Content uploaded by Gennady A. Chepurin
Author content
All content in this area was uploaded by Gennady A. Chepurin on Mar 07, 2016
Content may be subject to copyright.
F
EBRUARY
2000 311CARTON ET AL.
q2000 American Meteorological Society
A Simple Ocean Data Assimilation Analysis of the Global Upper Ocean 1950–95.
Part II: Results
J
AMES
A. C
ARTON
,G
ENNADY
C
HEPURIN
,
AND
X
IANHE
C
AO
Department of Meteorology, University of Maryland at College Park, College Park, Maryland
(Manuscript received 6 February 1998, in final form 10 February 1999)
ABSTRACT
The authors explore the accuracy of a comprehensive 46-year retrospective analysis of upper-ocean temper-
ature, salinity, and currents. The Simple Ocean Data Assimilation (SODA) analysis is global, spanning the
latitude range 628S–628N. The SODA analysis has been constructed using optimal interpolation data assimilation
combining numerical model forecasts with temperature and salinity profiles (MBT, XBT, CTD, and station), sea
surface temperature, and altimeter sea level. To determine the accuracy of the analysis, the authors present a
series of comparisons to independent observations at interannual and longer timescales and examine the structure
of well-known climate features such as the annual cycle, El Nin˜o, and the Pacific–North American (PNA)
anomaly pattern.
A comparison to tide-gauge time series records shows that 25%–35% of the variance is explained by the
analysis. Part of the variance that is not explained is due to unresolved mesoscale phenomena. Another part is
due to errors in the rate of water mass formation and errors in salinity estimates. Comparisons are presented to
altimeter sea level, WOCE global hydrographic sections, and to moored and surface drifter velocity. The results
of these comparisons are quite encouraging. The differences are largest in the eddy production regions of the
western boundary currents and the Antarctic Circumpolar Current. The differences are generally smaller in the
Tropics, although the major equatorial currents are too broad and weak.
The strongest basin-scale signal at interannual periods is associated with El Nin˜o. Examination of the zero-
lag correlation of global heat content shows the eastern and western tropical Pacific to be out of phase (correlation
20.4 to 20.6). The eastern Indian Ocean is in phase with the western Pacific and thus is out of phase with the
eastern Pacific. The North Pacific has a weak positive correlation with the eastern equatorial Pacific. Correlations
between eastern Pacific heat content and Atlantic heat content at interannual periods are modest. At longer
decadal periods the PNA wind pattern leads to broad patterns of correlation in heat content variability. Increases
in heat content in the central North Pacific are associated with decreases in heat content in the subtropical Pacific
and increases in the western tropical Pacific. Atlantic heat content is positively correlated with the central North
Pacific.
1. Introduction
Recently we have reported a new retrospective anal-
ysis of upper-ocean temperature, salinity, sea level, and
currents for the global ocean, 1950–1995 (Carton et al.
2000). The result of that research has been to provide
a synthesis of the historical oceanographic data record
in the form of a retrospective analysis. Here we describe
the results of an exploration of the accuracy of this
analysis.
The Simple Ocean Data Assimilation (SODA) algo-
rithm used to construct our retrospective analysis con-
sists of a numerical forecast model and an update pro-
cedure to provide corrections to the forecast. The update
Corresponding author address: Dr. James A. Carton, Department
of Meteorology, University of Maryland at College Park, 2417 Com-
puter and Space Science Building, College Park, MD 20742-2425.
E-mail: carton@metosrv2.umd.edu
procedure is based on a data assimilation algorithm
widely applied in atmospheric numerical weather pre-
diction called optimal interpolation (see Daley 1991).
In optimal interpolation the differences between ob-
served variables such as temperature, salinity, and sea
level and model forecasts of the same variables are used
to update the forecast. The interpolation coefficients,
otherwise known as gain matrices, are determined so as
to minimize the mean square error of the analysis. Our
implementation of this algorithm differs from the usual
implementation to the extent that we include spatial de-
pendence of the assumed error statistics and because of
our assumptions about bias in the model forecast.
SODA differs from the more sophisticated but much
more computationally intensive Kalman filter to the ex-
tent that we do not have prediction equations for the
temporal and spatial evolution of the error statistics.
Rather, the error statistics are determined a priori based
on a statistical analysis of errors from a preliminary
analysis. Our approach also bears similarity to 3D var-
312 V
OLUME
30JOURNAL OF PHYSICAL OCEANOGRAPHY
T
ABLE
1. Ocean analysis experiments presented in the text. Each
experiment covers the period 1950–96. Further description of these
experiments is provided in Carton et al. (2000).
Experiment Description
Control analysis Basic analysis with detrended winds
1 Basic analysis except assuming significant fore-
cast error bias
2 Simulation with no subsurface updating
3 Basic analysis with climatological monthly winds
4 Basic analysis with complete winds
5 Basic analysis except with salinity updating with
observed salinity, but without T/Serror covari-
ance
6 Basic analysis except without T/Serror covari-
ance
7 Basic analysis except replacing model with cli-
mate temperature
8 Basic analysis except that TOPEX/Poseidon al-
timeter sea level is excluded from updating
procedure
T
ABLE
2. Statistical comparison of control analysis and tide guage sea level for selected stations. Record length is given, along with overall
correlation (Cor1) and correlation of 5-yr low-pass filtered records (Cor2). Christmas Island record comes in two parts, each of which has
been detrended. The comparison is with the combined record. Correlations marked with an asterisk exceed the 95% test of significance.
Name Location Years Cor1 Cor2
Atlantic
San Juan (Puerto Rico)
Tenerife (Canary Is.)
Bermuda, Is.
La Coruna (Spain)
188279N, 668059W
288299N, 168149W
328229N, 648429W
438229N, 88249W
26
38
40
27
0.50*
20.02
0.46*
0.28
0.34
20.21
0.72*
0.31
Pacific
Rikitea Is.
Noumea II
Pago Pago Is.
Christmas Is. (two parts)
Pahnpei-b Is.
Majuro-b Is.
Kwajalein Is.
Yap Is.
Guam Is.
23889S, 1348579W
228189S, 1668269E
148179S, 1708419W
18599N, 1578299W
68599N, 1588149E
7869N, 1718229E
88449N, 1678449E
98319N, 138889E
138269N, 1448399E
20
22
40
14
17
17
45
21
41
0.15
0.52*
0.34*
0.86*
0.86*
0.79*
0.73*
0.66*
0.70*
0.43
0.77*
0.56*
0.88*
0.71*
0.81*
0.74*
0.91*
0.45*
Johnston Is.
Wake Is.
Hilo, Hawaii
Honolulu, Hawaii
French Frigate Shoals
Midway Is.
Sitka, Alaska
168459N, 1698319W
198179N, 1668379E
198449N, 155849W
218189N, 1578529W
238529N, 1668179W
288139N, 1778229W
57839N, 1358209W
41
38
45
45
17
35
35
0.66*
0.54*
0.44*
0.67*
0.44
0.07
0.07
0.40*
0.67*
0.35*
0.70*
0.77*
0.44*
0.44*
iational assimilation methods when the mean square er-
ror is minimized such as that described in Ji et al. (1995).
Comparison of the results shows correlations of heat
content exceeding 80% in the tropical Pacific (Chepurin
and Carton 1999). Bennett (1990) and Wunsch (1996)
provide comprehensive discussions of the alternatives
in formulating updating algorithms. Malonote-Rizzoli
(1996) reviews many current implementations.
The general circulation ocean model on which our
analysis is based uses the full Geophysical Fluid Dy-
namics Laboratory Modular Ocean Model 2.b primitive
equation code, with conventional choices for mixing,
etc. The domain of this analysis is global, extending
from 628Sto628N. The model horizontal resolution is
2.5830.58in the Tropics, expanding to a uniform 2.58
31.58resolution at midlatitude. No attempt is made to
resolve midlatitude eddy processes. At the polar bound-
aries the temperature and salinity fields are relaxed to
climatology. We make no attempt to model cryospheric
or deep-water formation processes explicitly. A weak
5-yr relaxation of the global temperature and salinity
fields is included in order to reduce forecast bias indeep-
water masses. Bottom topography is included. The mod-
el has 20 levels in the vertical, with 15-m resolution in
the upper 150 m. In this model sea level is obtained
through diagnostic calculation from the mass and mo-
mentum fields. Winds are provided by an analysis of
historical shipboard measurements by da Silva et al.
(1994) prior to 1993 and by the National Centers for
Environmental Prediction after that date.
The main datasets to constrain the model forecast are
the hydrographic data contained in the World Ocean
Atlas 1994 (WOA-94; Levitus et al. 1994), additional
hydrography obtained from a variety of sources, sea
surface temperature (Reynolds and Smith 1994), and
altimetry from the Geosat, ERS-1, and TOPEX/Posei-
don satellites. The hydrographic dataset exceeds 50 000
profiles per year for much of our 46-yr period of interest.
This dataset has been collected using several different
instruments. In the interval 1950–69, most of the tem-
perature measurements were made with the mechanical
bathythermograph. This subset reached a maximum of
70 000 measurements per year in 1968, but is limited
to sampling temperature in the upper 2–300 m. After
1969 the mechanical bathythermograph was largely re-
F
EBRUARY
2000 313CARTON ET AL.
F
IG
. 1. Observed and control analysis sea level time series at Kwa-
jalein Island (98N, 1688E). The seasonal cycle and a linear trend has
been removed from both records.
F
IG
. 2. Five-year low-pass filtered observed and control analysis
sea level at three locations: San Juan, Puerto Rico (188N, 668W),
Honolulu (218N, 1588W), and Kwajalein Island (98N, 1688E). A linear
trend has been removed from all records. Units are centimeters.
placed by the expendable bathythermograph. Two other
small subsets, conductivity–temperature–depth and bot-
tle measurements, are important because they provide
salinity as well as temperature and because they extend
more deeply into the water column.
The altimeter sea level used here is based on the
NASA Pathfinder Project version 2.1 to which we have
added all the standard corrections for geophysical ef-
fects and then averaged alongtrack into 18bins. No other
interpolation was carried out. The altimeter dataset be-
gins November 1986 with the Geosat Exact Repeat Mis-
sion. No altimeter data is available from the end of
Geosat in the fall of 1989 until the beginning of ERS-1
in spring 1992.
In addition to the basic analysis, which we refer to
as the control analysis, a series of analysis experiments
(see Table 1) has been carried out to determine some
of the properties and sensitivities of the analysis. The
control analysis and analysis experiments begin January
1950 and continue through December 1995. The anal-
ysis fields for each experiment consist of 552 monthly
averages of temperature, salinity, and the horizontal
components of velocity at 20 levels. Comparison of the
experiments reveals important sensitivities of the anal-
ysis to changes in parameters, boundary conditions, etc.
Some review of these results is provided in this paper.
A more extensive discussion is provided in Carton et
al. (2000).
Here our examination of analysis error focuses on
comparison to independent observations on interannual
and longer timescales. We limit our comparison to ex-
amination of correlations and root-mean-square (rms)
differences and a more detailed examination of one ex-
ample for each dataset. Section 2 examines the temporal
variability of the analysis by comparison to tide gauge
sea level time series from the Permanent Service for
Mean Sea Level. In Section 3 our examination focuses
on the spatial structure of analysis error. Datasets with
good spatial coverage include altimeter sea level, global
hydrographic sections, and surface drifter velocity.
2. Time series comparisons
In this section we present a comparison to island tide
gauge records. Comparison to temperature and salinity
at Bermuda is provided in Carton et al. (2000). Together
these datasets allow a detailed examination of the ac-
curacy of the analysis at interannual to decadal periods.
The behavior of the analysis at short seasonal periods
is also discussed in Carton et al. (2000).
Tide gauge comparison
Approximately 1800 tide gauge records are available
from the Permanent Service for Mean Sea Level or the
314 V
OLUME
30JOURNAL OF PHYSICAL OCEANOGRAPHY
F
IG
. 3. Sea level error estimated using three years of TOPEX/Poseidon altimetry. (a) Root-
mean-square difference between altimeter and expt 8 analysis sea level in which altimeter sea
level observations have been excluded from the analysis. (b) Rms difference between altimeter
and control analysis sea level in which temperature and salinity are constrained by altimeter
observations. Contour intervals are 2 and 1 cm.
Tropical Ocean Global Atmosphere sea level archives.
However, many gauges are in locations that are un-
suitable for observing the large-scale circulation, while
others have records that are too short for our purposes.
After examining the datasets we have identified 20 sta-
tions, mainly from islands in the North Atlantic and
Pacific (in the latitude band 238S–578N), with records
that each exceed 17 years in length and do not seem
excessively gappy or otherwise contaminated.
At each of the station locations we have carried out
a comparison between annually averaged observations
and control analysis sea-level time series. The compar-
ison is summarized in Table 2. Correlations between
observed and control analysis sea level is presented as
Cor1. The lowest correlations are for continental sta-
tions such as La Coruna, Spain, and for some of the
subtropical and midlatitude islands such as the Canary
Islands. The agreement at islands in the tropical Pacific
such as Kwajalein (98N, 1678E: Fig. 1) is generally ex-
cellent. A succession of high and low sea level events
in the record at Kwajalein reflects the importance of El
Nin˜o throughout the tropical Pacific. The earliest strong
event in this record is the El Nin˜o of 1957–58, which
is indicated by a rise in sea level followed by a 9-cm
drop. The strongest event overall is the 1982–83 El Nin˜o
when sea level dropped by 20 cm. The differences be-
tween observed and analyzed sea level appear to be of
longer than interannual timescale. We examine the de-
cadal behavior of these records below. Interestingly, the
anomaly correlations in Table 2 are quite comparable
to correlations reported previously by Miller and Cane
(1996) for much shorter 2-yr intervals that also included
the seasonal cycle.
In order to determine the quality of the comparison
at frequencies longer than interannual we next filter all
time series with a 5-yr low-pass filter. A linear trend is
also removed in order to eliminate unmodeled geo-
physical effects such as geologic uplift and global sea
level rise due to continental ice melt as well as model
bias. The resulting gauge comparisons are labeled Cor2
F
EBRUARY
2000 315CARTON ET AL.
T
ABLE
3. Rms differences between observed and control analysis
temperature and salinity along WOCE hydrographic sections.
Name Location Rms (T)
(8C) Rms (S)
(psu)
Atlantic
A1E
A11
A16
A9
AR4E
AR15
52.28N, 238W–148W
44.68S, 508W–128W
448–598N, 208W
198S, 368W–138W
58S–58N, 34.58W
68S–28N, 34.58W
0.43
1.29
0.32*
0.36*
0.93
0.74
0.047
0.195
0.031
0.156
0.138
0.113
Indian
I5 33.58S, 388–728E 0.45 0.058
Pacific
P1-2
P1-3
P4-2
P4-3
P4-4
P6W
478N, 1478W–1268W
478N, 1478W–1268W
9.38N, 1608E–1608W
9.38N, 1658W–1108W
9.38N, 1608E–1608W
308S, 1548E–1788E
0.29*
0.32*
0.62
0.54
0.51
0.66
0.130
0.091
0.074
0.062
0.053
0.052
P6C
P6E
P16S
P16C
P17S
P17C-2
P17C-3
32.38S, 1658W–1208W
32.38S, 1108W–888W
338S–178S, 150.38W
17.38S–9.38N, 150.68W
198C–68S, 1348W
68S–0.48N, 134.68W
0.48N–34.48N, 134.68W
0.55
0.56
0.50
0.72
0.51
0.44
0.82
0.060
0.073
0.067
0.116
0.124
0.067
0.105
PR3
PR13J
PR13N
PR16
PR20
348N–42.28N, 1448E
258N–488N, 1658E
43.28S, 1488E–1668E
18N–68N, 1108W
21.5N
1.36
0.78
1.41*
1.01
0.87
0.120
0.160
0.065
0.150
0.071
* Temperature (but not salinity) from these cruises has been as-
similated in the control analysis.
in Table 2. In two-thirds of the stations in the tropical
Pacific the correlation of observed and analysis sea level
is improved by low-pass filtering. In mid and high lat-
itude the correlation also improves, but generally re-
mains below 0.5. The arithmetic average correlation of
the two sets of records are 0.48 and 0.56. If these values
are representative they would suggest that the control
analysis explains 23% of the interannual sea level var-
iance, increasing to 31% on timescales between 5 and
25 years. A significant fraction of the unexplained var-
iance is due to mesoscale processes that cannot be re-
solved by our analysis and is of less interest to us (the
contribution of mesoscale variability is evident in the
comparison of the nearby Honolulu and Hilo time se-
ries).
The three pairs of low-pass filtered time series shown
in Fig. 2 include one (San Juan) that has a correlation
of less than 0.5 and two (Honolulu and Kwajalein) that
exceed 0.7. At all three stations a visual comparison
reveals substantial similarity. The main differences at
San Juan seem to be in the specific timing of the anom-
alies rather than their amplitude or duration. At Hon-
olulu the major differences between observed and anal-
ysis sea level occur prior to 1960, when the subsurface
data coverage was less complete.
3. Additional comparisons
The comparisons reported in this section involve da-
tasets with more limited temporal coverage but with
expanded spatial coverage (altimeter sea level, WOCE
hydrography, and drifter and moored currents).
a. Altimeter sea level
Although the time series comparisons discussed
above provide information about interannual to decadal
variability, they have limited spatial coverage and have
only indirect information about currents. The availabil-
ity in recent years of satellite altimeter sea level offers
us the opportunity to examine the accuracy of the anal-
ysis essentially globally. In order to make the compar-
ison to altimeter sea level independent we introduce a
new experiment, experiment 8, in which altimeter sea-
level information has been excluded from the updating
procedure. We limit our discussion hereto consider only
the TOPEX/Poseidon altimetry because of its low ob-
servation error.
The error in observed monthly averaged TOPEX/Po-
seidon altimetric sea level has been estimated to be in
the neighborhood of 2 cm in the tropical Pacific (Cheney
et al. 1994; Mitchum 1994). The rms difference between
observed and experiment 8 sea level exceeds the ob-
servation error by 2–4 cm (Fig. 3a). The difference is
lowest in the Tropics and somewhat loweron the eastern
side of the basin than on the western side. In the eastern
tropical basin the rms difference drops below 3 cm. For
the whole tropical belt 158S–158N the rms difference is
around 4.0 cm.
The rms difference increases in regions of high eddy
generation such as the regions of western boundary cur-
rent extensions and in the Antarctic Circumpolar Cur-
rent. A few ‘‘bull’s-eyes’’ appear in Fig. 3a. Close ex-
amination shows that these result from mislocated XBTs
that are then inconsistent with altimeter sea level. One
example of a mislocated XBT is at 188N, 1858W in Fig.
3a. The rms difference for the full 628S–628N domain
is 5.2 cm.
When altimetric sea level is used as a constraint on
the analysis temperature and salinity fields (but not pres-
sure since pressure is not a prognostic model variable),
the rms difference is reduced by 1–2 cm (Fig. 3b). In
the tropical Atlantic and Pacific the differences lie in
the range 2–3 cm. The average for the full tropical belt
158S–158N becomes 3.1 cm. In the mesoscale eddy pro-
duction regions of the midlatitudes the rms difference
again increases due to limitations in model physics and
poorer error statistics.
b. WOCE hydrographic transects
The World Ocean Circulation Experiment is an in-
ternational program designed to define the large-scale
316 V
OLUME
30JOURNAL OF PHYSICAL OCEANOGRAPHY
F
IG
. 4. Temperature and salinity with depth along WOCE meridional transect P17 near longitude 1348W
in the eastern Pacific Ocean during June–July 1991. The observed transect was made of three sections. Upper
panel shows the difference between observed and control analysis, middle panel shows observations, lower
panel shows analysis. (a) Temperature and (b) salinity.
structure of the ocean. The field program, including a
broad array of observations, has been concentrated dur-
ing the period since 1989. A primary dataset consists
of a series of high quality one-time hydrographic sur-
veys transecting the major oceans with some repeat sec-
tions. With the exceptions noted below, this dataset was
not included in the data archive of Levitus et al. (1994)
and consequently provides us with wonderful indepen-
dent data for comparison.
Not all of the WOCE hydrography is publicly avail-
able. From the more limited set of data available from
the Scripps Institution of Oceanography mirror site of
the WOCE hydrographic program Special Analysis
Center in Hamburg, we have extracted 16 transects list-
ed in Table 3. The temperature and salinity data along
each transect has been linearly interpolated on constant
depth surfaces to the model grid coordinates and then
compared to the monthly averaged temperature and sa-
linity fields from the control analysis. Along each sec-
tion the rms difference between observed and control
analysis temperature and salinity has been computed at
the model levels and averaged in depth and distance
F
EBRUARY
2000 317CARTON ET AL.
F
IG
.4.(Continued)
T
ABLE
4. Rms near-surface anomalous observed velocity compo-
nents and rms differences between anomalous observed and analysis
velocity components. Zonal and meridional anomalous velocity com-
ponents are computed with respect to the 1988–93 monthly clima-
tology and averaged over the Pacific basin between 308S and 408N.
Units: centimeters per second.
Year Rms (u9) Rms (
y
9) Rms (Du9) Rms (D
y
9)
1988
1989
1990
1991
1992
1993
1988–93
25.4
20.4
19.0
21.3
20.6
21.5
20.6
15.2
13.3
12.7
12.1
12.9
12.5
12.4
16.7
17.0
13.5
14.1
13.5
12.1
13.3
11.3
10.1
9.3
9.5
8.9
8.9
8.9
along the cross section. In some cases the transectshave
been decomposed into smaller sections when they span
more than a single month.
The large error in temperature for A11 can be un-
derstood because of its location in the poorly sampled
Southern Hemisphere and the high degree of eddy var-
iability. The errors for PR3 in the subtropical North
Pacific and PR16 in the tropical Pacific are more sur-
prising. The transects with large salinity errors generally
have two kinds of errors. Either the salinity errors are
large in the mixed layer (PR3, AR15, AR4E), indicating
problems with surface fluxes, or the errors are confined
below the mixed layer. Two of the transects with sub-
stantial salinity errors below the mixed layer (P1,
318 V
OLUME
30JOURNAL OF PHYSICAL OCEANOGRAPHY
F
IG
. 5. Comparison of anomalous observed and control analysis near-surface annual-averaged anomalous
currents in the tropical Pacific for the year 1991. Upper panel shows observations the control analysis, while
lower panel shows currents from currents based on drifter. Anomalous currents have been computed relative
to the 1988–93 monthly climatology. Observed and analysis tropical currents show a distinctive eastward
anomaly during this year.
PR13J) do not have large corresponding errors in tem-
perature, suggesting that the salinity errors are the result
of errors in horizontal advection.
One particularly interesting transect, labeled P17, cuts
through the eastern Pacific from 198Sto348N. This
transect was broken into several segments that are sep-
arated by vertical lines in Figs. 4a,b. At subtropical
latitudes comparison of observed and analysis temper-
ature fields shows that the most significant error is at
thermocline depths. Analysis temperature is too low by
18–28C suggesting that the thermocline is too shallow
by 10–20 m. The largest error is between 38N and 88N.
This band of latitudes corresponds to the North Equa-
torial Countercurrent trough that separates the northern
extension of the South Equatorial Current from the
North Equatorial Countercurrent. Weakness in the
F
EBRUARY
2000 319CARTON ET AL.
F
IG
. 6. Comparison of observed and control analysis zonal current
time series on the equator with depth at 1408W in the eastern Pacific.
Note the presence of strong interannual variability associated with
El Nin˜o.
trough implies that the transports in these two currents
are weak.
Comparison of observed and analysis salinity fields
also shows that significant error is present at thermocline
depths as well as in the mixed layer (Fig. 4b). South of
the equator the observed transect shows evidence of
subduction and equatorward transport of high salinity
subtropical water. The observed salinity maximum is
narrowly confined in depth within a few degrees of the
equator. In contrast, the analysis salinity shows a broad-
er, weaker subsurface salinity maximum. This difference
indicates that the analysis is subducting subtropical wa-
ter, but not as rapidly as is observed. One factor may
be that the mixed layer salinity reaches a maximum of
36.2 psu, 0.2 psu lower than observed. In contrast, the
analyzed mixed layer salinity between 68and 128Nis
more than 0.2 psu higher than observed. In summary,
the major sources of error include those associated with
errors in the mixed layer and subduction processes,
large-scale bias, and unresolved mesoscale variability.
c. Surface drifters
The most extensive spatial coverage of velocity mea-
surements is provided by the WOCE/TOGA surface
drifter velocity program (Niiler et al. 1999, manuscript
submitted to J. Phys. Oceanogr.). Although some mea-
surements were collected in the early 1980s, extensive
coverage is only available since 1988. We haveobtained
the data presented here from the Atlantic Oceanographic
Marine Laboratory/NOAA, where the drifter data has
been converted to Eulerian velocities and averaged into
8832831 month bins. We compute velocity anomalies
by removing the 1988–93 monthly climatology from the
observations and analyses. Table 4 shows the spatial
average of the rms velocity anomaly components, as
well as the rms differences between observed and anal-
ysis anomalous velocity components for each year. We
approach this comparison with trepidation since near-
surface velocity is a difficult field to simulate. The 6-
yr averages show that the rms anomalous zonal velocity
difference (13.3 cm s
21
) is 30% less than the rms zonal
velocity itself (20.6 cm s
21
). Some similarity in ob-
served and analysis variability is present in almost every
year in both zonal and meridional components.
The spatial pattern of anomalous velocity for 1991,
the year of the beginning of a strong El Nin˜o, is shown
in Fig. 5. The surface velocity during this year is char-
acterized by a strong 10 cm s
21
eastward surge along
and just north of the equator in response to therelaxation
of the trade winds [see Frankignoul et al. (1996) for
discussion of the surge]. South of the equator and north
of 108N the velocity is weakly westward. In the extra-
tropics the observed velocity components are confused.
The control analysis velocity also shows an eastward
surge of water of somewhat higher amplitude than ob-
served, extending not quite as far toward the coast of
South America. North and south of this surge westward
return currents are apparent, as observed. The analysis
velocity in the extratropics is of lower amplitude than
observed. We think that the reduced amplitude of an-
alyzed velocity at the oceanic mesoscale is the result of
dissipation of mesoscale eddies by the forecast model.
d. Equatorial Pacific moored current
The anomalous eastward surface velocity at the equator
during 1991 has corresponding subsurface changes. InFig.
6 we compare observed current from the Tropical Ocean–
Atmosphere mooring maintained by the Pacific Marine
Environmental Laboratory at 1408W. At this longitude,
close to the longitude of the section P17 shown in Fig. 4,
the westward South Equatorial Current is confined to the
upper 40 m where its annual average speed rarely exceeds
220 cm s
21
. During 1991 the South Equatorial Current
shoals to 20 m and actually disappears early in the year.
The analysis velocity at the same longitude as the mooring
shows a significant westward bias (Fig. 4). Thus, the South
Equatorial Current is too strong and the Equatorial Un-
320 V
OLUME
30JOURNAL OF PHYSICAL OCEANOGRAPHY
F
IG
. 7. Annual cycle of (a) kinetic energy and (b) sea level from the control analysis. The annual cycle
is defined as the annual harmonic of the Fourier series. Sea level has strong annual amplitude in the high
variability regions of the western boundary current. The maximum amplitude in these regions approaches
15 cm. Units are m
2
s
22
and cm.
dercurrent velocity is too weak at this location. The control
analysis surface South Equatorial Current is relatively
weak during 1991, while the Equatorial Undercurrent does
not weaken until 1992. From these results and additional
experiments we conclude that better meridional resolution
and stronger trade winds are required to improve the anal-
ysis of tropical currents.
4. Global statistics
The strongest signal in the mass and momentum fields
is the annual shift of heat and mass in response to shift-
ing winds and surface heat flux. We introduce our sta-
tistical analysis by discussing the annual cycle of two
key quantities, surface kinetic energy (u
2
1
y
2
)/2 and
sea level based on the 46-yr control analysis (Figs. 7a,b).
The annual cycle of a third variable, heat content, will
be discussed separately. The basin-scale structure of the
annual cycle of sea level is dominated by a pattern of
rising level in the summer hemisphere as a result of the
antisymmetry about the equator of winds and solar heat-
ing. Smaller-scale variations are also evident. Western
boundary current regions of the North Atlantic and
North Pacific have seasonal amplitudes approaching 15
cm (these results closely resemble those computed from
short 1-yr-long altimeter records by Cheney et al. 1994).
The Tropics also have distinct smaller-scale features.
The tropical Atlantic and Pacific show 4–6 cm varia-
tions in zonal bands resulting from seasonal changes in
the North Equatorial Countercurrent. The corresponding
amplitude of heat content in these regions (not shown)
is 2008Cm.
In contrast to sea level the annual cycle of kinetic
energy (Fig. 7b) is largest in the Tropics, reflecting the
seasonal changes in the tropical current system. The
annual amplitude of these currents exceeds 20 cm s
21
.
In the tropical Pacific the intensity of the current is
reduced somewhat near 1408W. The annual kinetic en-
F
EBRUARY
2000 321CARTON ET AL.
F
IG
.7.(Continued)
ergy in the Indian Ocean is elevated throughout, with
highest values along the western boundary in the region
of the seasonal Somali Current.
Seasonal variations in local heat storage, the time rate
of change of heat content, reflect the difference between
net surface heating and horizontal divergence of heat
transport. Because the ocean basins are bounded to the
east and west at most latitudes, the zonal average of
heat storage, shown in Figs. 8a–c for the three basins,
is the difference between net surface heating and the
meridional divergence of heat transport. Here storage is
computed by taking the center difference from succes-
sive monthly averages. These results can be directly
compared with the results of Hsiung et al. (1989). Heat
storage to a shallower 275-m depth, but using the ex-
panded WOA-94 dataset, is presented in Levitus and
Antonov (1997). Heat storage in the Pacific follows the
cycle of solar radiation with a maximum in the Northern
Hemisphere in June–August. The seasonal maximum is
somewhat lower than in either of the two previous stud-
ies. Maximum heat loss from the Southern Hemisphere
occurs perhaps a few weeks earlier and is of significantly
lower amplitude than Levitus and Antonov. The pattern
of heat storage near the equator resembles that of Lev-
itus and Antonov with a complex set of zonal bands of
heat gain and loss. The period July–October is one in
which the region from 108S–08is storing heat rapidly
as a result of the appearance of the cold tongue in the
eastern Pacific, while the region from 08–108N is losing
heat rapidly.
Heat storage in the Atlantic (Fig. 8b) generally re-
sembles heat storage in the Pacific. In the North Atlantic
maximum heat storage occurs somewhat earlier (May–
July). Heat is being exported mainly during early boreal
summer. In the tropical Atlantic rapid heat storage is
limited to the months June–September. South of the
equator storage occurs during September–February.
Heat storage in the tropical Indian Ocean (Fig. 8c) is
quite complicated. Between 208S and 58S the southern
Indian Ocean is gaining heat during June–October, a
period in which the sun is actually in the NorthernHemi-
sphere. This region is losing heat during November–
March, a period in which the sun has crossed over into
the Southern Hemisphere! These counterintuitive results
are generally consistent with those of Hsiung et al.
(1989) and Levitus and Antonov (1997).
322 V
OLUME
30JOURNAL OF PHYSICAL OCEANOGRAPHY
F
IG
. 8. Zonal average of monthly climatological 0/500 m heat
storage for the three ocean basins. Positive values indicate that the
ocean is gaining heat. Units are 8Cmmo
21
.
At periods longer than the annual cycle the distri-
bution of variance changes. In Fig. 9 we present a crude
decomposition of the spectral characteristics of the heat
content field by separating the variability into an inter-
annual band (1–5 yr) and a decadal band (5–25 yr). At
interannual periods the variability along the equator in
the Pacific is enhanced and shifted eastward, while off-
equatorial variability occurs in the western side of the
basin. The western tropical Indian Ocean has significant
variability, as do the eddy production regions of the
western boundary currents and Antarctic Circumpolar
Current. At decadal periods the Tropics become less
prominent relative to the subtropical and midlatitude
gyres. The shift of variability from the Tropics toward
midlatitude as the frequency decreases reflects the fun-
damental dynamical properties of the ocean.
Much of the heat content variability at interannual
and decadal periods in Fig. 9 is spatially incoherent. In
order to explore the sources of just that part of the heat
content signal that has basin scales we define three focal
areas, one in the central North Pacific (averaged, 308–
458N, 1808–2108W), a second in the eastern tropical
Pacific (averaged 58S–58N, 1508–908W), and a third in
the central North Atlantic (averaged 258–408N, 708–
308W). Note that the focal area in the North Atlantic is
smaller than the others, reflecting the smaller spatial
scales of variability there. The time series of area-av-
eraged heat content anomaly are presented in Fig. 10
(a linear trend has been removed, consistent with the
previous analysis). In order to define the spatial structure
of these modes we correlate the three heat content anom-
aly time series with heat content anomaly throughout
the global ocean. The three resulting correlation maps
are shown in Fig. 11.
The time series and spatial pattern of correlation with
North Pacific heat content is shown in the upper panels
of Figs. 10 and 11. The variability is substantially de-
cadal. The most prominent feature of the time series is
a cooling during the 1980s following a relatively warm
1970s. Heat content variability in the North Pacific has
been related to fluctuations in the Pacific–North Amer-
ica pattern of wind variation (Trenberth and Hurrell
1994; Graham 1994). Changes in the wind patterns in
1976–77 led to a cooling in the central basin that was
referred to as a climate shift by Miller et al. (1994).
This climate shift is evident in Fig. 10 but is followed
11 years later by a return to warm conditions, also noted
by Levitus and Antonov (1995).
By correlating the North Pacific time series with heat
content anomaly time series for the rest of the oceans
we can identify the spatial pattern of oceanic thermal
variability. The region that is most highly correlated
with the midlatitude North Pacific, interestingly, is the
western tropical and southern Pacific. Much of the cor-
relation results from a drop in heat content in the 1970s
that approximately coincides with the drop in the central
North Pacific. Between these two regions of positive
correlation is a region of negative correlation. This pat-
tern of alternating positive and negative correlation is
reminiscent of ideas of a slow advection of thermal
anomalies throughout the North Pacific (e.g., Latif and
Barnett 1994; Deser et al. 1996).
Next we turn our attention to the tropical Pacific. The
variability here is strongly interannual, reflecting the
importance of El Nin˜o. Individual peaks in the time
series can be identified with individual El Nin˜os. Ex-
amination of the time series shows that the amplitude
F
EBRUARY
2000 323CARTON ET AL.
F
IG
. 9. Root-mean-square 0/500 m heat content variability in the (a) interannual band (1–5 yr)
and (b) decadal band (5–25 yr). Units are 8Cm.
and frequency of heat content variability has been fairly
regular during the past five decades. Decadal variability
is also evident in the time series. The late 1970s were
a period of gradual warming coinciding with a warming
of SST (Wang 1995).
Examination of the spatial pattern of correlation with
tropical Pacific heat content shows the spatial structure
of the oceanic expression of El Nin˜o. For example, we
find that on either side of the zonal band of high cor-
relation near the equator are bands of negative corre-
lation. The band of negative correlation in the Northern
Hemisphere is most coherent, extending west-south-
westward from the west coast of North America. The
existence of off-equatorial bands of negative correlation
is a key feature of the delayed-oscillator theories of the
periodicity of El Nin˜o. The zone of negative correlation
extends into the eastern Indian Ocean, suggesting a con-
nection between the western Pacific and eastern Indian
Ocean. Examination of the lagged correlation shows that
the band of positive correlation near the equator shifts
eastward and poleward with time, while the off-equa-
torial bands of negative correlation shift westward.
The impact of El Nin˜o on the North and tropical
Atlantic has been examine in a number of studies (e.g.,
Carton and Huang 1994; Enfield and Meyer 1997). In
the tropical Atlantic Carton and Huang have proposed
that the changes in western tropical Atlantic winds as-
sociated with warm SST in the eastern tropical Pacific
leads to a buildup of heat in the western tropical At-
lantic. This anomalous heat, which occurs in the form
of anomalous deepening of the thermocline, eventually
leads to a deepening of the equatorial thermocline and
contributes to warming in the eastern Atlantic. Some
confirmation for these ideas is apparent in the weakly
positive correlation (0.2) between eastern tropical Pa-
cific heat content and western tropical Atlantic heat con-
tent.
The final time series in the lower panel of Fig. 10
represents the central North Atlantic Ocean. Heat con-
tent variability in the North Atlantic is also associated
with variability in the sector winds, in this case the North
Atlantic Oscillation pattern Atlantic winds (Hurrell and
van Loon 1997). The time series of heat content shows
rich decadal variability. The 1950s are relatively warm,
followed by a series of cool events in the early 1960s,
the late 1960s and early 1970s, and again in the late
1970s and early 1980s. The second of these cool events
has been documented by S. Levitus (Levitus 1990).
5. Conclusions
In a companion to this study Carton et al. (2000)
present a five-decade-long historical analysis of global
324 V
OLUME
30JOURNAL OF PHYSICAL OCEANOGRAPHY
F
IG
. 10. Time series of heat content anomaly from the seasonal
cycle for three regions, (a) North Pacific (averaged, 308–458N, 1808–
2108W), (b) Nino3 region of the eastern tropical Pacific (averaged
58S–58N, 1508–908W), and (c) North Atlantic (averaged 258–408N,
708–308W). The regions are indicated in the corresponding panels of
Fig. 11. Units are 8Cm.
upper-ocean temperature, salinity, sea level, and cur-
rents. The purpose of the analysis is to provide a uni-
formly gridded historical dataset for use in studies of
the ocean’s role in climate. Two ways to quantify the
accuracy of the analysis are by direct comparison to
independent observations and by examining identifiable
climate features such as the annual cycle, El Nin˜o, and
the Pacific–North American anomaly. Here wetake both
approaches.
We begin with a comparison to time series with lim-
ited spatial coverage and to global observation sets with
limited temporal coverage. The island tide-gauge time
series suggests that the analysis explains 25% of the
observed sea level variance at longer than annual fre-
quencies and more than 30% in the frequency band
between 5 and 25 years. The explained variance in-
creases in the tropical Pacific. Comparison to satellite
altimeter sea level shows a root-mean-square difference
of 4.0 cm in the Tropics 158S–158N and 5.2 cm globally
when this dataset is not assimilated. When altimetry is
assimilated, the rms difference in the tropical sea level
decreases to 3.1 cm. The comparison to a series of global
hydrographic sections shows average temperature and
salinity errors in the upper 500 m of 0.708C and 0.092
psu. The temperature errors are mainly concentrated at
thermocline depths. Salinity errors are distributed
through the mixed layer and pycnocline. A comparison
to moored and surface drifter velocity shows the anal-
ysis reproduces the qualitative behavior of surface ve-
locity in the Tropics.
We begin the discussion of climate features by con-
sideration of the annual cycle. At the annual period both
hemispheres show a strong response to seasonal forcing
of surface winds, heat, and freshwater. In the subtropics
and midlatitudes sea level varies in phase with seasonal
changes in solar radiation, except in the southern Indian
Ocean. Closer to the equator the phase of sea level un-
dergoes rapid reversals as the ocean responds to strong
annual variations in winds. The implied rates of heat
storage are consistent with previous studies.
The strongest basin-scale signal at interannual periods
is associated with El Nin˜o. To explore this pattern of
variability we examine the correlation of global heat
content with the heat content time series from a region
of the eastern equatorial Pacific (the Nino3 region) that
is itself considered an index of El Nin˜o. Our exami-
nation of the zero-lag correlation of global heat content
shows the eastern and western tropical Pacific to be out
of phase (correlation 20.4 to 20.6). The eastern Indian
Ocean is in phase with the western Pacific and thus is
out of phase with the eastern Pacific. The North Pacific
has a weak positive correlation with the eastern equa-
torial Pacific. Correlations between eastern Pacific heat
content and Atlantic heat content are modest.
At longer decadal periods the Pacific–North America
wind pattern leads to broad patterns of correlation in
heat content variability. Increases in heat content in the
central North Pacific are associated with decreases in
heat content in the subtropical Pacific and increases in
the western tropical Pacific. Atlantic heat content ispos-
itively correlated with the central North Pacific. The
Atlantic–Pacific relationship is confirmed by correlating
the heat content anomaly in the central North Atlantic
with global heat content.
In several respects the analysis is clearly inconsistent
with the observations. Major problems include:
1) The inability of the forecast model to produce ther-
mocline water masses in sufficient volume. Two ex-
amples of this problem are identified, the subtropical
mode water of the North Atlantic and the high sa-
linity subtropical water that enters the equatorial
thermocline from the south in the Pacific and At-
lantic. Interannual changes in these water masses
change the covariances of temperature and salinity
errors, and thus violates an assumption of our anal-
ysis. The major causes of insufficient thermocline
water-mass production are still not clear. Adjoint or
streamline assimilation techniques together with is-
F
EBRUARY
2000 325CARTON ET AL.
F
IG
. 11. Correlation of three heat content anomaly time series shown in Fig. 10 with global heat content
anomaly. (a) North Pacific, (b) tropical Pacific, and (c) North Atlantic.
opycnal coordinate forecast models may prove help-
ful.
2) The inability of the forecast model to produce re-
alistic mixed layer salinity. Our lack of knowledge
of historical surface salinity limits any estimate of
interannual fluctuations of salinity in the mixed layer
and contributes to the problems associated with in-
sufficient thermocline water-mass formation. Im-
provements in estimate of historical rainfall may help
address this problem.
3) The inability of the forecast model to simulate the
effects of important narrow topographic features.
Our lack of resolution limits the ability of the fore-
cast model to resolve features such as the Strait of
Gibraltar, the Windward Islands, or the Florida
Straits in the North Atlantic, all of which have pro-
326 V
OLUME
30JOURNAL OF PHYSICAL OCEANOGRAPHY
found effects on the general circulation. Lack of res-
olution limits mesoscale eddy production and alters
the dynamical balances in narrow currents such as
the Gulf Stream and the equatorial current systems.
On the other hand, increasing resolution will increase
production of mesoscale eddies. These eddies must
be treated as noise since they are not resolved by
the observation set prior to the availability of satellite
altimetry.
4) The errors in the Indian Ocean and the Southern
Hemisphere generally. The adequacy of the historical
subsurface data varies with location. Generally
speaking, the temperature of the North Atlantic and
major shipping routes of the North Pacific is rea-
sonably well sampled throughout most of the last
five decades. Sampling in the tropical Atlantic and
Pacific Oceans is intermittent, while the Indian
Ocean and the southern oceans are always inade-
quately sampled. Sampling of overlying atmospheric
variables is similarly limited. More salinity obser-
vations are also badly needed, as are measurements
below 500 m. Upgrades to WOA-94 are expected to
help fill in some of these data voids (S. Levitus 1997,
personal communication).
Despite these problems the authors feel greatly en-
couraged by the potential of the analysis for upper-ocean
climate studies on interannual to decadal timescales.
Acknowledgments. We are very grateful to a number
of people who have given us access to their datasets.
Mark Swenson and Zengxi Zhou of the Atlantic Ocean-
ographic Marine Laboratory have provided the monthly
averaged surface drifters, and Sydney Levitus and Rob-
ert Cheney and their colleagues at the National Ocean-
ographic Data Center/NOAA have provided access to
the hydrographic and altimeter data. We have benefited
from the datasets collected by principal investigators of
the Tropical Ocean Global Atmosphere/World Ocean-
ographic Circulation Experiments. Finally, we want to
express our gratitude for support from the Office of
Global Programs/NOAA under Grant NA66GP0269 and
the National Science Foundation under Grant
OCE9416894.
REFERENCES
Bennett, A. F., 1990: Inverse Methods in Physical Oceanography.
Cambridge University Press, 346 pp.
Carton, J. A., and B. Huang, 1994: Warm events in the tropical
Atlantic. J. Phys. Oceanogr., 24, 888–903.
, G. Chepurin, X. Cao, and B. S. Giese, 2000: A simple ocean
data assimilation analysis of the global upper ocean 1950–95.
Part I: Method. J. Phys. Oceanogr., 30, 294–309.
Cheney, R. E., L. Miller, and J. Lillibridge, 1994: Topex/Poseidon:
The 2 cm solution. J. Geophys. Res., 99, 24, 555–563.
Chepurin, G. A., and J. A. Carton, 1999: Comparison of retrospective
analyses of the global ocean heat content. Dyn. Atmos. Oceans,
submitted.
da Silva, A. M., C. C. Young, and S. Levitus, 1994: Atlas of Surface
Marine Data 1994, Vol. 1: Algorithms and Procedures. NOAA
Atlas NESDIS 6, U.S. Department of Commerce, NOAA, NES-
DIS, 83 pp.
Daley, R., 1991: Atmospheric Data Analysis. Cambridge University
Press, 457 pp.
Deser, C., M. A. Alexander, and M. S. Timlin, 1996: Upper-ocean
thermal variations in the North Pacific during 1970–91. J. Cli-
mate, 9, 1840–1855.
Enfield, D. B., and D. A. Meyer, 1997: Tropical Atlantic sea surface
temperature variability and its relation to El Nin˜o–Southern Os-
cillation. J. Geophys. Res., 102, 929–945.
Frankignoul, C., F. Bonjean, and G. Reverdin, 1996: Interannual var-
iability of surface currents in the tropical Pacific during 1987–
1993. J. Geophys. Res., 101, 3629–3647.
Graham, N. E., 1994: Decadal-scale climate variability in thetropical
and North Pacific during the 1970s and 1980s: Observations and
model results. Climate Dyn., 10, 135–162.
Hsiung, J., R. E. Newell, and T. Houghtby, 1989: The annual cycle
of oceanic heat storage and oceanic meridional heat transport.
Quart. J. Roy. Meteor. Soc., 115, 1–28.
Hurrell, J. W., and H. van Loon, 1997: Decadal variations in climate
associated with the North Atlantic Oscillation. Climatic Change,
36, 301–326.
Ji, M., A. Leetmaa, and J. Derber, 1995: An ocean analysis system
for seasonal to interannual climate studies. Mon. Wea. Rev., 123,
460–481.
Latif, M., and T. Barnett, 1994: Causes of decadal climate variability
over the North Pacific and North America. Science, 266, 634–
637.
Levitus, S., 1990: Interpentadal variability of steric sea level and
geopotential thickness of the North Atlantic Ocean, 1970–1974
versus 1955–1959. J. Geophys. Res., 95, 5233–5238.
, and J. Antonov, 1995: Observational evidence of interannual
to decadal-scale variability of the subsurface temperature–salin-
ity structure of the world ocean. Climatic Change, 31, 495–514.
, and , 1997: Climatological and Interannual Variability of
Temperature, Heat Storage, and Rate of Heat Storage in the
Upper Ocean. NESDIS Atlas Series 16, NOAA, 6 pp.
, T. P. Boyer, and J. Antonov, 1994: World Ocean Atlas, 1994.
Vol. 5: Interannual Variability of Upper Ocean Thermal Struc-
ture, NESDIS Atlas Series, NOAA, 176 pp.
Malanotte-Rizzoli, P., 1996: Modern Approaches to Data Assimilation
in Ocean Modeling. Elsevier, 455 pp.
Miller, A. J., D. R. Cayan, and T. P. Barnett, 1994: The 1976–77
climate shift of the Pacific Ocean. Oceanography, 7, 21–26.
Miller, R. N., and M. Cane, 1996: Tropical data assimilation: The-
oretical aspects. Modern Approaches to Data Assimilation in
Ocean Modeling, P. Malanotte-Rizzoli, Ed., Elsevier, 207–234.
Mitchum, G., 1994: Comparison of TOPEX sea surface heights and
tide gauge sea levels. J. Geophys. Res., 99, 24 541–24 553.
Reynolds, R. W., and T. M. Smith, 1994: Improved globalsea surface
temperature analysis using optimum interpolation. J. Climate, 7,
929–948.
Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere–ocean
variations in the Pacific. Climate Dyn., 9, 303–319.
Wang, B., 1995: Interdecadal changes in El Nin˜o onset in the last
four decades. J. Climate, 8, 267–285.
Wunsch, C., 1996: The Ocean Circulation Inverse Problem. Cam-
bridge University Press, 442 pp.
