pilot topaz reanalysis (2003-2008)
TRANSCRIPT
Pilot TOPAZ reanalysis (2003-2008)
Pavel Sakov, Francois Counillon and Laurent Bertino
Nansen Environmental and Remote Sensing Center, Norway
9 September 2010, Bergen
NERSCNERSC
OutlineAbout this talk
Data assimilation system design
Observations
Analysis diagnostics: DFS and SRF
Innovation statistics
ValidationMoviesDriftersIce concentrationSome intercomparisons
ICECSSHSSSSST
The Arctic
Parameter estimation
Conclusions
About this talk
This talk concentrates on:
◮ General DA system design and performance
◮ Some outcomes of the pilot (2003-2008) MyOcean TOPAZreanalysis
◮ In the context of what can be expected from the main(1990-2010) reanalysis
We will not talk about...
◮ TOPAZ3 → TOPAZ4 development (model + DA)
◮ Ensemble generation
◮ System spin-up
◮ Numerical and operational issues
About this talk
This talk concentrates on:
◮ General DA system design and performance
◮ Some outcomes of the pilot (2003-2008) MyOcean TOPAZreanalysis
◮ In the context of what can be expected from the main(1990-2010) reanalysis
We will not talk about...
◮ TOPAZ3 → TOPAZ4 development (model + DA)
◮ Ensemble generation
◮ System spin-up
◮ Numerical and operational issues
Data assimilation system design
◮ 1. Rank issues
Ensemble size = 100 → localisation radius = 300 km.
(Effective radius ∼ 90 km → ∼ 13 × 13 × 28 local domain.)
◮ 2. Assimilation window
1 week
◮ 3. Scheme
DEnKF, asynchronous
◮ 4. Localisation
Distance-based (non-adaptive) local analysis, with G&C tapering
◮ 5. Moderation
Adaptive observation pre-screening, inflation 1%, R-factor = 2.
◮ 6. Parameter estimation
MSSH, SST, π∗.
Observations
Type Number After SO Asynchronous
Track SLA 9 · 104 4 · 104 Yes
SST (Reynolds) 6 · 103 ” No
SST (OSTIA) 2 · 106 2.2 · 105 No
In-situ T 2 · 104 + 1.5 · 104 6 · 103 No
In-situ S 2 · 104 + 1.5 · 104 6 · 103 No
Ice conc. (AMSR) 1.6 · 105 105 No
Ice drift (CERSAT) 6 · 103 ” Yes
Total 2.3 · 106 4 · 105
Analysis diagnostics: DFS and SRF
DFS = “Degrees of Freedomof Signal”
DFS = tr(KH)
0 5 10 15 20 25 30 35 400
1
2
3
4
5
6
7SVD spectrum of ensemble anomalies
mode
mag
nitu
de
DFS = 0.82
forecastanalysis
SRF = “Spread ReductionFactor”
SRF =√
tr(HPf HTR−1)tr(HPaHTR−1)
− 1
0 1 2 3 4 5 6 7 8 9 10 110
0.05
0.1
0.15
0.2
0.25
0.3
Ensemble spread
step
spre
ad
<SRF> = 0.172
forecastanalysis
DFS = 0.27 + 0.23 + . . .
− 0.02 − 0.02 = 0.82
Total DFS and SRF (for 24 April 2008)
DFS and SRF for TSLA
DFS and SRF for SST
DFS and SRF for ICEC
DFS and SRF for T
DFS and SRF for S
DFS of UICE
DFS of VICE
Innovation statistics
innovation = d − Hxf = observations − forecast
120 oW
60oW
0o
60
o E
120 oE
180o W
0 o
20 oN
40 oN
60oN
80
o N
1
2
3
4
Innovation statistics for TSLA
2003 2004 2005 2006 2007 2008
0
0.05
0.1
0.15
0.2
0.25region 1
0
2000
4000
biasrmsσ
ensσ
tot
# obs.
2003 2004 2005 2006 2007 2008
0
0.05
0.1
0.15region 2
0
5000
2003 2004 2005 2006 2007 2008−0.04
−0.02
0
0.02
0.04
0.06
0.08
0.1
region 4
0
5000
Innovation statistics for SST
2003 2004 2005 2006 2007 2008
−1.5
−1
−0.5
0
0.5
1
1.5
region 1
0
2
x 104
biasrmsσ
ensσ
tot
# obs.
2003 2004 2005 2006 2007 2008
−1
−0.5
0
0.5
1
1.5region 2
0
2
x 104
2003 2004 2005 2006 2007 2008−1
−0.5
0
0.5
1
region 4
0
x 104
Innovation statistics for T, 0 − 10 m
2003 2004 2005 2006 2007 2008
−1
0
1
2
3
region 1
0
20
40
60
80
biasrmsσ
ensσ
tot
# obs.
2003 2004 2005 2006 2007 2008
−1
−0.5
0
0.5
1
1.5
2
region 2
0
50
100
150
2003 2004 2005 2006 2007 2008
−1.5
−1
−0.5
0
0.5
1
1.5
2
region 3
0
50
100
150
200
250
300
350
2003 2004 2005 2006 2007 2008
−1
−0.5
0
0.5
1
1.5
region 4
0
10
20
30
40
50
60
Innovation statistics for T, 100 − 200 m
2003 2004 2005 2006 2007 2008
−1
0
1
2
3
4
5
region 1
0
10
20
30
40
50
60
70
biasrmsσ
ensσ
tot
# obs.
2003 2004 2005 2006 2007 2008
−1
0
1
2
3
region 2
0
20
40
60
80
100
120
140
2003 2004 2005 2006 2007 2008−1
−0.5
0
0.5
1
1.5
2
2.5
region 3
0
200
400
600
800
2003 2004 2005 2006 2007 2008−1
0
1
2
3
4
region 4
0
20
40
60
80
Innovation statistics for T, 500 − 1000 m
2003 2004 2005 2006 2007 2008
0
1
2
3
region 1
0
50
100
150
biasrmsσ
ensσ
tot
# obs.
2003 2004 2005 2006 2007 2008
−1
−0.5
0
0.5
1
1.5
2
2.5region 2
0
50
100
150
2003 2004 2005 2006 2007 2008
−2
−1
0
1
2
3
region 3
0
50
100
150
200
250
2003 2004 2005 2006 2007 2008
−0.5
0
0.5
1
1.5
region 4
0
20
40
60
80
Innovation statistics for S, 0 − 10 m
2003 2004 2005 2006 2007 2008−0.5
0
0.5
1
region 1
0
10
20
30
40
50
60
70
biasrmsσ
ensσ
tot
# obs.
2003 2004 2005 2006 2007 2008−0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
region 2
0
50
100
150
2003 2004 2005 2006 2007 2008
−1
0
1
2
3
4
5
6
region 3
0
50
100
150
200
250
300
2003 2004 2005 2006 2007 2008
−0.5
0
0.5
1
region 4
0
10
20
30
40
50
60
Innovation statistics for S, 100 − 200 m
2003 2004 2005 2006 2007 2008−0.2
0
0.2
0.4
0.6
0.8
region 1
0
10
20
30
40
50
60
70
biasrmsσ
ensσ
tot
# obs.
2003 2004 2005 2006 2007 2008
−0.1
0
0.1
0.2
0.3
0.4
region 2
0
20
40
60
80
100
120
140
2003 2004 2005 2006 2007 2008
0
0.2
0.4
0.6
0.8
1
1.2
region 3
0
200
400
600
800
2003 2004 2005 2006 2007 2008
0
0.1
0.2
0.3
0.4
0.5
0.6
region 4
0
20
40
60
80
Innovation statistics for S, 500 − 1000 m
2003 2004 2005 2006 2007 2008
0
0.1
0.2
0.3
0.4
0.5region 1
0
50
100
150
200
biasrmsσ
ensσ
tot
# obs.
2003 2004 2005 2006 2007 2008−0.1
0
0.1
0.2
0.3
0.4
region 2
0
50
100
150
2003 2004 2005 2006 2007 2008
−0.1
0
0.1
0.2
0.3region 3
0
50
100
150
200
250
2003 2004 2005 2006 2007 2008
−0.1
0
0.1
0.2
0.3
region 4
0
20
40
60
80
Innovation statistics for ICEC
2003 2004 2005 2006 2007 2008
−0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0
1
2
3
4
5
x 104
biasrmsσ
ensσ
tot
# obs.
SLA, SSS and SST movies for 2007
• SSH
• Salinity
• Temperature
SLA/SSH versus drifters - 9 January 2008 (± 4 days)
SSH versus drifters - 9 January 2008 (± 4 days)
SSH versus drifters - 9 January 2008 (± 4 days)
Ice concentration - 17 August 2005
Ice concentration - 30 April 2008
ICEC on 9 January 2008: reanalysis versus observations
ICEC on 9 January 2008: reanalysis versus TOPAZ3
ICEC on 9 January 2008: reanalysis versus free run
SLA on 9 January 2008: reanalysis versus TOPAZ3
SLA on 9 January 2008: reanalysis versus free run
SSS on 9 January 2008: reanalysis versus GDEM
climatology
SSS on 9 January 2008: reanalysis versus PHC climatology
SSS on 9 January 2008: reanalysis versus TOPAZ3
SSS on 9 January 2008: reanalysis versus free run
SST on 9 January 2008: reanalysis versus GDEM
climatology
SST on 9 January 2008: reanalysis versus TOPAZ3
SST on 9 January 2008: reanalysis versus free run
Arctic: SSH and Ice thickness movies for 2007
• SSH
• Ice thickness
(Thanks to Alexander Korablev for providing in-situ observation data for North
Atlantic and Arctic)
SSS, January 2007 and 2008
S at 100m, January 2007 and 2008
Parameter estimation
◮ Started from 20 February 2008
◮ Results shown after 10 cycles
To be done
Decisions:
◮ Use of parameter estimation in the main reanalysis
Validations:
◮ Time/area average profiles
◮ Taylor diagrams
◮ Some fancy movies with drifters, perhaps
Conclusions
◮ The TOPAZ reanalysis system has been developed and isclose now to its final shape.
◮ It proved to be robust and computationally effective. Beinggranted 100 CPUs on Hexagon, it will be possible to completethe 20-year reanalysis in 1 year.
◮ The quality of the analysis seems to be good, both in terms ofdynamical balance and match to observations and climatology.There seem to be a substantial improvement compared toTOPAZ3, particularly in the Gulf Stream and Arctic regions.
◮ The final shape of the system is to be decided by outcomes ofparameter estimation experiment.
Conclusions
◮ The TOPAZ reanalysis system has been developed and isclose now to its final shape.
◮ It proved to be robust and computationally effective. Beinggranted 100 CPUs on Hexagon, it will be possible to completethe 20-year reanalysis in 1 year.
◮ The quality of the analysis seems to be good, both in terms ofdynamical balance and match to observations and climatology.There seem to be a substantial improvement compared toTOPAZ3, particularly in the Gulf Stream and Arctic regions.
◮ The final shape of the system is to be decided by outcomes ofparameter estimation experiment.
Conclusions
◮ The TOPAZ reanalysis system has been developed and isclose now to its final shape.
◮ It proved to be robust and computationally effective. Beinggranted 100 CPUs on Hexagon, it will be possible to completethe 20-year reanalysis in 1 year.
◮ The quality of the analysis seems to be good, both in terms ofdynamical balance and match to observations and climatology.There seem to be a substantial improvement compared toTOPAZ3, particularly in the Gulf Stream and Arctic regions.
◮ The final shape of the system is to be decided by outcomes ofparameter estimation experiment.
Conclusions
◮ The TOPAZ reanalysis system has been developed and isclose now to its final shape.
◮ It proved to be robust and computationally effective. Beinggranted 100 CPUs on Hexagon, it will be possible to completethe 20-year reanalysis in 1 year.
◮ The quality of the analysis seems to be good, both in terms ofdynamical balance and match to observations and climatology.There seem to be a substantial improvement compared toTOPAZ3, particularly in the Gulf Stream and Arctic regions.
◮ The final shape of the system is to be decided by outcomes ofparameter estimation experiment.
References
Bertino, L. and K. A. Lisæter, 2008: The TOPAZ monitoring and prediction system for the Atlantic and
Arctic Oceans. J. Oper. Oceanogr., 1, 15–19.
Oke, P. R., G. B. Brassington, D. A. Griffin, and A. Schiller, 2008: The Bluelink ocean data assimilation
system (BODAS). Ocean Model., 21, 46–70.
— 2010: Ocean data assimilation: a case for ensemble optimal interpolation. AMOJ, 59, 67–76.
Sakov, P., G. Evensen, and L. Bertino, 2010: Asynchronous data assimilation with the EnKF. Tellus, 62A,
24–29, doi:10.1111/j.1600-0870.2009.00417.x.
Sakov, P. and P. R. Oke, 2008: A deterministic formulation of the ensemble Kalman filter: an alternative to
ensemble square root filters. Tellus, 60A, 361–371.
Simmons, U. S. D. D., A. and S. Kobayashi, 2007: PHC: A global ocean hydrography with a high-quality
Arctic Ocean. J. Climate, 14, 25–35.
Teague, C. M. J., W. J. and P. J. Hogan, 1990: Generalized Digital Environmental Model and Levitus
climatologies. J. Geophys. Res., 95, 7167–7183.