4 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
5 |
============================== |
============================== |
6 |
|
|
7 |
|
o new AD-test experiment: obcs_ctrl |
8 |
|
(From Matt MITgcm_Contrib/SOSE/BoxAdj) to test OBCS as control vector. |
9 |
|
o pkg/dic: |
10 |
|
fix two bugs in handling water vapour pressure in carbonate chemistry |
11 |
|
and air-sea fluxes. Found by Val Bennington and Galen McKinley |
12 |
|
(add DIC Option = WATERVAP_BUG to recover old results) |
13 |
|
update output of tutorial_global_oce_biogeo (fwd+AD) & tutorial_dic_adjoffline |
14 |
|
o adding allowInteriorFreezing option to check for water below freezing point |
15 |
|
at depth and bring the negative heat anomaly to the surface level |
16 |
|
o pkgs _fields_load: |
17 |
|
- extend GET_PERIODIC_INTERVAL to deal with non-periodic case (from pkg/rbcs) |
18 |
|
- use new S/R GET_PERIODIC_INTERVAL in pkgs: |
19 |
|
bulk_force, cfc, dic, obcs (+ ctrl/ctrl_get_gen_rec), offline & rbcs to get |
20 |
|
interpolation weights and time reccord number. (left: cheapaml_fields_load). |
21 |
|
- fix multi-threading in aim_fields_load.F |
22 |
|
- apply EXCH to temp arrays [0,1] after reading in dic_fields_load.F |
23 |
|
o external_fields_load: |
24 |
|
new S/R GET_PERIODIC_INTERVAL to calculate which time-record (and linear |
25 |
|
interpolation weights) to use in a periodic time-series ; do not divide by |
26 |
|
time-step to convert to integer: |
27 |
|
- this fixes case where externForcingPeriod is not multiple of time-step |
28 |
|
(Thanks to Christopher L. Wolfe) |
29 |
|
- this fixes case where externForcingPeriod is an odd multiple of time-step |
30 |
|
(was wrong by half of time-step); update results of tutorial_global_oce_in_p. |
31 |
|
store index of time-record which is currently loaded: |
32 |
|
- works without any assumption on sequence of calls to EXTERNAL_FIELDS_LOAD ; |
33 |
|
- needed for the adjoint but not yet used because of extensive recomputations; |
34 |
|
(to be fixed soon) |
35 |
|
|
36 |
|
checkpoint62v (2011/04/02) |
37 |
|
o verification/1D_ocean_ice_column added to test Ian Fenty's |
38 |
|
pkg/seaice_growth.F modifications for improving sea ice code differentiability |
39 |
|
o update output of exp4, internal_wave, seaice_obcs, tutorial_plume_on_slope |
40 |
|
(new velocity statistics after changing maskInW & maskInS definition) |
41 |
|
o pkg/generic_advdiff with OBCS: |
42 |
|
use maksInW & maskInS in advective flux and in biharmonic diffusion calculation: |
43 |
|
equivalent to assumption of no tracer gradient outside the OB region. |
44 |
|
This will affect the results if using obcs+biharmonic diffusion. |
45 |
|
update output_adm.txt of experiment bottom_ctrl_5x5 (not expected since no obc) |
46 |
|
o pkg/obcs: |
47 |
|
change definition of maskInW & maskInS: leave OB edges inside (maskIn=1) |
48 |
|
(e.g., maskInW(OB_Ie)=1 ) so that velocity normal-component at OB is still |
49 |
|
in Interior region. No impact on solution but affects Velocity monitor stats. |
50 |
|
o pkg/ptracers (adjoint) |
51 |
|
New option AUTODIFF_PTRACERS_SPLIT_FILES for writing separate |
52 |
|
tape file for each tracer |
53 |
|
o model/inc/src and pkg/autodiff: add parameters that allow to turn |
54 |
|
of parts of the adjoint code at run time. For now, there are only |
55 |
|
useKPPinAdMode, useGMrediInAdMode, and useSEAICEinAdMode. These |
56 |
|
parameters are set in data.pkg. |
57 |
|
This is the lazy coward's version of doing this, but still useful for |
58 |
|
debugging. In order to do it right, everything needs to be moved to |
59 |
|
pkg/autodiff. |
60 |
|
o pkg/autodiff: fix global_sum_tile_ad.F by resetting adsumPhi at |
61 |
|
the end; this fixes the adjoint balancing code for open boundaries. |
62 |
|
o pkg/ctrl: |
63 |
|
- change behavior of xx_${varname}period to be analogous to exf-conventions: |
64 |
|
xx_${varname}period = 0., means time-independent control part |
65 |
|
xx_${varname}period = -12., means one-year-periodic monthly control part |
66 |
|
- introduce helper routine ctrl_init_rec (part of ctrl_init.F) to |
67 |
|
clean up ctrl_init a little, remove unused variables |
68 |
|
- ctrl_init: fixed the handling of the start/endrec for obcs-variables |
69 |
|
- move "call cal_GetMonthsRec" from s/r ctrl_get_gen to s/r ctrl_get_gen_rec |
70 |
|
|
71 |
|
checkpoint62u (2011/03/15) |
72 |
|
o model/src: |
73 |
|
- for the adjoint, change NonLin-FreeSurf initialisation sequence: fill-in |
74 |
|
h0Fac from INITIALISE_FIXED and reset hFac to h0Fac in INITIALISE_VARIA ; |
75 |
|
- in initialise_varia.F (except for ALLOW_DEPTH_CONTROL) remove 2nd call to |
76 |
|
INI_DEPTHS, INI_MASKS_ETC & INI_LINEAR_PHISURF. |
77 |
|
o pkg/ctrl/obcs/ecco: |
78 |
|
- remove obsolete, unused, and partially broken code: |
79 |
|
ctrl_obcsvol/ctrl_obcsbal/ctrl_volflux and code within |
80 |
|
ALLOW_CTRL_OBCS_BALANCE |
81 |
|
- put a stop statement into ecco/cost_obcsvol.F, but also add |
82 |
|
cost_obcsvol to ecco_ad_diff.list |
83 |
|
o pkg/ecco: fix counting cost function contributions for obcs (code |
84 |
|
by Matt M.) and diffkr for diagnostics |
85 |
|
o pkg/ecco, pkg/ctrl, pkg/grdchk: |
86 |
|
add cost penalty terms for initial uvel, vvel, eta controls |
87 |
|
and I/O for prior error files |
88 |
|
o pkg/mdsio: |
89 |
|
move mds_check4file.F from eesupp/src to mdsio_check4file.F ; add option |
90 |
|
(new argument) to check for tiled file in mdsioLocalDir : this fix the |
91 |
|
restart from pickup_som , pickup_dic (& dic_co2atm) with mdsioLocalDir. |
92 |
|
o pkg/ctrl: |
93 |
|
- ctrl_obcsbal.F, ctrl_obcsvol.F: replace last calls of ctrl_getrec by |
94 |
|
ctrl_get_gen_rec, put a stop statement into ctrl_getrec. Remove |
95 |
|
unused and superfluous variables |
96 |
|
- ctrl_get_gen_rec.F: catch the case of xx_genperiod=0. |
97 |
|
o pkg/ecco |
98 |
|
- enable adjoint of Stevens OB |
99 |
|
o pkg/ctrl/ecco/grdchk: |
100 |
|
- remove global declaration of tmpfldx/yz and tmpfldx/yz2. |
101 |
|
Fortunately, we do not have any tests for the numerous cpp-flagged |
102 |
|
option of "obcs as control parameters" so we will never know |
103 |
|
if this is an improvement (but at least now things compile). |
104 |
|
- remove an obsolete initialization from ecco_cost_init_barfiles.F |
105 |
|
- fix some problems within never used ifdefs to make the code compile |
106 |
|
o pkg/obcs: |
107 |
|
- Refine Interior mask definition to work with more complex OB (e.g. |
108 |
|
if OB_Ie < OB_Iw); In case default setting is not sufficient, allows to |
109 |
|
read in Interior mask from file. Check if Interior mask is consistent |
110 |
|
with OB list of indices. |
111 |
o pkg/seaice |
o pkg/seaice |
112 |
- beginning code for a seaice tracer package, |
- beginning code for a seaice tracer package, |
113 |
but currently only used for seaice age: SEAICE_AGE |
but currently only used for seaice age: SEAICE_AGE |