tindex.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
---
tindex.rst (5696B)
---
1 .. include:: ../../../global.txt
2
3 .. contents::
4
5 .. _sec-regional:
6
7 Modeling individual outlet glaciers
8 ===================================
9
10 PISM was created to model ice sheets in entirety; in this context we can assume that ice
11 does not extend to the edge of the computational domain and there is no need to provide
12 lateral boundary conditions.
13
14 However, in some interesting cases the ice *does* extend to the edge of the domain, but we
15 can assume that changes near the boundary do not affect the behavior in the region we want
16 to model. Two examples come to mind.
17
18 1. Modeling an individual alpine glacier: there may be ice near a domain boundary, but we
19 can select the domain so that corresponding ice masses are not connected to the glacier
20 we're modeling. Note that it is not always possible to "remove" glaciers we don't care
21 about because they may re-appear due to the mass balance forcing.
22 2. Modeling an outlet glacier in an ice sheet: we cut out a region from an ice sheet, so
23 ice *will* extend to the edge of the domain, but we can select the domain so that it
24 contains the whole *drainage basin* of the outlet glacier of interest.
25
26 In some ways this case is harder to model: the shape and size of drainage basins of an
27 ice sheet changes with its geometry, so we have to assume that our simulation is short
28 enough to ensure that the drainage basin we're modeling remains roughly the same in
29 shape and extent.
30
31 See :ref:`sec-jako` for an example.
32
33 Use
34
35 .. code:: bash
36
37 pismr -regional ...
38
39 to enable PISM's "regional mode."
40
41 Ideally, modeling a region containing an ice mass extending to the edge of the domain
42 would use the following time-dependent lateral boundary conditions
43
44 - energy balance: enthalpy flux across the boundary,
45 - mass continuity: mass flux across the boundary,
46 - stress balance: sliding speed at the boundary.
47
48 PISM's regional mode uses a special mask :var:`no_model_mask` (zeros in the interior of
49 the modeling domain, ones at the edge of the domain or in other areas that are *not
50 modeled*) to implement modifications at domain boundaries. This mask is saved to output
51 files and read back in when the model is re-started. Set :config:`regional.no_model_strip`
52 during bootstrapping to create a "non-modeled" strip of a given width along the domain
53 boundary.
54
55 Energy
56 ------
57
58 PISM assumes that ice enthalpy and the basal melt rate (i.e. parts of the model state that
59 capture the energy state) near the boundary of the domain *remain constant*: at the end of
60 each time step updated enthalpy and basal melt rate are re-set to values read from an
61 input file or computed during bootstrapping at all grid points where :var:`no_model_mask`
62 is `1`.
63
64 Stress balance
65 --------------
66
67 When prescribing the sliding velocity, the :var:`no_model_mask` overrides the basal
68 sliding B.C. mask: all :var:`no_model_mask` locations are *also* the Dirichlet B.C.
69 locations for the sliding velocity. This makes it possible to prescribe the sliding
70 velocity of the ice across the domain boundary. Set
71 :config:`stress_balance.ssa.dirichlet_bc` to ``true`` to enable this feature.
72
73 In many cases it makes sense to *disable* sliding at the boundary. When the sliding
74 velocity near the boundary is not prescribed, PISM sets the basal yield stress to a high
75 value (see :config:`regional.no_model_yield_stress`).
76
77 The domain in PISM "wraps around", which means that we can not accurately compute
78 gradients near the boundary in the non-periodic case.
79
80 Note, though, that updating the velocity field requires computing the gravitational
81 driving stress, which depends on gradients of the ice thickness and surface elevation.
82
83 To avoid using finite differences across the domain boundary when computing these
84 gradients, PISM stores ice thickness and surface elevation near the edge of the domain and
85 uses them to modify surface elevation and thickness gradients.
86
87 .. note::
88
89 In the SIA stress balance model, prescribing ice thickness and surface elevation near
90 the edge of the domain is equivalent to prescribing the *flux* across the domain
91 boundary.
92
93 To use *zero* surface elevation and thickness gradients, set
94 :config:`regional.zero_gradient`. (This disables SIA flow across the boundary.)
95
96 .. warning::
97
98 High surface elevation and ice thickness gradients near the domain boundary *will*
99 affect time-stepping even if they do now affect model evolution.
100
101 The resulting high SIA diffusivity will force PISM to take unreasonably short time
102 steps, wasting computational time.
103
104 Consider setting :config:`regional.zero_gradient` if you see high SIA diffusivities
105 near domain boundaries (save :var:`diffusivity_staggered` to check).
106
107 Mass continuity
108 ---------------
109
110 PISM uses the SSA Dirichlet B.C. mask as the ice thickness Dirichlet B.C. mask, i.e. ice
111 thickness is fixed wherever the sliding velocity is fixed. (In other words, PISM allows
112 prescribing the *ice flux* at a given location.)
113
114 This means that the *ice thickness does not evolve* in the :var:`no_model_mask` area.
115
116 Mass balance adjustment
117 ^^^^^^^^^^^^^^^^^^^^^^^
118
119 Prescribing the ice thickness near the boundary when the ice in the interior of the domain
120 thins would lead to high thickness and surface elevation gradients at the inner boundary
121 of the "non-modeled" strip. Use :ref:`sec-surface-forcing` to keep the ice geometry from
122 deviating from the target *without* sharp transitions at the boundary from fixed to
123 evolving ice thickness.
124
125 Calving
126 ^^^^^^^
127
128 Set :config:`geometry.front_retreat.wrap_around` to ``true`` to allow calving front retreat
129 due to calving to "wrap around" the computational domain. This may be necessary in some
130 regional synthetic-geometry setups.