Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007 Slide 1 Observing System experiments with ECWMF operational ocean analysis (ORA-S3) The new ECMWF operational ocean analysis system - Historical reanalysis and real time - The ORA-S3 analysis system - Impacts of data assimilation (mean/variability/forecast skill) Results from OSEs - Impact on the ocean state - Impact on forecasts - Impact on climate variability
33
Embed
Slide 1 Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007 Observing System experiments with ECWMF operational ocean analysis (ORA-S3) The new.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 1
Observing System experiments with ECWMF operational ocean analysis (ORA-S3)
The new ECMWF operational ocean analysis system
- Historical reanalysis and real time
- The ORA-S3 analysis system
- Impacts of data assimilation (mean/variability/forecast skill)
Results from OSEs
- Impact on the ocean state
- Impact on forecasts
- Impact on climate variability
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 2
Delayed Ocean Analysis ~12 days
Real Time Ocean Analysis ~Real time
ECMWF:
Weather and Climate Dynamical Forecasts
ECMWF:
Weather and Climate Dynamical Forecasts
10-Day Medium-Range
Forecasts
10-Day Medium-Range
Forecasts
Seasonal Forecasts
Seasonal Forecasts
Monthly Forecasts
Monthly Forecasts
Atmospheric model
Wave model
Ocean model
Atmospheric model
Wave model
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 3
Coupled Hindcasts, needed to estimate climatological PDF, require a historical ocean reanalysis
Real time Probabilistic Coupled
Forecasttime
Ocean reanalysis
Quality of reanalysis affects the climatological
PDF
Consistency between historical and real-time initial initial conditions is required
Main Objective: to provide ocean Initial conditions for coupled forecasts
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 5
•Ocean model: HOPE (~1x1, equatorial refinement)
•Assimilation Method OI (3D OI).
•ERA-40 fluxes to initialize ocean.
•Retrospective Ocean Reanalysis back to 1959.
•Assimilation of T
•Assimilation of salinity data.
•Assimilation of altimeter-derived sea level anomalies.
•Multivariate on-line Bias Correction .
•Balanced relationships (T-S, ρ-U)
•10 days assimilation windows, increment spread in time
ORA-S3 Ocean Re-Analysis System
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 6
Observations used in the S3 ocean analysis
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 7
Observation Monitoring
60°S 60°S
30°S30°S
0° 0°
30°N30°N
60°N 60°N
60°E
60°E
120°E
120°E
180°
180°
120°W
120°W
60°W
60°W
0°
0°
Moorings: 909 profiles
Argo floats: 2520 profiles
XBT probes: 394 profiles
Fully Rejected: 795 profiles
Fully Accepted: 2280 profiles
Partially Accepted: 748 profiles
(at least one per profile)
SuperObs: 1404 profiles
10 days period centered on 20070529
S3 ocean analysis
monitoring (temp)
In situ observation
60°S 60°S
30°S30°S
0° 0°
30°N30°N
60°N 60°N
60°E
60°E
120°E
120°E
180°
180°
120°W
120°W
60°W
60°W
0°
0°
60°S 60°S
30°S30°S
0° 0°
30°N30°N
60°N 60°N
60°E
60°E
120°E
120°E
180°
180°
120°W
120°W
60°W
60°W
0°
0°
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 8
Altimeter product1. Ingredients:
2. Assimilation of detrendend sea level, taking care of removing the spatial average from the altimeter data:
alt' Observed SLA from T/P+ERS+GFO+Jason+ENVISATRespect to 7 year mean of measurements.Weekly anomalies, twice a week.Global gridded maps
A Mean Sea LevelChoice: MSL from an analysis where no altimeter has been assimilated
''
'''' average spatial ~
; ~
'
altalt
altaltaltalt
There are MSL products derived from GRACE (Rio4/5 from CLS, NASA, …) but the
choice of the reference global mean is not trivial and the system can be quite sensitive to
this choice. Better assimilation methods are needed to make optimal use of the Gravity
product
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 9
Sequential Assimilation of data streams
T/S
conserved
alt T/S
Changedinsituinsitu ST ,
aa ST ,
T/S
conservedinsituT
', aa ST
1. Assimilation of Sea level anomalies
2. Assimilation of Subsurface temperature
3. Assimilation of Salinity
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 10
Bias evolution vector-equation
Some notation (Temperature,Salinity,Velocity)
T
UTS
TT
TUST
T
T
fkkkk
akk
fk
fk
LLKLbbbb
USTx
xHydbbxx
,, ; ,,
; ,,
~ ~
; ~1
k
U
S
T
a
kU
S
T
kU
S
T
a
kU
S
T
d
L
L
K
b
b
b
b
b
b
b
b
b~
1
prescribed (constant/seasonal)
k
fkk
fk
b
bbb ; 1
Balmaseda et al 2007, QJRMS
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 11
Effect of the pressure-gradient correction
50OE 100OE 150OE 160OW 110OW 60OW 10OW
Longitude
500
400
300
200
100
0
Depth
(m
etr
es)
500
400
300
200
100
0Plot resolution is 1.4063 in x and 10 in yZonal section at 0.00 deg NICODE=178 contoured every 0.0002 XXXHOPE gcm:: 0001
•Upper 300m, there is a large degree of coherence in ORAS3, ORA-nobs, Lev. The largest signals are in ORAS3 (SYNERGY?)
•Deeper Ocean: In ORA-nobs the decadal signals do not penetrate deep enough?
•OSEs indicate that 2002-2003 upper ocean cooling is robust
•Cooling after 2003 in ORAS3 is a consequence of ARGO in the Southern Oceans.The ARGO SOLO/FSI are not responsible for the post-2004 cooling
Global T300
2001 2002 2003 2004 2005 2006Time
-0.04
-0.02
0.00
0.02
0.04ORAS3ORA-nobsNOSOLONOARGO
Global T700
2001 2002 2003 2004 2005 2006Time
-0.08
-0.06
-0.04
-0.02
0.00
0.02
0.04
ORAS3ORA-nobsNOSOLONOARGO
Global T3000
2001 2002 2003 2004 2005 2006Time
-0.02
-0.01
0.00
0.01
0.02
ORAS3ORA-nobsNOSOLONOARGO
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 26
Spatial distribution of trends in heat content
Taux (x 0.01N/m2)
Tauy (x 0.01N/m2)T300 (deg C)
1982-2006 mean minus 1959-1981 mean
How reliable are the trends in ERA40 winds?
SST (deg C)
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 27
Comparison with ocean observations
392
Observations: Oceanic Climate Change and Sea Level Chapter 5
decrease during 1980 to 1983. The 0 to 700 m layer cooled at a rate of 1.2 W m–2 during this period. Most of this cooling occurred in the PaciÞc Ocean and may have been associated with the reversal in polarity of the PDO (Stephens et al., 2001; Levitus et al., 2005c, see also Section 3.6.3). Examination of the geographical distribution of the differences in 0 to 700 m heat content between the 1977–1981 and 1965–1969 pentads and the 1986–1990 and 1977–1981 pentads shows that the pattern of heat content change has spatial scales of entire ocean basins and is also found in similar analyses by Ishii et al. (2006). The PaciÞc Ocean dominates the decadal variations of global heat content during these two periods. The origin of this variability is not well understood.
Based on model experiments, it has been suggested that errors resulting from the highly inhomogeneous distribution of ocean observations in space and time (see Appendix 5.A.1) could lead to spurious variability in the analysis (e.g., Gregory et al., 2004, AchutaRao et al., 2006). As discussed in the appendix, even in periods with overall good coverage in the observing system, large regions in Southern Hemisphere (SH) are not well sampled, and their contribution to global heat content variability is less certain. However, the large-scale nature of heat content variability, the similarity of the Levitus et al. (2005a) and the Ishii et al. (2006) analyses and new results showing a decrease in the
global heat content in a period with much better data coverage (Lyman et al., 2006), gives conÞdence that there is substantial inter-decadal variability in global ocean heat content.
5.2.2.3 Implications for Earth’s Heat Balance
To place the changes of ocean heat content in perspective, Figure 5.4 provides updated estimates of the change in heat content of various components of the Earth’s climate system for the period 1961 to 2003 (Levitus et al., 2005a). This includes changes in heat content of the lithosphere (Beltrami et al., 2002), the atmosphere (e.g., Trenberth et al., 2001) and the total heat of fusion due to melting of i) glaciers, ice caps and the Antarctic and Greenland Ice Sheets (see Chapter 4) and ii) arctic sea ice (Hilmer and Lemke, 2000). The increase in ocean heat content is much larger than any other store of energy in the Earth’s heat balance over the two periods 1961 to 2003 and 1993 to 2003, and accounts for more than 90% of the possible increase in heat content of the Earth system during these periods. Ocean heat content variability is thus a critical variable for detecting the effects of the observed increase in greenhouse gases in the Earth’s atmosphere and for resolving the Earth’s overall energy balance. It is noteworthy that whereas ice melt from glaciers, ice caps and ice sheets is very important in the sea level budget
Figure 5.3. Linear trend (1955Ð2003) of zonally averaged temperature in the upper 1,500 m of the water column of the Atlantic, PaciÞc, Indian and World Oceans.The contour interval is 0.05¡C per decade, and the dark solid line is the zero contour. Red shading indicates values equal to or greater than 0.025¡C per decade and blue shading indicates values equal to or less than Ð0.025¡C per decade. Based on the work of Levitus et al. (2005a).
ORA-S3 IPCC-AR4 (LEVITUS)CI=0.05 deg/decade
Similarities
•Equatorial cooling
•Warmer subtropics
•Cooling at ~60N
Comments
•Trends in ERA40 winds seem robust
•Stronger features in ORA-S3, more structure
•Circulation changes as well as mixed layer changes
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 28
Comparison with ocean observations
392
Observations: Oceanic Climate Change and Sea Level Chapter 5
decrease during 1980 to 1983. The 0 to 700 m layer cooled at a rate of 1.2 W m–2 during this period. Most of this cooling occurred in the PaciÞc Ocean and may have been associated with the reversal in polarity of the PDO (Stephens et al., 2001; Levitus et al., 2005c, see also Section 3.6.3). Examination of the geographical distribution of the differences in 0 to 700 m heat content between the 1977–1981 and 1965–1969 pentads and the 1986–1990 and 1977–1981 pentads shows that the pattern of heat content change has spatial scales of entire ocean basins and is also found in similar analyses by Ishii et al. (2006). The PaciÞc Ocean dominates the decadal variations of global heat content during these two periods. The origin of this variability is not well understood.
Based on model experiments, it has been suggested that errors resulting from the highly inhomogeneous distribution of ocean observations in space and time (see Appendix 5.A.1) could lead to spurious variability in the analysis (e.g., Gregory et al., 2004, AchutaRao et al., 2006). As discussed in the appendix, even in periods with overall good coverage in the observing system, large regions in Southern Hemisphere (SH) are not well sampled, and their contribution to global heat content variability is less certain. However, the large-scale nature of heat content variability, the similarity of the Levitus et al. (2005a) and the Ishii et al. (2006) analyses and new results showing a decrease in the
global heat content in a period with much better data coverage (Lyman et al., 2006), gives conÞdence that there is substantial inter-decadal variability in global ocean heat content.
5.2.2.3 Implications for Earth’s Heat Balance
To place the changes of ocean heat content in perspective, Figure 5.4 provides updated estimates of the change in heat content of various components of the Earth’s climate system for the period 1961 to 2003 (Levitus et al., 2005a). This includes changes in heat content of the lithosphere (Beltrami et al., 2002), the atmosphere (e.g., Trenberth et al., 2001) and the total heat of fusion due to melting of i) glaciers, ice caps and the Antarctic and Greenland Ice Sheets (see Chapter 4) and ii) arctic sea ice (Hilmer and Lemke, 2000). The increase in ocean heat content is much larger than any other store of energy in the Earth’s heat balance over the two periods 1961 to 2003 and 1993 to 2003, and accounts for more than 90% of the possible increase in heat content of the Earth system during these periods. Ocean heat content variability is thus a critical variable for detecting the effects of the observed increase in greenhouse gases in the Earth’s atmosphere and for resolving the Earth’s overall energy balance. It is noteworthy that whereas ice melt from glaciers, ice caps and ice sheets is very important in the sea level budget
Figure 5.3. Linear trend (1955Ð2003) of zonally averaged temperature in the upper 1,500 m of the water column of the Atlantic, PaciÞc, Indian and World Oceans.The contour interval is 0.05¡C per decade, and the dark solid line is the zero contour. Red shading indicates values equal to or greater than 0.025¡C per decade and blue shading indicates values equal to or less than Ð0.025¡C per decade. Based on the work of Levitus et al. (2005a).
ORA-S3 IPCC-AR4 (LEVITUS) CI=0.05 deg/decade
Atlantic and Indian
Largest warming is in the Atlantic
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 29
Attribution of Sea level changes
Trends
1961-2003: SL (IPCC) =1.8 mm/yr
SH (IPCC) =0.5 mm/yr
SH ORA-S3 (1960-2003)=0.9mm/yr
SH ORA-nobs “ =0.5mm/yr
ORAS3 gets closer…..
1993-2003: SL (IPCC) =3.1 mm/yr
SH (IPCC) =1.6 mm/yr
SH ORA-S3 (1993-2003)=2.1mm/yr
SH ORA-nobs “ =1.1 mm/yr
consistent with others
2002 onwards??Effect of ARGO?
Altimeter problems?
Sea level changes= Mass + Volume (SH)
Steric Height (SH) can be estimated from ORAS3
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 30
Attribution of Sea Level Change (OSES)Steric height contribution to sea level rise
2002 2003 2004 2005 2006 2007Time
-0.020
-0.015
-0.010
-0.005
0.000
0.005
ORA-S3ORA-nobsNOSOLONOARGO
•Argo is responsible for the decay in SH in ORAS3
•SOLO/FSI have little impact
•But even without Argo, the trend in SH stabilizes after 2002
•While the SL from altimeter keeps increasing…If we believe the altimeter
•This would imply a mass increase of 2mm/yr (twice as large as the latest IPCC)
•Worrying: either the estimates are wrong, or a lot of continental ice is melting
Mass contribution to sea level rise
2002 2003 2004 2005 2006 2007Time
-0.01
0.00
0.01
0.02
0.03
ORA-S3ORA-nobsNOSOLONOARGOIPCC-AR4
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 31
Impact of data assimilation in the MOC
ORAS3
ORA-nobs
ORAS3
ORA-nobs
Bryden05
Cunningham07
•Assimilation improves the estimation of the MOC
•Downward trend ~4% decade in ORAS3, ~2% decade in ORA-nobs
RMS fit to observations in the NATL
Balmaseda et al, GRL 2007
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 33
SummaryState estimation:
- Both ARGO temperature and salinity have a large information content.
- Argo is instrumental in correcting the salinity of the ORA-S3 analysis
- The ARGO data is best used in combination with the altimeter information.
Seasonal forecast skill:- Argo/Altimeter/Moorings contribute to the improvement of the skill of seasonal forecast of SST.
- Their contribution is often complementary: Argo has larger effect in the Western Pacific and Indian Ocean. Altimeter’s impact is larger in Atlantic and Eastern Pacific
Climate variability:- The profound impact of Argo on the analysis should be taken into account when analysing the climate variability
from ORA-S3.
- OSEs indicate a deceleration in the ocean warming and global SH after 2003.
- The variability in the ORA-S3 salinity may not be reliable
Other comments:- A new observing system SHOULD NEVER HAVE a negative impact.
- In the Seasonal Forecast, the inability to improve predictions in the Equatorial Atlantic is symptomatic of errors in the model/analysis.
- In future reanalysis, the information provided by Argo could be used in retrospect, for instance via bias-correction algorithms (or improved models).
Magdalena A. Balmaseda, OSE Workshop, Paris 5-7 November 2007
Slide 34
What if the Observations have negative impact?
In the Analysis?
- Model error not taken into account
- Wrong Specification of Background error
- Wrong Specification of Observation error
In the forecast?
- The analysis error has not been reduced
- The analysis error has been reduced in total, but the error has increased in the directions of larger error growth.