/[MITgcm]/MITgcm/doc/OpenBound_ctrl.txt
ViewVC logotype

Contents of /MITgcm/doc/OpenBound_ctrl.txt

Parent Directory Parent Directory | Revision Log Revision Log | View Revision Graph Revision Graph


Revision 1.1.2.1 - (show annotations) (download)
Thu Jun 19 15:38:17 2003 UTC (19 years, 2 months ago) by heimbach
Branch: ecco-branch
CVS Tags: ecco_c50_e33, ecco_c51_e34d, ecco_c51_e34e, ecco_c51_e34f, ecco_c51_e34g, ecco_c51_e34a, ecco_c51_e34b, ecco_c51_e34c, ecco_c50_e33a, ecco_c51_e34
Changes since 1.1: +111 -0 lines
File MIME type: text/plain
adding Gebbie notes on obcs.

1 %%%%% OPEN BOUNDARY CONTROL: NOTES %%%%%%%%%%
2 %%%%%
3 %%%%% G. GEBBIE, GEBBIE@MIT.EDU, 27-MAY-2003
4
5 SECTION 1: COMPILE-TIME PARAMETERS AND FLAGS.
6
7 1. Open boundary control is set by ALLOW_OBCSN_CONTROL for the northern
8 boundary (and OBCSS, OBCSW, OBCSE for South, West, and East
9 boundaries) in ECCO_CPPOPTIONS.h .
10
11 2. In parallel setups, there has been a problem controlling the barotropic
12 component of both the tangential and normal velocities. Switch
13 NO_CONTROL_BAROTROPIC_VELOCITY in ECCO_CPPOPTIONS.h to throw this
14 information out. The adjoint model still computes the gradient of
15 the cost function w.r.t. barotropic velocity but CTRL_PACK will only
16 pack zero values for this gradient. Below, I describe how gradients w.r.t.
17 barotropic velocity are computed in the first place.
18
19 3. parameter "nobcs" in SIZE.h refers to number of control variable
20 types on the boundary .
21 1 = TEMPERATURE
22 2 = TEMP., SALINITY
23 3 = TEMP., SAL., NORMAL VELOCITY
24 4 = TEMP., SAL., NORMAL VELOCITY, TANGENTIAL VELOCITY.
25
26 4. There are extra cost function routines.
27
28 O.B. regularization:
29 ALLOW_OBCSN_COST_CONTRIBUTION (OBCSS,OBCSW,OBCSE) computes a penalty
30 for deviations from the first guess open boundary field. The weights
31 for Temperature and Salinity are equivalent to those that are used
32 for the Levitus climatology. (For the ECCO 2 degree O.B.C.'s, I see
33 no reason that they should be more strongly weighted than Levitus.)
34 For the weighting of O.B. velocities, append a third column to "data.err"
35 which is a background velocity magnitude as a function of depth. Also,
36 append a fifth item to the first row of "data.err" which is an expected
37 magnitude for barotropic velocity adjustments.
38
39 O.B. Performance:
40 OBCS_AGEOS_COST_CONTRIBUTION penalizes deviations from thermal wind
41 balance on the boundaries. This is computed from the model monthly
42 average fields, and only puts a weak constraint on the Temperature,
43 Salinity, and Normal Velocity fields. Also, this balance is not required
44 in the upper layers of the ocean, but this is hard-coded into
45 cost_obcs_ageos at this time.
46
47 OBCS_VOLFLUX_COST_CONTRIBUTION penalizes an open boundary volume flux
48 which does not equal zero. This subroutine is presently obsolete and
49 needs to be rewritten.
50
51 SECTION 2: FORM OF THE O.B. CONTROL FILES, xx_obcsX and adxx_obcsX.
52
53 The north and south files have the size (sNx,Nr,Nt*nobcs).
54 West and east, (sNy,Nr,Nt*nobcs).
55 Nt, the number of O.B. control time periods is set in "data.exf"
56 with entries of this form:
57 obcsXstartdate1=19920116,
58 obcsXstartdate2=220000,
59 obcsXperiod=2592000.0,
60 where X=N,S,W,E.
61
62 For nobcs=4, note that the 3rd record index will alternate with
63 Temperature, Salinity, Meridional Velocity and Zonal Velocity for
64 the North and South boundaries. But the E,W boundaries will have
65 T,S,Zonal Velocity, Meridional Velocity, because of the definition
66 of Tangential and Normal Velocity.
67
68 Finally, both components of the velocity field are not filled with
69 the actual velocity adjustments. Instead,
70 level 1 contains the depth-integrated average velocity, or barotropic
71 velocity, and levels 2->Nr contain the baroclinic part of the velocity.
72 The baroclinic velocity and level 1 must be computed in such a way to
73 balance the depth-integrated volume flux of the baroclinic velocity.
74
75 OUTSTANDING ISSUES:
76 1. The code automatically balances the volume flux by the open
77 boundary control velocity adjustments. Theoretically, this
78 can be relaxed and the O.B. volume flux does not need to be
79 perfectly balanced. Add a switch for automatically balancing volume
80 flux (or not). Also automatic balance will only work if Northern
81 boundary exists.
82
83 2. In parallel mode, the gradients with respect to open boundary
84 barotropic velocity have been shown to be incorrect. Also, the
85 gradients w.r.t. tangential barotropic velocity reflect the tile
86 boundaries.
87
88 3. "ctrl_volflux" is the routine that calculates the volume flux
89 due to the O.B. control velocity adjustments. It must assume
90 that all the open boundaries have the same calendar.
91
92 4. Make a switch for changing the form of adxx_obcs ? (i.e.
93 split barotropic flag? )
94
95 Also, there are forward o.b.c.s issues:
96
97 1. Global slice files can not be read.
98 Slices can not be exchanged.
99
100 2. Must not be any slope in topography at boundary.
101
102 3. Must not be any chance of double periodicity at boundary.
103
104 4. Tile boundaries evident at open boundaries in spinup.
105
106 5. O.B. meta files don't work.
107
108
109
110
111

  ViewVC Help
Powered by ViewVC 1.1.22