Linking sea surface temperature, surface
flux, and heat content in the North
Atlantic: what can we learn about
predictability?
LuAnne Thompson
School of Oceanography
University of Washington
Kathryn Kelly (UW/APL), James Booth (NASA/GISS),
Maylis Garcia (NESTA, Paris)
NASA Ocean Surface Topography Science Team
Question:
Can the heat content be used to predict surface heat
flux in the North Atlantic? Where and when?
How does it compare against SST?
Answer
Use sea level as a proxy for upper ocean heat content
to find:
Basin wide interannual relationships
Basin wide seasonal relationships
Net surface flux from the atmosphere to
the oceans Watt/m2: implied ocean heat
transport convergence, maximum
around 30-40N
Out of the ocean
Into the ocean
Role of heat transport: mean balance
Heat is lost from the Gulf Stream to the
atmosphere (about 0.2 PWatts)
Maximum ocean meridional heat transport
about 1 PW at about 20-25N
Heat released
to atmosphere
Heat
transported by
the Ocean
Heat Budget (temporally varying)
dTocean
rocean H ocean c p
= rocean H ocean c pÑ iuT + Qsurface
dt
Heat storage
rate in upper
ocean
=
Heat Transport
Convergence
+
Heat exchanged
with atmosphere
UinTin
Heat
transported by
the Gulf
Stream
Tocean
UoutTout
Surface heat
flux
Local regional heat budget on interannual
time scales behave similarly to the mean
Dong and Kelly (2004) and Kelly and Dong
(2004) for Gulf Stream
Heat Transport
Convergence
Heat stored
in ocean
Heat loss to
atmosphere
Tocean increases
Heat transport convergence interannual variations on regional
scales much larger than the surface heat flux
Ratio of interannual surface heat flux to heat storage rate from
ECCO2 in North Atlantic (300 km Gaussian smoother)
Using observations to look at the relationship
between the heat content and surface flux smooth
both with 500 km Gaussian smoother
Observational
Analysis
variables
Sea surface
height (SSH)
Turbulent heat
flux Qturb
And net suface
heat flux Qnet
Source
Comment
Monthly maps of sea
level anomaly from
Ssalto/Duacs
1/3° x 1/3°, Mercator
grid, Aviso
OAflux: Objectively
Analyzed air-sea fluxes
for the Global Oceans
(Yu and Weller, 2007)
ISCCP for radiative
fluxes
Used as proxy for
upper ocean heat
content
Fluxes are
positive for
warming the
ocean.
Using sea level as a proxy for heat content.
1993-1999 (see also Jayne et al 2003)
Local sea level determined by thermosteric (thermal
expansion), and halosteric (haline contraction).
Thermosteric dominates in tropics and subtropics
Willis and
Roemmich
(2004). Seasonal
correlations
between upper
700 m ocean
heat content
(mostly XBTs)
and sea level,
1993-1999
Interannual relationships between SSH and Heat flux
60oN
100
90
50oN
80
70
2
o
40 N
60
1
50
o
30 N
40
3
30
o
20 N
20
o
80 W
Sea Surface Height m
Mode Water
o
40 W
North GS
o
0
10
Eastern Basin
5
50
0
0
−5
1995 2000 2005 2010
1995 2000 2005 2010
1995 2000 2005 2010
−50
−Turbulent heat flux W/m2
o
10 N
Sea Surface Height m
5
0
−5
1995
Mode Water: sea level leads turbulent heat flux
by about 5 months
2000
4
Lag Correlations between sea level (SST)
−1 8. 41.5, −68.5, −5
and surface
flux in mode water:
both lead the surface flux by a few months
12.
36.5,
−62.5,
−4
9. 48.5,
−25.5,
10. 29.5,
−30.5, −5
on interannual
time−4
scales. (mode
water)
1
Sea Level auto
correlation
7. 41.5,
Heat
flux −54.5
auto
11.
13.5, −40.5, −5
Correlation
Lagged
correlation: sea
level heat flux
0.5
6. 38.5, −37.5
0 10. 29.5, −30.5, −8
−0.5
Correlation
4
−0.5
Sea level leads
−1
−24 −12
On monthly time scales,
SSH has longer
11. 13.5, −40.5, −5
persistence and shows
more predictive skill than
SST.
8. 41.5, −68.5
12. 36.5, −62.5, −1
Heat Flux Leads
0
12
24
−2412.
−12
12 −5
24
36.5,0−62.5,
0
12
24
−24 −12
Sea level leads
−24 −12
0 −30.5
12 24
10. 29.5,
44 −24
−24−12
−120 012
−24 −12
0
Heat Flux
SST auto
Leads
correlation
−24 −12
0 −40.5
12 24 −24 −12
0 −62.5
12 24
11. 13.5,
12. 36.5,
Lagged
correlation: SSH
heat flux
Lagged
correlation: SST
heat flux
12 −24
24−12
24
0
12
24
−24 −12
0
12
24
12
Sea Level
auto
correlation
Heat flux auto
Correlation
12 24 −24 −
Lagged
correlation:
sea level heat
flux
SST auto
correlation
Heat flux auto
correlation
Lagged
correlation: SST
heat flux
Correlation of low frequency turbulent heat flux: typically SSH leadsby 3-5
months
SST has more skill than SSH up to one year ahead
SSH has more skill more than a year ahead
0.7
0.7
SSH 50oN
0.6
0.6
Correation
60oN
o
0.5
30 N
o
0.4
20 N
o
0.3
0.3
10 N
o
0.2
0.2
60oN
0.7
0.7
SST 50ooN
60 N
0.6
12
0.6
24
40
50oN
N
o
0.5
10
o
8
0.4
o
30
40 N
N
o
o
20
30 N
N
o
o
10
20 N
N
6
0.3
4
0.2
2
Months ofCorreation
lead
40 N
0.5
0.4
0.5
22
20
0.4
18
0.3
16
0.2
14
Correlation
Second year
Correlation
Months of
lead
First year
Question: where does the 3-5 month lead come
from? Could be seasonality in the
relationship?
Investigate the seasonal relationships between
heat content and surface flux
1. Make 12 times series, one for each month, of
anomalous heat content and surface flux
2. Investigate heat content and surface flux
persistence
3. Investigate correlations between heat
content and surface flux between different
seasons
Mode Water
North GS
Eastern Basin
Sea Surface Height m
-Turbulent heat flux Watt m-2
2010
2000
March
5
2000
July
0
−5
December
Perform the lagged correlations between surface flux and heat
content and look for the places where SSH predicts surface flux
(warmer ocean predicts heat flux out of the ocean)
Mode Water
December
North GS
July
Eastern Basin
March
Perform the lagged correlations between surface flux and heat
content and look for the places where SST predicts surface flux
(warmer ocean predicts heat flux out of the ocean)
Mode Water
December
North GS
July
Eastern Basin
March
Look for longest run of significant correlations for each months at
each location
Length of run
Months of
predicatiblity
correlation
Correlation
Regions and numbers of months of predictability for
turbulent heat flux from sea level (heat content) for each
month of the year
Length of patch of correlation
Correlation higher for SSH than SST in some places
SSH
SST
SSH generally has more predictability than SST
SSH
SST
Length of patch of correlation: generally longer for SSH
than for SST
SSH
SST
Conclusions
1.
The 20 year satellite altimeter record allows investigation of the role of regional
upper ocean heat content in ocean-atmosphere interactions.
2.
The heat content (and SST) in the ocean is correlated with surface flux on
interannual time scales in the Gulf Stream, with SSH leading by 3-5 months with
a warmer ocean fluxing heat from the ocean to the atmosphere
3.
Both SST and SSH have predictive skill throughout the year and throughout the
basin
SST and SSH predictive skill depends on season
SST and SSH have different patterns of seasonal predictive skill
4.
The length of the runs of high correlations and the correlation coefficient are
generally higher for SSH than for SST
5.
Most of the signal comes from latent heat flux, with the correlations also seen in
net surface heat flux (not shown).
6.
Mode water and northern recirculation gyre show large correlations between
SSH and surface flux with heat content in the mode water controlling December
surface flux, and northern recirculation gyre controlling July surface flux
Implication for coupling between ocean and
atmosphere: preliminary results for clouds
1.
Minobe et al (2008, 2010) showed that in the climatological mean, south of the
GS, the SST forces the formation of mid-level clouds in winter, while north of the
Gulf Stream is a region of stratus formation in the summer.
2.
We find:
January latent heat flux correlated with mid-level cloud fraction 0.75 (0.29)
July Latent heat flux (-) correlated with low-level cloud fraction -0.45 (0.29)
3.
Local forcing of deep heating in summer not obvious in SST, possibly in SSH
X
X