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 (3559B)
---
1 # laboratory validation example
2
3 This example is a validation of [PISM's](http://www.pism-docs.org) isothermal
4 SIA numerical model using a laboratory experiment with a
5 [Xanthan gum](http://en.wikipedia.org/wiki/Xanthan_gum) suspension in water.
6 This fluid is more strongly shear-thinning than ice but it has nearly the same
7 density. This example is documented in section 12.2 of the PISM User's Manual.
8
9 The source of the set-up is the "constant flux" experiment in
10
11 R. Sayag and M. G. Worster, 2013. *Axisymmetric gravity currents of
12 power-law fluids over a rigid horizontal surface*, J. Fluid Mech. 716,
13 [doi:10.1017/jfm.2012.545](http://dx.doi.org/10.1017/jfm.2012.545).
14
15 See also
16
17 R. Sayag, S. S. Pegler, and M. G. Worster, 2012. *Floating extensional flows*,
18 Physics of Fluids 24 (9),
19 [doi:10.1063/1.4747184](http://dx.doi.org/10.1063/1.4747184).
20
21 In the laboratory experiment, fluid is pushed through the bottom of a flat table
22 from a tube of radius 8 mm (R. Sayag, personal communication) at a mass rate of
23 about 3 g/s (Sayag & Worster, 2013). The glaciological analog is an ice sheet
24 on a flat bed fed by positive surface mass balance in the vicinity of the dome,
25 but with zero surface mass balance everywhere else.
26
27 Sayag & Worster (2013) estimate n = 5.9 using regression of laboratory
28 measurements of the radius; see Figures 2(c) and 2(d) in the paper. More
29 precisely, they compare radius data to a similarity solution of the thickness
30 evolution equation to infer the exponent n.
31
32 See `preprocess.py` for settings of various parameters which are appropriate to
33 this fluid problem. The final total mass of fluid, about 1 kg, is about 18
34 orders of magnitude smaller than the mass of the Greenland ice sheet.)
35
36 The flow rate in the pipe is assumed to be constant across the pipe,
37 so the input "climate" has given `climatic_mass_balance` which is constant in
38 the pipe and zero outside the pipe.
39
40 ## basic usage
41
42 The preprocessing stage builds a NetCDF file suitable for PISM bootstrapping.
43 It is built at exactly the run-time resolution in order to make the flux
44 into the center of the "ice" sheet have the correct value given the small size
45 of the positive mass flux area. Also it creates `gumparams.nc` which contains
46 `pism_overrides` attributes, the special parameters for this experiment.
47
48 $ ./preprocess.py
49
50 Now view `initlab52.nc`. Only the `climatic_mass_balance` variable is
51 interesting.
52
53 Now we run for 746 model seconds (Sayag & Worster, 2013) on a 10 mm grid
54 (520 mm / 52 subintervals) using 4 processors:
55
56 $ ./rungum.sh 4 52 &> out.lab52
57
58 This run generates `out.lab52` from `stderr` and `stdout`,
59 and it also generates diagnostic NetCDF files `ts_lab52.nc` and `ex_lab52.nc`.
60 It takes about 5 minutes on a 2013 laptop.
61
62 Results are better on finer grids because the input pipe radius is only 8 mm.
63 For example, this uses a 5 mm grid, and takes about an hour to run:
64
65 $ ./preprocess.py -Mx 104 -o initlab104.nc
66 $ ./rungum.sh 4 104 &> out.lab104
67
68 You can compare multiple runs to the experimental data on radius (R. Sayag,
69 personal communication):
70
71 $ ./showradius.py -o foo.png -d constantflux3.txt ts_lab*.nc
72
73 ## higher resolutions
74
75 For a 2.5 mm grid taking several hours to run, do:
76
77 $ ./preprocess.py -Mx 208 -o initlab208.nc
78 $ ./rungum.sh 4 208 &> out.lab208
79
80 More processors (NN) are recommended for a 1.0 mm grid:
81
82 $ ./preprocess.py -Mx 520 -o initlab520.nc
83 $ ./rungum.sh NN 520 &> out.lab520
84
85 To experiment with different configuration constants, edit and rerun
86 `preprocess.py`.
87