advancing ocean data assimilation and reanalysis
TRANSCRIPT
ADVANCING OCEAN DATA ASSIMILATION AT NCEPS. Penny1,2, D. Behringer,2 J. Carton1, E. Kalnay11University of Maryland, 2National Centers for Environmental Prediction (NCEP)
NOAA Climate Reanalysis Task Force Technical Workshop, May 4-5, 2015
SUMMARY
The Hybrid-GODAS
OSSE experiments
21-year Ensemble Reanalysis
THE HYBRID-GAIN 3DVAR/LETKFPenny, S.G., 2014: The Hybrid Local Ensemble Transform Kalman Filter. Mon. Wea. Rev., 142, 2139–2149.
K = KP +αKB −αKBHKP = KP +αKB I − HKP⎡⎣ ⎤⎦
KP = PbHT HPbHT + R( )−1KB = BHT HBHT + R( )−1
LETKF Gain
3DVar GainAdaptive
contraction
5 10 15 20 25 30 35 40
5
10
15
20
25
30
35
40
Ensemble size (k)
Obs
erva
tion
coun
t (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
5 10 15 20 25 30 35 40
5
10
15
20
25
30
35
40
Ensemble size (k)
Obs
erva
tion
coun
t (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
5 10 15 20 25 30 35 40
5
10
15
20
25
30
35
40
Ensemble size (k)
Obs
erva
tion
coun
t (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
5 10 15 20 25 30 35 40
5
10
15
20
25
30
35
40
Ensemble size (k)O
bser
vatio
n co
unt (
l)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
(a) α = 0.1 (b) α = 0.2 (c) α = 0.55 10 15 20 25 30 35 40
5
10
15
20
25
30
35
40
Ensemble size (k)
Obs
erva
tion
coun
t (l)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
5 10 15 20 25 30 35 40
5
10
15
20
25
30
35
40
Ensemble size (k)O
bser
vatio
n co
unt (
l)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
(a) α = 0.2 (b) α = 0.5
LETKFHybrid-
CovarianceHybrid-Gain
Ensemble Size Obs
erva
tion
Cou
nt
Analysis Error
Lorenz-96 system3DVa
r
HYBRID-GODASThe current system uses GFDL’s MOM4p1(1/2x1/4º) (LETKF is currently compatible with MOM versions 4,4p1,5,6)Localization is applied in the horizontal with 700km sigma-radius at the equator, decreasing to 200km at the poles.No localization is applied in the vertical, analysis weights are applied equally throughout all depths.Analysis variables: Temperature, Salinity, U/V velocities
Currently assimilating:Temperature ProfilesSSTSSH
Salinity ProfilesSSSOcean Color
Penny, S.G., D. Behringer, J. Carton, E. Kalnay, 2015: A
Hybrid Global Ocean Data Assimilation System at NCEP.
Mon. Wea. Rev., (Submitted for publication).
OBSERVING SYSTEM SIMULATION EXPERIMENT (OSSE)Experiments:
Nature‘Perfect’ 3DVar
3DVarLETKFHybrid
Ens. Size:1112828
Observations locations match the observations used in the Climate Forecast System Reanalysis (CFSR)
Forcing:R2R2
1 and 161-281-28
Imposed Bias:nonenone
R2 vs. 1, R2 vs. 16R2 vs. <1-28>R2 vs. <1-28>
Surface forcing perturbations come from a subsampling of the 56 members from the 20th Century Reanalysis (20CR)
Observation errors are identical between all DA experiments
91 92 93 94 95 96 97 98 99−0.1
−0.08
−0.06
−0.04
−0.02
0
0.02
date
bias
(ºC
)
91 92 93 94 95 96 97 98 99−6
−4
−2
0
2
4
6x 10−3
date
bias
(psu
)
91 92 93 94 95 96 97 98 990
0.1
0.2
0.3
0.4
0.5
date
rmse
(ºC
)
Temperature
3DVar (16)3DVar (01)Perfect Fluxes & ICsLETKFHYBRID
91 92 93 94 95 96 97 98 990
0.02
0.04
0.06
0.08
0.1
date
rmse
(psu
)
SalinityOSSE Results: RMSE and BIAS
3DVar
Hybrid
Hybrid
LETKF
LETKF
Perfect Surface Forcing 3DVar LETKF HybridNature Run (truth) PNature Run (truth)
OSSE Results: Error in the 20ºC Isotherm Depth
The Hybrid-GODAS generally reduces errors where there is disagreement between 3DVar and LETKF
OSSE Results: RMSE vs. background spread and BIAS in upper ocean
−0.1 −0.05 0 0.05 0.1
5
45
85
125
165
205
262
585
bias (ºC)
Temperature
dept
h (m
)
−0.015 −0.01 −0.005 0 0.005 0.01 0.015
5
45
85
125
165
205
262
585
bias (psu)
Salinityde
pth
(m)
0 0.1 0.2 0.3 0.4 0.5 0.6
5
45
85
125
165
205
262
585
rmse (ºC)
Temperature
dept
h (m
)
0 0.2 0.4 0.6 0.8 10 0.02 0.04 0.06 0.08
5
45
85
125
165
205
262
585
rmse (psu)
Salinity
dept
h (m
)
LETKFHYBRID
3DVar (1)3DVar (16)
The bias imposed on LETKF via surface forcing is reduced with the Hybrid
Daily global mean background ens. spread
Global mean 3DVar background error
3DVar
LETKFHybrid
OSSE SUMMARY
Ensemble approaches reduced RMSE of forecast and analysis errors vs. 3DVar
The Hybrid reduced biases imposed on LETKF via surface forcing conditions
21-YEAR REANALYSIS (1991-2011)Observation data mirrors the CFSR (T/S Profiles: XBT, TAO/TRITON, ARGO, etc.)
56-members, surface fluxes centered at R2 with perturbations from 20CR (T62)
The LETKF component does not use synthetic salinity.
The 3DVar component does use synthetic salinity.
The following results show Hybrid-GODAS vs. 3DVar-GODAS using identical models, observations, and observation errors.
OBSERVED-MINUS-FORECAST GLOBAL RMSD AND BIAS (0-700M)
92 94 96 98 00 02 04 06 08 100
1
2
3
4
rmsd
(ºC
)
Temperature
92 94 96 98 00 02 04 06 08 10−1
−0.5
0
0.5
1
date
bias
(ºC
)
Temperature
Temperature (O-F) RMSD and BIAS reduced using the Hybrid-GODAS (5-day forecasts)
3DVar
Hybrid-GODAS
3-month moving averages
21-Year Reanalysis Results (1991-2011)
OBSERVED-MINUS-FORECAST GLOBAL RMSD AND BIAS (0-700M)
Salinity (O-F) RMSD and BIAS reduced using the Hybrid-GODAS (5-day forecasts)
92 94 96 98 00 02 04 06 08 100
0.2
0.4
0.6
0.8
rmsd
(psu
)
Salinity
92 94 96 98 00 02 04 06 08 10−0.5−0.4−0.3−0.2−0.1
00.10.20.30.40.5
date
bias
(psu
)
Salinity
3DVar
Hybrid-GODAS
3-month moving averages
21-Year Reanalysis Results (1991-2011)
21-YEAR REANALYSIS (1991-2011)
Remaining results show the Hybrid-GODAS vs. the Operational GODAS
The operational GODAS uses: - MOM3, with 1x1/3º resolution - 3DVar with repeatedly reused observations, and assimilation of altimetry since 2007
Comparisons are made to Altimetry and the Met Office EN4 monthly objective analysisPurpose: Indicate impacts of Hybrid on monthly to seasonal timescales
Good, S. A., M. J. Martin and N. A. Rayner, 2013. EN4: quality controlled ocean temperature and salinity profiles and
monthly objective analyses with uncertainty estimates, Journal of Geophysical Research: Oceans, 118, 6704-6716.
CORRELATION WITH ALTIMETRYA summary of improvements over the operational 3DVar-GODAS:
Hybrid-GODAS sea level Anomaly Correlations (ACs) are generally improved across the global ocean.Sea level ACs and Root Mean Square Deviations (RMSDs) are improved particularly in the Tropical Pacific, Equatorial Atlantic, Southern Pacific, Southern Atlantic, and in the Southern Indian Ocean
21-Year Reanalysis Results (1991-2011)
Comparison with Altimetry SSH (1995-2011)Anomaly Correlation
RMSD Plots provided by Yan Xue
Altimetry assimilated 2007-2011 No Altimetry assimilated
Comparison with Altimetry SSH (1995-2011)Anomaly Correlation
RMSD Plots provided by Yan Xue
Focus: Tropical PacificAltimetry assimilated 2007-2011 No Altimetry assimilated
SEA SURFACE SALINITY (SSS)Summary of improvements over the operational 3DVar-GODAS
compared to EN4 monthly analysis:Hybrid-GODAS captures the ENSO cycle in Equatorial Pacific SSSAnomaly Correlations (ACs) in SSS and upper ocean salinity are increased throughout most of the global oceanRoot Mean Square Deviations (RMSDs) of SSS are reduced throughout most of the global oceanRMSDs of upper ocean salinity (S300) are reduced in the Tropical Pacific and Southern OceanRMSDs of S300 are increased in the Pacific extra-tropics and equatorial Atlantic
21-Year Reanalysis Results (1991-2011)
SEA SURFACE SALINITY AND EL NIÑO PREDICTABILITY “…in addition to the passive response, salinity variability may also play an active role in ENSO evolution, and thus important in forecasting El Niño events. By comparing two forecast experiments in which the interannual variability of salinity in the ocean initial states is either included or excluded, the salinity variability is shown to be essential to correctly forecast the 2007/08 La Niña starting from April 2007.”
Zhu, J., B. Huang, R-H. Zhang, Z-Z. Hu, A. Kumar, M.A. Balmaseda, L. Marx, J.L. Kinter, 2014: Salinity Anomaly as a Trigger for ENSO events. Nature, 4 : 6821.
Niño-3.4 SST anomalies (2001-2010) for observations (black), CTL (blue) and noS (red). Solid curves represent the forecast ensemble mean, and shaded areas the forecast ensemble spread.
Sea Surface Salinity (5m) Anomaly, (psu) 5ºS-5ºN
Plots provided by Yan Xue
HYBRID-GODAS captures the ENSO cycle in Tropical Pacific SSS
Anomaly Correlation
RMSD
Comparison with EN4 Sea Surface Salinity at 5m, 1995-2011
Plots provided by Yan Xue
Comparison with EN4 S300, (1995-2011)
Anomaly Correlation
RMSD
Plots provided by Yan Xue
Anomaly Correlation
RMSD
Comparison with EN4 Equatorial Salinity (1995-2011)Plots provided by Yan Xue
TEMPERATURESummary of improvements over the operational 3DVar-GODAS
compared to EN4 monthly analysis:Hybrid-GODAS increases Anomaly Correlations (ACs) and reduces Root Mean Square Deviations (RMSDs) of upper ocean temperature in the far western and eastern equatorial PacificHowever,There are reduced upper ocean temperature ACs outside of the Tropical Pacific
21-Year Reanalysis Results (1991-2011)
Comparison with EN4 Equatorial Temperature (1995-2011)Anomaly
Correlation
RMSD
Focus: Equatorial Pacific
Comparison with EN4 Temperature 0-300m (1995-2011)Anomaly
Correlation
RMSD
Focus: Equatorial Pacific
Comparison with EN4 Temperature 0-300m (1995-2011)Anomaly
Correlation
RMSD
The Hybrid-GODAS provides significant improvements over the current operational 3DVar-based GODAS.
Possible degradation in Atlantic basin and equatorial Indian Ocean heat content (compared to EN4).
Inclusion of satellite observations will improve upon this in situ assimilation baseline.
Coupling with a high-res atmospheric ensemble will improve representation of surface forcing uncertainty.
CONCLUSIONS