[Mitgcm-support] Control c20000630 vs filters kf021c, kf022e, kf025e, kf026a, and kf029.
mitgcm-support at dev.mitgcm.org
mitgcm-support at dev.mitgcm.org
Wed Jul 9 15:54:57 EDT 2003
HyperNews notification of new message. See:
http://escher.JPL.NASA.GOV:2000/HyperNews/get/forums/assim/106/1/2.html
Comparison of Kalman filters kf021c, kf022e, kf025e, kf026a, and kf029
with control integration (c20000630);
A full description of Kalman filter estimates is here[1] . Brief summary
of selected estimates follows:
kf021c : BT filter, NCEP Q
kf022e : BT filter, calibrated Q
kf025e : BC-BT combined filter, calibrated Q
kf026a : BC-BT combined filter, NCEP Q
kf029 : BC-BT filter combining "P"s of kf024i and kf022e.
kf024f : BC filter, calibrated Q
kf024i : BC filter, Same as kf024f but with tapered P
towards NS open boundary
Statistical comparison tables are here[2] . The following table
summarizes the skill of the various filters relative to bottom pressure
data:
(Resid. Var. Simulation) - (Resid. Var. Assimilation)
Skill = 100 * -----------------------------------------------------
Variance Data
Longitude Latitude kf019 kf021c kf022e kf025e kf026a kf029
77.58 -37.88 -8.42 -9.34 -11.81 -12.52 -10.05 -12.47
77.58 -37.90 -6.61 -5.96 -7.03 -7.87 -6.30 -7.89
53.47 -46.87 8.61 9.36 10.26 10.50 9.44 10.45
51.94 -46.40 5.03 6.28 5.85 6.31 6.93 6.26
303.64 -58.36 -2.08 -1.25 -3.15 -2.70 -1.79 -2.58
301.62 -54.94 -1.11 0.87 -0.92 -1.18 1.75 -1.24
301.62 -54.94 -2.22 1.97 -6.13 -10.71 -0.34 -10.57
302.50 -56.84 -0.83 -0.04 -0.02 0.28 -0.10 -0.54
305.28 -60.85 -0.21 2.22 -1.05 -0.04 2.07 0.08
305.28 -60.85 3.21 8.36 7.06 5.97 7.21 5.94
69.36 -47.66 4.54 3.74 -1.88 -2.67 3.16 -2.67
323.99 -32.00 0.77 0.03 -9.22 -9.92 -0.68 -10.01
304.51 -59.73 -0.48 0.48 0.02 -0.49 -0.15 -0.46
301.61 -54.94 1.72 11.05 9.70 8.46 10.89 8.36
301.61 -54.94 7.33 8.18 -3.02 -2.21 8.43 -2.20
305.28 -60.85 13.78 17.66 9.55 9.27 17.98 9.27
305.29 -60.85 -9.75 -4.82 -22.57 -19.89 -5.24 -19.75
146.22 -44.11 -2.71 -2.15 -1.80 -4.96 -2.04 -5.10
17.80 -34.59 0.54 4.57 -2.28 -0.73 3.72 -1.02
2.34 -42.99 13.08 16.95 18.56 19.01 17.43 18.99
3.02 -54.34 7.39 5.75 3.20 3.58 5.86 3.59
357.36 -70.14 -31.28 -27.35 -50.59 -55.32 -29.54 -55.63
146.22 -44.12 3.44 1.68 -1.55 -7.57 -2.80 -7.57
139.85 -65.56 -5.71 -3.86 -17.35 -18.09 -4.53 -18.04
357.39 -70.13 -5.06 -0.04 -13.35 -14.76 0.07 -14.58
==================================================================
AVERAGE SKILL: -0.28 1.77 -3.58 -4.33 1.26 -4.38
On average, kf021c is the most skillful assimilation, closely followed
by kf026a, that is, the assimilations that use NCEP Q for the barotropic
filter. The least skillful assimilation are kf022e, kf025e, and kf029,
that is the assimilations that make use of calibrated Q for the
barotropic filter.
Maps of simulation/assimilation skill versus pressure gauge data for the
the various Kalman filters follow:
kf019[3]
kf021c[4]
kf022e[5]
kf025e[6]
kf026a[7]
kf029[8]
A combined figure of coherency-squared for the various estimates
relative to data is shown here[9] . All Kalman filter estimates are
substantially better than pure simulation for periods greater than 20
days. The figure shows that most of the degradation of kf022e, kf025e,
and kf029 (calibrated Q) relative to kf021c and kf026a (NCEP Q) occurs
at periods shorter than 15 days. At longer periods, the
coherency-squared of all the filter estimates relative to data are
comparable.
-----------------------------------------------------------------------
[1] http://escher:2000/hosts/escher/escher6/medea/btang/kf_bc_runs/kf_run.txt
[2] http://escher:2000/hosts/triton/dm1/dimitri/data/gloup/matlab/skill.txt
[3] http://escher:2000/hosts/triton/dm1/dimitri/data/gloup/FIGS/skill.ps
[4] http://escher:2000/hosts/triton/dm1/dimitri/data/gloup/FIGS/skill_kf021c.ps
[5] http://escher:2000/hosts/triton/dm1/dimitri/data/gloup/FIGS/skill_kf022e.ps
[6] http://escher:2000/hosts/triton/dm1/dimitri/data/gloup/FIGS/skill_kf025e.ps
[7] http://escher:2000/hosts/triton/dm1/dimitri/data/gloup/FIGS/skill_kf026a.ps
[8] http://escher:2000/hosts/triton/dm1/dimitri/data/gloup/FIGS/skill_kf029.ps
[9] http://escher:2000/hosts/triton/dm1/dimitri/data/gloup/FIGS/CohSq3.ps
More information about the MITgcm-support
mailing list