See http://eleanorfrajka.com/moc-from-space/ Slides from IUGG meeting in Prague: Estimating the Atlantic overturning circulation at 26N from satellite altimetry.
Estimating the Atlantic overturning at 26N using satellite altimetry [IUGG]
1. Eleanor Frajka-Williams (Univ of Southampton)
Grace (NASA/JPL)
RRS Discovery
1
Estimating the Atlantic
overturning at 26N
using satellite altimetry
[IUGG general assembly in Prague, Jun 2015]
Questions? @EleanorFrajka
2. [Kulbrodt et al, 2007]
Overturning circulation
2
RAPID-MOCHA project:
Observations of the time-varying large-scale ocean circulation
Funded by UK NERC, NSF and NOAA
3. Single value (the MOC) or components?
• Components help us understand where and why the MOC is changing
• But the actual value of the MOC is also important
3
What do we really want to know?
Volume or Heat transport?
MOC timescales of variability:
• Eddies on 20-100 day timescales (Clement et al. 2014; Frajka-Williams et al. 2013)
• Wind-variability on interannual timescales (Yang & Johns 2014)
• Buoyancy-driven variability …?
[Johns et al., 2011]
4. [Frajka-Williams 2015]
4
In this talk:
Introduce a proxy for the MOC at 26N
that recovers over 90% of the
interannual variability of
the RAPID time series from 2004-2014.
Tell you why it doesn’t replace the in
situ observations.
5. Data: RAPID transbasin transport
5
MOC = EK + GS + UMO
For details of the method, see McCarthy et al. 2015, Measuring the MOC
EK (meridional Ekman) from ERA-Interim
GS (Gulf Stream) from Florida Cable
UMO (upper mid-ocean transport, Bahamas to Africa) from
current meter & dynamic height moorings
6. Data: RAPID transbasin transport
6
MOC = EK + GS + UMO
For details of the method, see McCarthy et al. 2015, Measuring the MOC
EK (meridional Ekman) from ERA-Interim
GS (Gulf Stream) from Florida Cable
UMO (upper mid-ocean transport, Bahamas to Africa) from
current meter & dynamic height moorings
7. Method
Temporal:
Remove seasonal cycle
1.5 year Tukey filter
7
AVISO Sea level
anomaly (SLA):
RAPID upper mid-ocean transport time series (UMO):
Focus on the interannual variability…
Remove eddies…
Spatial:
Smooth (5x10 deg):
Regress RAPID UMO against SLA
9. 9
[Frajka-Williams 2015]
[Frajka-Williams 2015]
UMO transport is proportional to thermocline depth at
the west.
Deeper (more negative) thermocline depth means
stronger (more negative) UMO transport.
SLA vs transbasin transport UMO
10. 10
[Frajka-Williams 2015]
UMO transport is proportional to thermocline depth at
the west.
2 cm change in SLA results in a 1 Sv change in UMO
SLA vs transbasin transport UMO
[Frajka-Williams 2015]
11. [Frajka-Williams 2015]
From SLA: MOC* = EK + GS + UMO*
Using SLA for UMO, determine MOC
11
From RAPID: MOC = EK + GS + UMO
EK from ERA-Interim since 1979
GS from Florida Cable since 1982
UMO* from SLA since 1993
[Frajka-Williams 2015]MOC* since 1993
12. 12
This MOC* recovers over 90% of the variability of the RAPID MOC.
(note: the two are not independent since both use the same GS and Ek.)
Can we just use SLA to investigate longer term MOC changes?
[Frajka-Williams 2015]
Using SLA for UMO, determine MOC
13. Single value (the MOC) or components?
• Components help us understand where and why the MOC is changing
• But the actual value of the MOC is also important
13
Recall:What do we really want to know?
Volume or Heat transport?
MOC timescales of variability:
• Eddies on 20-100 day timescales (Clement et al. 2014; Frajka-Williams et al. 2013)
• Wind-variability on interannual timescales (Yang & Johns 2014)
• Buoyancy-driven variability …?
[Johns et al., 2011]
14. To date, MOC interannual variability has been dominated by wind-forcing
(debatable, but evidence suggests yes).
This is consistent with model-based studies (e.g., Yeager 2015; Pillar et al. 2015)
• RAPID observations demonstrate that most of the interannual variability originates in Ekman & UMO transport.
• SLA reconstruction works because UMO-SLA relationship is strong.
Buoyancy-driven variability occurs on longer time scales
(e.g., Yeager 2015; Pillar et al. 2015)
• Under buoyancy forcing/on longer timescales, not clear that the UMO-SLA relationship would be as strong.
14
Why not just use SLA proxy?
15. The SLA proxy provides a 20-year proxy for MOC variability.
IF the SLA-UMO relationship is stationery,
then we can use it to look at
lower frequency MOC changes.
Suggests that:
• Trend over 2004-2014 does not
continue back in time
• Moderate reduction (1 Sv) between
1994 decade & 2004 decade
[Frajka-Williams 2015]
15
Even so…
Thank you!
See: http://eleanorfrajka.com/moc-from-space Questions? @EleanorFrajka