116 boundary layer theory, and at least at coarser resolutions the problem is with PISM's
117 numerical solution, not with Schoof's semi-analytic theory. The result improves under grid
118 refinement, however. Results from grid mode 3 with 6 km spacing, instead of 12 km in mode
119 1, are the right part of :numref:`fig-MISMIPmodel1exper1aA7`. The corresponding
120 results from grid mode 2, with 1.2 km spacing, are in Figure
121 :numref:`fig-MISMIPmode2results`. Note that the difference between the numerical grounding
122 line location and the semi-analytical location has been reduced from 76 km for grid mode 1
123 to 16 km for grid mode 2 (a factor of about 5), by using a grid refinement from 12 km to
124 1.2 km (a factor of about 10).
125