JOURNAL
OF GEOPHYSICAL
RESEARCH,
VOL. 103, NO. C13, PAGES 30,973-30,984, DECEMBER
15, 1998
Flow of Bottom and Deep Water in the Amirante
Passage and Mascarene Basin
GregoryC. Johnson,
1 DavidL. Musgrave,
2 BruceA. Warren,3 Amy Ffield,
and Donald B. Olson s
Abstract. In the Indian Ocean the Amirante Passageis the sill through which relatively cold, fresh, oxygen-rich,and nutrient-poor bottom water spreadsnorthward
into the Somali Basin from the Mascarene Basin. The passageis also a conduit
through which relatively warm, salty, oxygen-poor,and nutrient-rich deep water
spreads south. Previous estimates for northward transport of bottom water in the
passagehave been made from station pairs and sectionswithout benefit of tracer
measurements. Previous estimates of southward transport of deep water are scarce.
Three hydrographic sections were made across the passagein 1995 and 1996 as
part of the World OceanCirculationExperiment(WOCE). Two WOCE sections
were also made perpendicular to the western boundary in the Mascarene Basin,
just south of the passage. The geostrophicshear field is used with the salinity,
dissolvedoxygen, and silica distributions to select a range of zero-velocity surfaces
(ZVSs) on potential isothermsfrom 1.0ø to 1.1øC(hencea rangeof geostrophic
transports)for whichthe flow directionis consistentwith the tracer distributions.
Objective mapping is usedto obtain flux estimatesbelow the deepestcommonlevel
of station pairs. Estimates in the Mascarene Basin result in a bottom water volume
transportfrom2.5 to 3.8 x 10• m3 s-1 northward
towardthe passage
belowthe
ZVSs and a deep-water transport between the ZVSs and 2.5øC from 11.6 to 6.4
x 10• ma s-• southward.Estimateswithinthe passage
resultin transportsfrom
1.0 to 1.7 x 10• m a s-• northward for the bottom water and from 8.6 to 3.8 x
10• m3 s-• southwardfor the deepwater.
1.
Introduction
Maps of water properties on, and depths of, potentialdensity surfaces illustrate some aspects of these flow
patterns, especially the large-scale control by bottom
A powerful dynamical framework for the large-scale
abyssalcirculationconsistsof interior upwellingfed by
topography[Mantylaand Reid, 1995].
high-latitude sinking through deep western-boundary
Estimates of net transport below about 2000 dbar
currents[Stommeland Arons, 1960]. In the South into the Indian Ocean near 32øS range from a minimum
Indian Ocean, zonal hydrographic sections have revealed three separate northward flows in deep westernboundary currents supportedby three roughly meridional boundaries defined by continental rises and ocean
ridgesthat separatethe Indian Oceaninto three setsof
of zero as derived from hydrographic stations so sparse
as to missthe boundarycurrents[Fu, 1986],as well
as from a combination of heavily smoothed climatol-
ogywith a general-circulation
model[LeeandMarotzke,
of27 x 106m3 s-1 asderived
from
deepbasins[Warren,1981; Tooleand Warren,1993]. 1997],to a maximum
a closely spaced hydrographic section with the zero-
•Pacific Marine Environmental Laboratory, NOAA,
Seattle, Washington.
2Universityof AlaskaFairbanks/SFOS.
3Department of Physical Oceanography,Woods Hole
velocitysurface(ZVS) for geostrophic-velocity
computationinferredfromwater-property
distributions
[Toole
and Warren,1993].The mostrecentobservation-based
estimate for the total northward flow of abyssalwaters
into the Indian Oceanacross32øSlatitude [Robbins
and
OceanographicInstitution, Vv•oodsHole, Massachusetts.
1997]is 11.9+ 2.7 x 10• kgs-• (converting
mass
4Lamont-Doherty Earth Observatory, Columbia Univer- Toole,
transport
to
volume
transport
involves
dividing
by
the
sity, Palisades, New York.
in
situ
density,
so
10
•
kg
s
-•
•
10
•
m
a
s-l),
obtained
5Rosensteil School of Marine and Atmospheric Science,
University of Miami, Miami, Florida.
by using the silica budget to adjust the referencevelocCopyright1998 by theAmericanGeophysicalUnion.
ities of Tooleand Warren[1993]. Furtherquantifying
Papernumber1998JC900027.
the northward mass and property fluxes of the abyssal
flow, and the return fluxes southward, will help to elu-
0148-0227/98/1998JC900027 $09.00
cidate the role that the thermohaline
30,973
circulation
in the
30,974
JOHNSON ET AL.- AMIRANTE PASSAGE DEEP CIRCULATION
Indian Ocean plays in global heat and freshwaterbud-
Two of the twelve studies and five of the sixteen hy-
drographictransport estimatesof abyssalflow in the
The westernseriesof deep basinsin the Indian Ocean, westernbasins(Table 1) are acrossthe AmirantePasfrom south to north, consistsof the Crozet, Madagas- sage. The 3650-misobath[Smithand Sandwell,1997]
the SomaliBasin (Figure 1) exceptfor the
car, Mascarene,Somali,andArabianBasins(Figure1). encloses
to the southand the OwenFre[cture
There are at least twelve published studies using six- AmirantePassage
teen hydrographicsectionsspanningpart or all of these Zone (10øN, 57øE) to the north. Thus the Amirante
westernbasinsto make geostrophictransport estimates Passageis the route by which all bottom water from
for northwardabyssalflow (Table 1). Theseestimates the south enters the Somali Basin. All previous botstart at 7-11 x 106m3 s-1 at 32øSand generallyde- tom water transports in the passageare calculatedrelcrease northward to 1-5 x 106 m 3 s-1 near and north
ative to and below a 3800-dbar ZVS, a choicewhich,
of the AmirantePassage(9øS,53øE),the deepestcon- for a deep ZVS abovethe passage,generallymaximizes
nection between the Mascarene and Somali Basins, near northward transport of bottom water. Two of these
9øS,53øE (Figure 1). ZVSs for theseestimatesrange estimates are actually made using two sets of station
between 1000 and 4200 dbar, tending to deepen from pairs within the passagewhich eachyield rangesof relsouth to north and from west to east. For many of atively small transportestimates[Fieux and Swallow,
thesetransport estimates,visual inspectionof potential 1988]. For theseestimates,geostrophic
velocityis extemperature or density sectionssuggeststhat the ZVS trapolated as a constant below the deepest common
choices are made to maximize northward baroclinic velevel(4200dbar) for the stationpairs,but differentasgets.
locity in the bottom water, assuminga middepth zerovelocity surface.
20øN
sumptions about where this flow is located above the
highly varying topographyleads to different valuesfor
the resulting transport increments, and hence for the
ranges in the total estimates. The extrapolations account for 0.3-1.1 x 106 m 3 s-1 and 0.1-1.3 x 106 m 3 s-1
of thesetransport estimates. The other three transports
are based on sections of four to six stations each within
15øN
the passage[Barton and Hill, 1989]. Thesehigherresolution sectionsmore than double the transport estimates obtained from the station pairs. Extrapolation
of velocitiesbelow the deepestcommon levelsin these
10øN
sections
accounts
for 0.9, 3.6, and 1.5 x 106m3 s-• of
5øN
the transport estimates. Thus, while the distribution
of geostrophicvelocitiesacrossthe passageis better defined by the sectionsthan by the station pairs, in both
casesthe extrapolated velocity accountsfor roughlyhalf
the transport estimate.
This investigationexaminesthree recent,denselysampled hydrographic sectionsthat span the entire Amirante Passage,together with the three sectionsprevi-
0ø
5øS
10øS
ouslyanalyzedby BartonandHill [1989],andincludes
two more just south of the passagein the Mascarene
Basin. This regionalanalysisof deep and bottom water
transports improveson previous studies in four ways.
20øS
First, estimates of transport below the deepest common level of station pairs are achievedthrough objective mapping rather than assuming a uniform veloc25øS
ity, making consistentuse of all the hydrographicinformation gathered. Second,water property distributions are combinedwith the geostrophicshear to select
30øS
a plausible range of ZVSs, again incorporatingmore information than was •sed in previous studies. Third,
the boundary between bottom and deep water is de35ø•0o
E 45øE50øE55øE60øE65øE70øE75øEfined as the ZVS, and the upper limit of deep water
is set by examination of the water property distribuFigure 1. Map (Mercatorprojection)of the western
tions. This procedure achievesimproved estimatesof
Indian Ocean with basin names and the 3650-m isobath
(thin line) [Smithand Sandwell,1997]that definesthe northward transport of bottom water, as well as estiSomali Basin. The location around the Amirante Pasmates of south•vardtransport of deep water in the region. Fourth, the relatively large number of sectionsansage(insidethick line) is shownin detail in Figure2.
15øS
o.
JOHNSON ET AL.: AMIRANTE
PASSAGE DEEP CIRCULATION
30,975
Table 1. Summaryof PublishedNorthwardDeep TransportEstimatesin the WesternBasinsof the Indian
Ocean, ProceedingFrom South to North
Lat.,
ZVS,
Transport,
Location
øS
dbar
106 ms S--1
Comments
Reference
Crozet Basin
48
2500
3.2
Madagascarand
32
1000-3600
11.0
>ZVS
>2000
Warren [1978]
Tooleand Warren [1993]
32
1000-3600 7.4*
Crozet
Basins
Madagascarand
Crozet
dbar
>-•2000 dbar,
32-23
3500 m
6.1
>ZVS, along ridge
MadagascarBasin
MadagascarBasin
MadagascarBasin
35-20
23
23
3500 m
3100
3100
5.2
5.7
3.6
>ZVS
Mascarene
Basin
18
1400-4200
5.2
>ZVS, repeat section
2000 dbar,
DWBC only
Mascarene
Basin
12
>ZVS
Southwest Indian
Robbinsand Toole[1997]
silica constraint
Basins
Warren [1978]
Ridge
ZVS, nonzonal
3600
4.4
Amirante Passage 9
3800
0.8-2.1 and
Amirante Passage
9
3800
3.7, 5.1, and 2.8
Somali Basin
4-2
1.2øC 0
3.2 and 4.6
Arabian Basin
12øN
3000
3.2, 2.3, and 1.2
Arabian Basin
7ø-10øN
1.7øC 0
0.5 and 4.8
0.9-1.7
ZVS, two sets
of station pairs
ZVS, three sections
ZVS, two sections,
DWBC only
ZVS, three sections
ZVS, two sections,
DWBC only
Swallowand Pollard [1988]
Warren [1974]
Fieux et al. [1986]
Warren [1981]
Warren [1974]
Fieux and Swallow[1988]
Barton and Hill [1989]
Johnsonet al. [1991b]
Quadfaselet al. [1997]
Johnsonet al. [199la]
ZVS, zero-velocity
surface;DWBC, deep'western-boundary
current. The third studyis a reanalysisof the second,
otherwise,differenthydrographic
data setsare usedfor eachestimate.Most of thesestudiesassumea constantvelocity
below the deepestcommonlevel of each station pair.
'109 kg s-•.
alyzed, with closelyspacedstations,allowsfor improved three sections(Figure 2; CD22, CD24, and CD24S)
statistical uncertainties in the transport estimates.
takenacrossthe passage
in April andJuneof 1987[Barton and Hill, 1989]are reanalyzedhere. Potentialtem-
perature •, in situ density,and specificvolume anomaly
are calculated from the 2-dbar CTD data, filtered with
As part of the U.S. componentof the World Ocean a 50-dbar half-width Hanning filter, and subsampledat
CirculationExperiment (WOCE), three hydrographic 50-dbar intervals. The square of buoyancy frequency,
sectionswere occupied acrossthe entire Amirante Pas- N 2 -- g/p Op/Oz,is alsocalculated
from the 2-dbar
sage and two sectionsjust south of the passagein the CTD data. Here g is the local gravitational acceleraMascareneBasin (Figure 2). Data examinedare con- tion and the densityp is locally referencedto the central
tinuousconductivity-temperature-depth
(CTD) profiles pressureover which the derivative is evaluated; hence
of temperature T and salinity $ versuspressureP, to- the effect of compressibilityis excludedbut that of the
gether with discretebottle measurements
of dissolved adiabatictemperaturegradientis included.The N 2 esoxygen,02, and silica, H4SiO4, at selectedpressures. timates are filtered with a 250-dbar half-width Hanning
Stations used here for volume transport calculationsin- filter before subsamplingbecause, as derivatives, they
clude 70-91 from cruise IR7N, occupied in April 1995, are inherently noisy.
Geostrophicflux calculationsuse horizontal specific
stations 720-738 from cruiseI7N, occupiedin July 1995,
and stations 1197-1215 from I2, occupied in January volume anomaly gradientsbetweenstation pairs. In re1996. The mean station spacingis 37 km, the minimum gionswherethe bottom slopes,this meansthat somead
is 11 km, and the maximumis 75 km. Stationswereoc- hoc method (usuallyapplyingzerovelocity,a constant
cupiedto within 10 m of the bottom. These CTD pro- velocity from above,or a constantvertical shear from
files and discrete bottle data are preliminary but are above)is often usedto extendthesecalculationsbelow
already thought to comply with WOCE one-timesur- the deepest common level of station pairs, into what
vey standardsfor accuracyand precision(T. Joyceand are customarilycalledthe bottom triangles.Instead of
C. Corry, Requirementsfor WOCE hydrographicpro- this ad hoc procedure, here water properties in these
grammedata reporting,unpublished
manuscript,1994) regionsare extrapolated systematicallyusing objective
1983]. The powerof
and are not likely to changemuch when final calibra- mappingtechniques[Roeromich,
tions are completed. In addition, the CTD data from the method is in the combined use of vertical and hor-
2. Data and Mapping
30,976
JOHNSON ET AL.: AMIRANTE
PASSAGE DEEP CIRCULATION
7øS
3. Water Property
Distributions
Geostrophictransport estimates often rely on water
property distributions to help determine referencevelocities, especiallyfor deep currentswhere velocitiesare
small. In the basinsin questionthe bottom water, flowing from southto north, is colder,fresher,more oxygenrich, and more silica-poor than the southward flowing
8øS
9øS
deepwater aboveit [e.g., Wyrtki,1971; Warren,1981;
Spenceret al., 1982; Mantyla and Reid, 1995]. The
operating assumption of this interpretation of the circulation is that the water-property patterns are primarily governedby advection, so that on a given isopycnal
the direction of flow tends to parallel property isopleths
10øS
rather
than to cross them.
The reference velocities
for
11øS
the geostrophicvelocity profiles are constrainedby the
velocity directions consistentwith the isopycnaltracer
gradients. To the extent that variable mixing along
or across isopycnalsinfluences the property distribu12øS
tions, the choicesof referencevelocitiesmay be questionable, and, while one expects that the net flow of
southern-sourcewater should be northward, one cannot
13ø•
OE 52OE 53OE 54OE 55OE 56OEreally argue that the flow everywhereshouldbe northward. Other sourcesof error in inferring mean transFigure 2. Detailedmap (Mercatorprojection)in the ports from synopticsurveysof the density field include
Amirante Passage(near 9øS,53øE) regionwith areas time-dependentgeostrophicmotions suchas eddiesand
between3000-m,4000-m,and 5000-misobaths(black) planetary waves and ageostrophicmotions such as in-
increasingly
shaded[SmithandSandwell,
1997]andstation locations(seelegend). SeeFigure I for map lo-
ternal
waves and tides.
Isopycnal property gradients are not discernable to
cation within western Indian Ocean. Sections across
the Amirante Passageare roughly oriented southwest within measurement error and mesoscale noise below
to northeast, whereas those in the MascareneBasin are 1.6øC for the WOCE sections from within the Amioriented northwest to southeast.
rante Passage to at least 12.25øS, 55øE in the Mascarene Basin, the southeast extent of transport estimates made here. However, larger-scale distributions
of water properties suggestgeneral northward abyssal
flow in this western series of basins that extends upizontal gradients to extrapolate into the unsampledar- wardto depthsbetween3400and 4400m [e.g.,Wyrtki,
eas, resulting in flux calculationsthat make consistent 1971; Warren, 1981; Spencer et al., 1982; Mantyla and
use of all available density information. For the map- Reid,1995].Tracerdata ($, 02, andHaSiO4)areplot-
ping we assumean error-to-signalenergyof 1/16 and ted as curvesversusOin the regionof study (Figure2)
covariance with 100-kin lateral and 500-m
from the WOCE I7N section(Figure 3) and are also
a Gaussian
vertical correlation length scales. These length scales
are chosenbecausethe shear signature of the boundary
current appears to be about 100 km wide in the vicinity of the Amirante Passage and potential isopycnals
fall about
500 m over that
distance.
Median
station
contoured
in vertical
sections with
O as a vertical
axis
overa wider latitude range(Figure4). Thesedata are
usedto support this assertionand make other inferences
about deepflow directions. In addition, at least at 18øS,
water characteristicsof southern origin indicate northward flow as well in a thin band against Madagascar
spacingis 33 km in the passageand the median magnitude of the bottom pressuredifferencebetween stations that reachesup to at least2000m [Warren,1981].
is 300 dbar; hence the formal errors in the bottom triThe O-Scurves(Figure3) for sectionI7N in the region
angles are small because the bottom triangle regions of study (Figure 2) consistof a tight linear mixingline
are usually within a correlation length scale of actual from the cold, fresh bottom end-member up to about
station data in both the lateral and vertical directions.
1.6øC. Above this mixing line is a distinct $ maxiThese small formal errors mean that data outside the
mum near 1.9øC. This $ maximum is separated from
bottom triangles in both directions strongly influence the salty northern intermediate water above by a set
the mapping within them, with the ratio of their in- of $ minima that vary from 2.3ø to 4.5øC. Extremely.
fluencegovernedby the Gaussiancovarianceand their fresh minima are found between 3 ø and 4øC at three
distance from the mapped region scaledby the correla- stations. The cold, fresh end-member near the bottom
is Circumpolar Deep Water spreadingnorthward from
tion length scales.
JOHNSON ET AL.: AMIRANTE
PASSAGE DEEP CIRCULATION
5
the south (Figure 4). The S maximumnear 1.9øCis
a
4.5
30,977
a signature of the Indian Deep Water spreadingsouth-
..................
ward from the north (Figure 4). The S minimumnear
3øC is again suggestiveof water with a southern origin
locatedabovethe Indian Deep Water (Figure4). Here
the term "spreading" is used to signify the combined
effectsof advection and diffusionon the water property
3
2.5
fields.
.....................
Dissolved 02 values are highest near the bottom and,
over the large scale, decreasetoward the north, indicating northwardspreadingof oxygen-richCircumpolar
1.5
DeepWater from the south[Tooleand Warren,1993].
The 0-02 curves have a maximum curvature at about
1.6øC (Figure 3), and the low valuesthere are sugges-
0.5
34.68
34.72
34.7
34.74
34.78
34.76
Salinity (PSS-78)
fromthe north (Figure4). Anothercurvaturemaximum
of oppositesignnear 3øC (Figure 3) againsuggests
either oxygen-richwater of a southernorigin above the
Indian Deep Water (Figure 4) or a regionof little net
flow. Nitrate and phosphatecurves(not shown)are
very strongly anticorrelatedwith the tendencyof 02,
so they do not add much additional information.
5
The 0-H4SiO4 curves have a deep maximum near
3
2.5
tive of oxygen-poorIndian Deep Water spreadingsouth
1.4øC (Figure 3), indicatingsouthwardspreadingof
silica-richIndian Deep Water of northernorigin (Figure 4). Valuesdecreasebelow the maximum in a relatively tight, linear fashion(Figure3), a resultof north-
...................
ward spreading of relatively silica-poor Circumpolar
Deep Water somewherebelow the H4SiO4 maximum
0.5
1O0
120
140 1601
02(gmol
kg-)
180
200
(Figure 4). Again, near 3øC, there is a curvetoward
lowervalues(Figure 3), suggestive
of either the influence of silica-poorwaters of southernorigin above the
Indian DeepWater H4SiO4maximum(Figure4) or lit-
5
tle net flow in this temperature range.
The planetary componentof potential vorticity Q is
4.5
another conservative tracer in the absenceof mixing,
friction,or significant
relativevorticity[Pedlosky,
1987].
HereQ = f/gN 2, wheref isthelocalCoriolis
parame-
3.5
ter. Overall, there is an interior minimum in the magnitude of Q between 1.4ø and 1.6øC, and magnitudesare
againvery low nearthe oceanfloor (Figure5). Above
1.6øC,magnitudesincreasetoward the main pycnocline.
There is an interior maximum
near 1øC.
Maps of Q on three deepisopycnalsurfaceshavebeen
publishedfor the Indian Ocean[O'Dwyerand Williams,
1997,Figures8b, 11b, and 15b]. Theseisopycnalsur-
rich, and silica-poorvaluesof southernorigin near 3øC
suggestsan upper bound for the Indian Deep Water
facesare near 0 = 1.85ø, 1.35ø, and 1.15øC in the region
of study. While they are not ideal for tracing the extrema describedabove, the lightest surface,with lowest
magnitudesof Q from the equatorto 30øSfoundin"the
western basin system, suggeststhat the interior minimum at 1.4ø to 1.6øC (Figure 5) is a signatureof a
plumeof southwardflowingIndian DeepWater. Indian
Deep Water originatingin the north with a small positive Q and movingsouthwardwould haveto modify Q
to a small negative value near the equator by mixing
processes,probably at the western boundary, as seen
near 2.5øC.
from 1.4ø to 1.6øC. Near the bottom in the Circumpo-
0.5
90
100
110
1120
H4SiO
4(gmol
kg-)
130
Figure $. Propertycurvesof (a) S, (b) 02, and (c)
H4SiO4versus0. Stations720-738 from WOCE section
I7N (opencircleson Figure2) are used.The Circumpolar Deep Water near the bottom is fresh, oxygenrich, and relatively silica-poorcomparedwith the Indian Deep Water with a salty,oxygen-poor,
silica-rich
corefrom 1.4ø-1.9øC. A tendencytoward fresh,oxygen-
30,978
JOHNSON
ET AL.' AMIRANTE
PASSAGE DEEP CIRCULATION
Potential
Temperature
(øC)
Salinity(PSS-78)
5
lOOO
1500
2000
2500
3.5
3000
3
3500
4000
2
45OO
1.5
34.73-
5000
5500
15øS
10øS
5øS
I'"'"'"'l"'"'"l'""'l'"'"'
r4D
0
0
0
0ON
5ON
10øN
15øS
'i"""'"l'"'"'"l'"'"'"l'""""l'"l
0
0
0
0
10øS
5øS
I'"'"'"'l'"'""?"""r'"'"
0
.1•
•
0
0
0
0øN
5øN
10øN
"1""'""1'"'"'"1""""'1""""'1'"1
0
0
0
0
0
.1•
Silicate
(•moles
kg-l)
Oxygen
(•moles
kg-l)
5
4.5
120
4
130
,-3.5
140
3
150
•2.5
2'
60
155
1
60øE $OøE
1.5
Figure 4. Distribution of deep water propertiesalong I7N from station 709 in the Mascarene
Basin to station 794 in the Arabian Sea: (a) verticalsectionof objectivelymapped0 contoured
with depthasthe verticalaxis,(b) verticalsectionof $ with 8 as the verticalaxis,(c) vertical
sectionof 02 with 0 as the verticalaxis,and (d) verticalsectionof H4SiO4with 0 as the vertical
axis. Station locations are given at the bottom of each plot. Bottle $, 02 and H4SiO4 are
linearly interpolatedto 0 at eachstation and then smoothedbeforecontouring.The inset map
in Figure 4c showsthe station locations. The Amirante Passageis betweenstations 731 and 740.
The topographicfeaturesnear 5øS and 5øN are the SeychellesBank and the CarlsbergRidge,
respectively.
lar Deep Water, the magnitude of Q is again reduced
by vertical mixing and the no-flux bottom boundary
condition, but above that a finite negative value is carried north in the Circumpolar Deep Water, corresponding to the magnitude maximum near 1.0øC. The two
densersurfacesof O'Dywerand Williams[1997]showa
Water flowingsouth[Reidand Lonsdale,
1974],rather
than a core layer of high-magnitudeQ water as suggested above. This hypothesis is also consistentwith
higher Q magnitudesnear the westernboundary,but
the dynamical foundation for this interpretation has
beencalledinto questionusingmodelsto providecoun-
similar maximum in the magnitude of Q alongthe west- terexamples[Warren,
1977].
ern boundary, just east of Madagascar, consistentwith
a Circumpolar Deep Water influencespreadingnorth- 4. Volume
Fluxes
ward in the Western
Basin.
An alternate explanation of similar maxima in the
Verticalsectionsof potentialtemperature(Figure5)
magnitudeof Q in the Pacific Oceanis that they reveal give a qualitative indication of where the deep geoboundariesbetween homogenouswater massesof Cir- strophic shear is located. Geostrophic shear is located
cumpolar Deep Water flowing north and Pacific Deep where potential isotherms tilt strongly and stratifica-
JOHNSON ET AL-
AMIRANTE
PASSAGE DEEP CIRCULATION
CD22 CD24 CD24S
ß' .• v '
IR7N [Mascarene Basin]
•.-....'•'•'"•-•'.'...'•
$0,979
I7N [Mascarene Basin]
.'-x .'• ."•'•Y'/•'-•
....
.......
,.?.>.•
...:.:....:.-
...
:',•-'¾..;•..
..
'•½•...•::"
j.•":•::•::'
:•.......
•-::::•:
:•:•'• }•:
:
:•
..................
......................
:.•:.•:.;
3000
3500
4000
4500
I
-5
5000
-4
-3
-2
-1
I
100 km
Q [10-12m-1s-1]
Figure 15. Deepverticalprofilesof potentialvorticity(Q, shadedat 1.0 x 10-12 m-1 s-1
intervalsandcontoured
with thin whitelinesat 0.5 x 10-12 m-1 s-1 intervals)with potential
temperature(0, contouredwith thick dark linesat 0.1øCintervals)for the hydrographic
sections
acrossthe AmirantePassage(six left profiles)and MascareneBasin(two right profiles).Station
locations(seeFigure 2) are shownby tick marks at the tops and bottomsof the sections.The
westernedgeof eachsectionis toward the left side. The vertical exaggerationis 500:1.
tion is high (hencethe magnitudeof Q is large),regions tial isotherm,2.5øC (near 1950 m), a reasonableupper
centered around 0.9ø to 1.1øC. As detailed above, this
high Q magnitude is either associatedwith northward
flow of Circumpolar Deep Water or reveals the boundary betweenCircumpolar Deep Water belowand Indian
Deep Water above. In either case,the shear is suchthat
overall northward velocity increasestoward the bottom
around these potential isotherms.
boundary for the extent of Indian Deep Water in the
region.
In the absenceof other information, there is no partic-
ular reasonto pick a ZVS at a singleP, 0, potential density, within minimumor maximumshearlayers,or even
assumethat a ZVS exists everywhere. Here we choose
to use 0 surfacesfor a ZVS and compute geostrophic
The water property distributions(Figures3 and 4) volume transports for a set of ZVSs between 0.7ø and
suggestthat the cold, fresh, oxygen-rich,nutrient-poor 2.0øC at 0.1øC intervals. Where the lowesttemperature
CircumpolarDeep Water shouldbe flowingnorth some- is higher than the ZVS, the bottom is usedas the ZVS.
where below 1.4øC, and at the very least within the Use of a set of isopycnalor even level surfaceswould cerrelatively homogenousbottom layer. These same dis- tainly givesimilar results. This exerciseis performedfor
tributions suggest that the warmer, saltier, oxygen- all three WOCE sectionsacrossthe Amirante Passage
poor, nutrient-rich Indian Deep Water should be flow- and the three partial historical sectionsthere, as well
ing south between 1.4ø and 1.9øC. The property distri- as for the two WOCE sections in the Mascarene Basin.
butions, especiallythe salinity minimum strengthening The volume transports for each sectionin each ZVS are
to the south near 3øC, make an intermediate poten- summed in 0.1øC bins. The values in bins for each sec-
30,980
JOHNSON ET AL.' AMIRANTE PASSAGE DEEP CIRCULATION
I
I
I
I
I
I ..........
! ......... I
I
I
I
I
I
3.1
3.1
b3
a 3
2.9
2.9
2.8
2.8
2.7
--0•••_ 2.7
o
2.6
2.6
2.5
2.5
2.4
2.4
2.3
2.3
N
2.2
2.2
•2.1
L•2.1
m
2
2
m 1.9
1.9
¸
•
1.8
0 1.7
1.6
1.6
1.5
1.5
1.4
1.4
1.3
1.3
1.2
1.2
1.1
1.1
1
0.9
0.9
0.8
0.8
0.7
0.7
0.7 0.8 0.9
1
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
ZeroVelocitySurfaceO
(oC)
0.7 0.8 0.9
1
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1o8 1.9
2
ZeroVelocitySurfaceO
(oC)
Figure6. Contour
plotsofmeangeostrophic
volume
transport
(in106mas-•) per0.1øCbinasa
functionof zerovelocitysurfacespecifiedby potentialtemperature.Regionsof positivetransport
are shaded.The requirements
that bottomwater be flowingnorth, deepwater between1.4ø and
1.9øCbe flowingsouth,and the zerovelocitysurfacebe the samein the (a) AmirantePassage
and(b) Mascarene
Basinconstrain
the zerovelocitysurface
to lie between
0.9ø and1.1øC.
tion are then averagedto find mean volume transports between 1.4 ø and 1.9øC is northward in one of the loca-
(Figure6) andstandarddeviations
(not shown)inthe
tions. Estimates for ZVSs from 0.7 ø to 1.3øC result in
passageand the basin, with allowancesmade for the a rangeof volumetransports(Table 2). A comparison
partial nature of the coverageof the Amirante Passage of transportsbetweenthe sectionscan be summarized
by the net transportdifferencebetweensectionsand
by the older sections.
An examinationof the mean transports(Figure 6) the root meansquareof differencesbetweenindividual
suggeststhat to obtain flowsin the Amirante Passage 0.1øCtransportbins(Table2, last column).Owingto
and the Mascarene Basin that are consistent with the
the closeproximityof the westernendsof the Mascarene
water propertiesin an overallsense(significantnorth- Basin and AmirantePassagesections(Figure2), and
ward flow at the bottom, southward flow from 1.4ø to the extensive shallow barrier of the Mascarene Plateau
1.9øC), the ZVS must residebetween0.9ø and 1.1øC, to the east (Figure 1), it seemsplausibleto eliminate
that is, within the region of deep geostrophicshear. the 0.9øC ZVS by this comparisonbecausethe transfor it are muchlargerthan the 1.0ø and
With a ZVS at a lower temperature there is little or port differences
no northward bottom flow of Circumpolar Deep Water 1.1øC ZVSs both overall and in individual bins. While
in the Amirante Passage.With a ZVS at a higher tem- there is not a large depth changeover the remaining
perature someportion of the flow of Indian Deep Water ZVS rangeof 1.0ø to 1.1øC,it is within the high-shear
JOHNSON ET AL.' AMIRANTE
PASSAGE DEEP CIRCULATION
30,981
Table 2. MeanandStandard
Deviations
forVolumeTransports
(in 106m3 S--1)in the Mascarene
Basinand the Amirante PassageFrom VariousZVSs to the Bottom (Bottom Water or Circumpolar
DeepWater) and FromTheseZVSsto 2.5øC(DeepWater or Indian DeepWater)
ZVS,
øC
Mascarene Basin
ZVS to Bottom
1.3
1.2
1.1
1.0
0.9
0.8
0.7
5.4
4.8
3.8
2.5
1.4
0.5
0.0
-4- 1.1
5= 1.0
+ 0.8
:k: 0.6
+ 0.5
5= 0.3
5= 0.0
Amirante Passage
ZVS to 2.5øC
-1.4
-3.3
-6.4
-11.6
-17.6
-25.2
-31.3
ZVS to Bottom
5= 0.4
5= 0.4
+ 0.6
:k: 0.8
+ 0.8
5= 0.9
5= 1.8
2.8
2.4
1.7
1.0
0.4
0.0
0.0
5= 1.4
5= 1.2
+ 0.9
+ 0.6
+ 0.3
5= 0.0
5= 0.0
ZVS to 2.5øC
1.1
-0.3
-3.8
-8.6
-11.8
-16.6
-16.1
5= 1.0
5= 1.4
+ 1.9
:k: 2.6
+ 3.2
5= 3.5
5= 3.2
Difference
Bottom
-0.1
-0.4
-0.6
-1.5
-4.8
-8.1
-15.2
to 2.5øC
5= 1.7
-4- 1.8
+ 1.6
:k: 1.5
+ 1.9
5= 2.4
5= 3.9
Deviations
for volumetransports
aregivenin 106m3 s-•. The ZVSsshownrangeovermostof the deep
geostrophicshear. The differencecolumnshowsthe net imbalanceof the differencebetweenthe Mascarene
Basin and the Amirante Passagebetweenthe bottom and 2.5øC with the root mean squareof differences
for individual
0.1øC bins between the two locations.
region(Figure5), sosmallchanges
in ZVS depthresult on the practice of objective mapping for extrapolation.
in large changesin transport.
Within the Amirante Passagethe 1.0ø-1.1øC range
In the Amirante Passage, that quantity ranges from
of ZVSs resultsin transports(Table 2) from 1.0 to 1.7
0.6 to 1.0 x 106 m • s-• northward in the bottom water and from 0.4 to 0.1 x 106 m • s-1 southward in the
x 106 m 3 s-1 northward for the bottom water below
the ZVSs and from 8.6 to 3.8 x 106 m • s-1 southward
deep water for ZVSs from 1.0ø to 1.1øC. In the Mascarene Basin the transport in the bottom triangles is
for the deepwater betweenthem and 2.5øC (the upper 0.9-1.3 x 106 m 3 s-1 northward in the bottom water
bound chosenfor the Indian Deep Water). Fluxes in and0.2-0.2 x 106m• s-1 southward
in the deepwater.
0.1øC• bins (Figure7) revealthat the bulk of the bot- Therefore, in the Mascarene Basin and the Amirante
tom water transport occurs around 0.85øC, and above
0.8øC the mean fluxes exceed their standard
deviations
within the bottom water. In the deep water, southward
transportis at a maximumat 1.55øC(Figure 7), near
the core of the low-magnitudeQ water. This relation is
consistentwith the hypothesisthat this low-magnitude
Q is a signatureof southwardspreadingof Indian Deep
Water (Figure 5). Below 1.6ø-1.4øC,mean fluxesexceed their standard deviations in the deep water.
In the MascareneBasin the 1.0ø-1.1øCrangeof ZVSs
resultsin a bottom water transport(Table 2) between
the ZVSs and the sea floor from 2.5 to 3.8 x 106 ma s-1
northward toward the passageand a deep water trans-
portbetween
theZVSsand2.5øCfrom11.6to 6.4 x 106
ma s-1 southward.Fluxesin 0.1øC0 bins (Figure7)
show the most transport near 0.75øC in the bottom
water, consistentwith warming of bottom water from
Passage, only a few percent of the deep water transports are estimated to reside in the bottom triangles,
since little of the deep water is found in the bottom triangles. However, over half of the total bottom water
transport in the Amirante Passageand a third of the
bottom water transport in the MascareneBasin are estimated to residein the bottom triangles. Extrapolation,
even if by the more sophisticatedprocedure of objective mapping, still accountsfor a large fraction of the
total bottom water transport reported in all Amirante
Passagestudies.
5.
Discussion
At higher latitudes in the Indian and Pacific Oceans,
bands of southern deep water at the western boundaries appear to flow north at middepth. Our conclu-
betweenthesesectionsand the Amirante Passage.The sionsabout transports depend on identifying the ZVS
maximum in southwarddeep water transport is again with an isothermalsurface,an intuitively appealingasat 1.55øC, consistentwith the findingsin the Amirante sumption with no surer computational alternative, but
Passage. While the small standard deviations for the difficult to reconcilewith data at higher latitudes. As
MascareneBasin suggestthe two sectionsgoing into remarkedpreviously[Warren, 1981], at 18øSa thick
the estimate have similar geostrophicshear, much less band of relatively fresh, oxygen-rich water was interweight shouldbe placedon their magnituderelative to posedat middepth between Madagascar and the Indian
the mean than the same relation in the Amirante PasDeep Water offshore. This feature in the water propsage, where statistics are derived from six sections.
erties indicated that the Circumpolar Deep Water was
A final quantity of interest is the amount of trans- flowing northward not only beneath the Indian Deep
port assignedto the bottom triangles, because that Water, but also beside and to the west of it, so that the
shows how much of the total estimate is dependent ZVS there deviated greatly from a single isotherm. No
30,982
JOHNSON ET AL.: AMIRANTE
PASSAGE DEEP CIRCULATION
3
a
2.5
,•
I
ß i-•,
ß
1.5
'1
•
:
, I
•'
t
I•
........ L•..•
..... •
' I
!
I-H
r
i,
I
'1
I
ß
•_•,m
.
[
i
i
..
-1
-2
0
1
2
•
6•
-1
-2
0
1
2
Transport
(106m3s-1)
Transport
(106
m3s-])
3
2.5
m
'
•
'
1.5 '1.........
i
I •
.t
P
.f.j
-t
m'
i
:
•
i
i
:
' I]I
.......... :.........
-1
;......1..
'1
ß ,--•___•
-
1
........ i............
- •
-2
-
I
I
I
'
ßU3C•
'
:'; ' .
0
1
2
Transport
(106m• s-])
-2
-1
0
1
Transport
(106m3s-])
Figure 7. Plots of mean geostrophic
volumetransport (bars) and standarddeviations(error
bars)in 0.1øCbinsin the AmirantePassage(Figures7a and 7c) and Mascarene
Basin(Figures
7b and 7d) for zero-velocity
surfacesof 1.0øC(Figures7c and 7d) and 1.1øC(Figures7a and
7b). CircumpolarDeep Water flowsnorth belowthe zero-velocitysurfaces,and Indian Deep
Water flows south above them. The net flow of Indian Deep Water southwardexceedsthat of
the Circumpolar Deep Water northward.
such anomalous middepth feature was found near the
Amirante Passage,however,so the ZVS may indeedlie
closeto an isothermal surfacethere, as assumed,but the
of these narrow bands of northward
flow at these low-
latitude passagesis uncertain. It seems possible that
they may yet be found to the west of the deep passages,
circumstances to the south demonstrate that this need banked against shallowertopography.
Measurements at the Samoan Passagesuggesta net
not be necessarilyso. We are somewhatreassuredby
a potentiallyanalogous
situationin the SouthPacific, northward Circumpolar Deep Water transport of 7.8
et al., 1996],severaltimes
wherean array of deepcurrentmetersat 32.5øSreveals x 106 ma s-• [Roemmich
a narrow band of northward flowing Circumpolar Deep larger than the estimate for the Amirante Passage.In
Water betweenthe Kermadec Ridge and the southward addition, deep values of the chlorofiuorocarbonsCFCflowingPacificDeepWater offshore[T. WhitworthIII 11 and CFC-12 are undetectable in the recent WOCE
et al., 1998],but near10øSin the Samoan
Passage
(an surveysof the Amirante Passage.By comparison,CFCabyssalconstrictionin the SouthPacificOceansimilar 11 has been found in the major Pacific deep westernto the Amirante Passagein the SouthIndian Ocean) boundary current as far north as the SamoanPassageas
the ZVS appearsto be well approximated
by the 1.2øC of March 1996 (J. Bullister,personalcommunication).
potentialisotherm[Roemmich
et al., 1996]. The fate
Since the CFCs are transient
tracers that are introduced
JOHNSON ET AL.' AMIRANTE
into the ocean in regionswhere water is exposedto the
surface, this differencesuggeststhat less Circumpolar
Deep Water reachesthe Amirante Passagethan reaches
the Samoan Passage,consistentwith the transport estimates. Other possibilitiesfor the CFC concentration
contrastare that CircumpolarDeep Water reachingthe
Amirante Passagetakes a longer path, is slower, or is
diluted more with less recently ventilated waters than
that reaching the Samoan Passage.
Our range of valuesfor northward flow of Circumpolar Deep Water is at the low end of the general run of
previouslyreported valuesfor northward flow of abyssal
PASSAGE DEEP CIRCULATION
30,983
A more complete understanding of the northward
flow of bottom water and the southward flow of deep
water requires a basin-wide synthesisof the WOCE Indian Ocean data. One peculiar feature in the Somali
Basin that may have some bearing on the southward
flow of Indian Deep Water is the strong thermostad
centered near 1.5øC near 3øS off the African Coast visi-
ble in historicalhydrographic
sections
there[Johnson
et
al., 1991b;Warrenand Johnson,1992].This bolushas
not been discussedpreviously, but its low-magnitude
Q connectsit with the low-magnitude Q water traveling south through the Amirante Passageand Mas-
water in the westernbasinsof the Indian Ocean (Ta- careneBasin[O'Dwyerand Williams,1997].This lowble 1). The presentestimatesin the AmirantePassage magnitudeQ likely originatesas a result of Indian Deep
agreewell with thoseof Fieux and Swallow[1988],but Water crossingthe equator near the African Coast, or
while our estimates are low becausewe place the ZVS
within the deep shear rather than above it, theirs are
likely low becausethe station pairs used did not completely span the passage. In contrast with our ZVS
choice, the larger estimates for northward Circumpolar Deep Water flo•v place the ZVS mostly above the
strongdeepshear. Hence most previousstudiesimplic-
itly underestimatedthe significanceof southwardflow
of Indian Deep Water with respectto this analysisby
choosinglighter ZVSs. In contrastwith the historical
assumptionof little southwardflow in the region,the
estimatehere is of a larger southwardflow of deepwa-
it could alternatively be supplied south of the equator
by flow from the east.
Asnotedbyprevious
investigators
[FieuxandSwallow,1988;Barton andHill, 1989],flowof bottomwater
into the Somali Basin through the Amirante Passage
has no exit save through upwelling. Using the volume
of the SomaliBasinbelow4000m, 1.5 x l015 m3 for
the 1.1øCZVS, andthat below4300m, 0.8 x 1015m3,
for the 1.0øC ZVS, togetherwith the volumetransports
reported here for thoseZVSs, givesa Circumpolar Deep
Water residence time near 25 years, which could be
halved or doubled by transport uncertainties. A mass
ter than the northward flow of bottom water beneath it
budgetusinga basinareaof 2.8 and2.1 x 10TMm2 for
(Figure7). Given that Indian DeepWater is presum- the 1.1ø and 1.0øC ZVSs, respectively,requiresdeepupably transformedCircumpolarDeep Water returning wellingvelocities
near5 x 107m s-1 , whichagaincould
southward, one asks how this could be. The answer be halved or doubled within uncertainty. Finally, ap-
may lie in the fact that the Indian Ocean has many
deepbasins. The aggregateof CircumpolarDeep Water upwelledin the easternbasinsto a depthat whichit
is no longerconstrainedby their topographymight be
expectedto flow westwardand then return southward
as Indian Deep Water in a largedeepwestern-boundary
current alongMadagascar,but whetherthat aggregate
is largeenoughto supplyour estimatedsouthward
flow
is not yet known.
plyinga bottomgeothermal
heatingof 0.05W m-2 and
a deep vertical temperature gradient within the basin
of 4 x l0 -4 øC m-1 [Johnson,
1990]to the abyssal
heat budgetdemandsheatingof 0.2 and 0.6 W m-2
and vertical diffusivities of 1.5 and 3.5 x 10-4 m 2 s-1
at the 1.0ø and 1.1•øCZVSs, respectively.Again, these
values can be halved or doubled by the transport uncertainties. The vertical mixing estimateslie within the
range of previousabyssalmixing estimatesfor the So-
The ZVS range chosen here produces noticeably mali Basin[Fieuxand Swallow,1988;Barton and Hill,
smaller flowsof both Circumpolar Deep Water and In- 1989].
dian Deep Water in the Amirante Passagethan in the
Additional evidenceof significant mixing is found in
Mascarene
Basin(Figure7). If thereis little net flowin the coldest water within the Amirante Trench, the isothe sidesof the box defined by the sections,this result lated portionof the passage
deeperthan 4750m [Smith
impliesa significant
(about1.8 x l06 m3 s-1) conver- and Sandwell,1997]. This evidenceis in the distribusion of Circumpolar Deep Water to Indian Deep Water in the northern Mascarene Basin, suggestiveof a
very vigorousmeridionaloverturningcirculationwhere
northward flowingbottom water returns as southward
flowingdeep water. However,it shouldbe noted here
that a rather large gap existsbetweenthe east end of
the Amirante Passagesectionsand the east end of the
MascareneBasinsections(Figure2), and the difference
in transportsbetweenthe two setsof sections,if significant, could simply be owing to flow through this large
gap.
tion
of 0 within
the trench
from
the six sections an-
alyzed here. In both 1987 (CD24S versusCD22 and
CD24) and in 1995-1996(I2 versusIR7N and I7N), 0
within the trench increasesby about 0.04øC over about
100 km from south to north with little indication
of tem-
poral variability, suggestingstrongmixing near the bottom there, heating the abyssalwater as it flows northward. A heat budget for the near-bottom isothermsin
the vicinity of the SamoanPassagealso implies signif-
icant mixing in the coldestwater there [Roeromichet
al., 1996].
30,984
JOHNSON ET AL.: AMIRANTE
Acknowledgments.
G.C.J. was funded by the National Oceanic and Atmospheric Administration Office of
Global Programs through the Climate and Global Change
Program.
A.F. was funded through NOAA grant
NA66GP0267. D.L.M., B.A.W., and D.B.O. were funded
by the National ScienceFoundation through grants OCE9401989, OCE-9413156, and OCE-9413165, respectively.
None of this analysis would have been possiblewithout the
hard work on the part of the shipboardscientificparties, the
officersand crew of the R/V Knorr and NOAA Ship Malcolm Baldrige, and those who helped to calibrate, process,
and analyze the data after the completion of the cruises.
PMEL
contribution
1872.
References
Barton, E. D., and A. E. Hi]], Abyssal flow through the
Amirante Trench (westernIndian Ocean), DeepSea Res.,
Part A, 36, 1121-1126, 1989.
Fieux, M., and J. C. Swallow, Flow of deep water into the
Somali Basin, Deep Sea Res., Part A, 35, 303-309, 1988.
Fieux, M., F. Schott, and J. C. Swallow, Deep boundary
currents in the western Indian Ocean revisited, Deep Sea
Res., Part A, 33, 415-426, 1986.
Fu, L.-L., Mass, heat and fleshwater fluxes in the South
Indian Ocean, J. Phys. Oceanogr., 16, 1683-1693, 1986.
Johnson, G. C., Near-Equatorial Deep Circulation in the
Indian and Pacific Oceans,Ph.D. thesis, 157 pp., MIT/
WHOI, Joint Program in Oceanogr., Woods Hole, Mass.,
1990.
Johnson, G. C., B. A. Warren, and D. B. Olson, A deep
boundary current in the Arabian Basin, Deep Sea Res.,
Part A, 38, 653-661, 1991a.
Johnson, G. C., B. A. Warren, and D. B. Olson, Flow of
bottom water in the Somali Basin, Deep Sea Res., Part
A, 38, 637-652, 1991b.
Lee, T., and J. Marotzke, Inferring meridional mass and
heat transports of the Indian Ocean by fitting a general
circulation model to climatological data, J. Geophys.Res.,
PASSAGE DEEP CIRCULATION
data and derived fields, J. Phys. Oceanogr., 13, 15441549, 1983.
Roemmich, D., S. Hautala, and D. Rudnick, Northward
abyssaltransport through the Samoan passageand adjacent regions, J. Geophys.Res., 101, 14,039-14,055, 1996.
Smith, W. H. F., and D. T. Sandwell, Global seafloor topography from satellite altimetry and ship depth soundings,
Science, 277, 1956-1962, 1997.
Spencer, D., W. S. Broecker, H. Craig, and R. F. VVeiss,
GEOSECS Indian Ocean Expedition, vol. 6, Sections and
Profiles, 140 q- xiv pp., Nat. Sci. Found., Washington,
D.C., 1982.
Stommel, H., and A. B. Arons, On the abyssalcirculation
of the world ocean, I, Stationary planetary flow patterns
on a sphere, Deep Sea Res., 6, 140-154, 1960.
Swallow, J. C., and R. T. Pollard, Flow of bottom water
through Madagascar Basin, Deep Sea Res., Part A, 35,
1437-1440, 1988.
Toole, J. M., and B. A. Warren, A hydrographicsection
acrossthe subtropical South Indian Ocean, Deep Sea Res.,
Part I, J0, 1973-2019, 1993.
Warren, B. A., Deep flow in the Madagascarand Mascarene
basins, Deep Sea Res., Part A, 21, 1-21, 1974.
Warren, B. A., Shapesof deepdensity-depthcurves,J. Phys.
Oceanogr.,7, 338-344, 1977.
Warren, B. A., Bottom water transport through the Southwest Indian Ridge, Deep Sea Res., 25, 315-321, 1978.
Warren, B. A., Transindian hydrographicpropertiesat Lat.
18øS:Property distributions and circulation in the South
Indian Ocean, Deep Sea Res., Part A, 28, 759-788, 1981.
Warren, B. A., and G. C. Johnson, Deep currents in the
Arabian Sea in 1987, Mar. Geol., 104, 279-288, 1992.
Whitworth, T., III, B. A. Warren, W. D. Nowlin, Jr., S.
B. Rutz, R. D. Pillsbury, and M. L. Moore. On the deep
western-boundary current in the SouthwestPacific Ocean,
Prog. Oceanogr.,in press, 1998.
Wyrtki, K., OceanographicAtlas of the International Indian
OceanExpedition, 521 q- xi pp., Super. of Doc., U.S. Gov.
Print. Off., Washington, D.C., 1971.
102, 10,585-10,602, 1997.
Mantyla, A. W., and J. L. Reid, On the origins of deep and
bottom waters of the Indian Ocean, J. Geophys.Res., 100,
2417-2439, 1995.
O'Dwyer, J., and R. G. Williams, The climatological distribution of potential vorticity over the abyssal ocean, J.
Phys. Oceanogr., 27, 2488-2506, 1997.
Pedlosky, J., GeophysicalFluid Dynamics, 710 pp., SpringerVerlag, New York, 1987.
Quadfasel, D., J. Fischer, F. Schott, and L. Stramma, Deep
water exchange through the Owen Fracture Zone in the
Arabian Sea, Geophys. Res. Lett., 2•, 2805-2808, 1997.
Reid, J. L., and P. F. Lonsdale, On the flow of water through
the Samoan Passage, J. Phys. Oceanogr.,J, 58-73, 1974.
Robbins, P. E., and J. M. Toole, The dissolvedsilica budget
as a constraint on the meridional overturning circulation
of the Indian Ocean, Deep Sea Res., Part I, JJ, 879-906,
1997.
Roemmich, D., Optimal estimation of hydrographic station
A. Ffield, Lamont-Doherty Earth Observatory,Columbia
University, Palisades, NY 10964. (e-mail: affield@ldeo.
columbia.edu)
G. C. Johnson,Pacific Marine EnvironmentalLaboratory,
NOAA, 7600 Sand Point Way, NE, Building 3, Seattle, WA
98115-0070.(e-mail: [email protected])
D. L. Musgrave, University of Alaska Fairbanks/SFOS,
P.O. Box 757220,Fairbanks,AK 99775-7220.(e-mail: [email protected])
D. B. Olson, RosenstielSchoolof Marine and Atmospheric
Science,University of Miami, 4600 RickenbackerCauseway,
Miami, FL 33149. (e-maih [email protected])
B. A. Warren, Department of Physical Oceanography,
Woods Hole Oceanographic Institution, Woods Hole, MA
02543. (e-mail: [email protected])
(ReceivedAugust 19, 1997; revisedAugust 10, 1998;
acceptedSeptember10, 1998.)