You are on page 1of 14

The seasonal variability of the circulation in the South Indian Ocean:

Model and observations


R.P. Matano
a,

, E.J. Beier
b
, P.T. Strub
a
a
College of Oceanic and Atmos. Sciences, Oregon State University, Corvallis, OR, 97331-5503, United States
b
CICESE Unidad La Paz, La Paz, B.C.S. Mexico
Received 20 November 2006; received in revised form 12 January 2008; accepted 16 January 2008
Available online 14 February 2008
Abstract
This article compares the seasonal variability patterns of the South Indian Ocean circulation derived from a global, eddy-
permitting, numerical model and altimeter observations. The seasonal variability of the Indian Ocean circulation is driven by the
inflow from the Indonesian Passages and by the local wind forcing. Our analysis indicates that the influence of the Indonesian
throughflow is confined to the easternmost portion of the basin, while the influence of the wind stress forcing is important
everywhere. Model and observations indicate that, between ~105E and 75E, the seasonal variability is characterized by the
southwestward propagation of an annual wave over a period of ~4 months. Preliminary calculations using Pathfinder data also
indicate that, in the western region, there are seasonal perturbations that originate in the tropics and propagate poleward through the
Mozambique Channel. Our calculations, however, did not find the connections between the tropical and the Agulhas Current
variability suggested by earlier modeling studies.
2008 Elsevier B.V. All rights reserved.
Keywords: Western boundary; Seasonal variability; Modeling; South Indian Ocean; Model and observation
1. Introduction
The South Indian Ocean circulation includes a trop-
ical cyclonic gyre, which extends from the equator to
approximately 17S, and a subtropical anticyclonic gyre
that extends from approximately 17S to 47S (see
Lutjeharms, 2007 and references therein) (Fig. 1). These
gyres are the conduit through which Pacific waters
entrained by the Indonesian throughflow are discharged
onto the eastern Atlantic by the Agulhas retroflection
(Gordon, 1985; Gordon et al., 1999; Lutjeharms, 1996).
Since the entrainment from the Pacific occurs in the
tropics, and the detrainment onto the Atlantic in the
subtropics, the Indian Ocean circulation has to transfer
mass and energy between its gyres. These transfers are
largely accomplished by eddies and narrow currents
within the Mozambique Channel and east of Madagascar
(De Ruijter et al., 2002, 2004; Schouten et al., 2002;
Ridderinkhof and de Ruijter, 2003; Quartly and Srokosz,
2004; Lutjeharms, 2007). The variability of these eddies
and currents are linked to the large-scale circulation and
are therefore modulated by its low-frequency variability
(e.g., Schouten et al., 2002; Palastanga et al., 2006). Thus,
our ability to understand the dynamical mechanisms that
participate in the global interocean exchanges is tied to
our ability to understand the process that controls the
Available online at www.sciencedirect.com
Journal of Marine Systems 74 (2008) 315328
www.elsevier.com/locate/jmarsys

Corresponding author.
E-mail address: rmatano@coas.oregonstate.edu (R.P. Matano).
0924-7963/$ - see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmarsys.2008.01.007
variability of the Indian Ocean gyres. In this article we
investigate one of the most distinctive modes of low-
frequency variability of the Indian Ocean circulation, the
seasonal cycle.
Previously we investigated the seasonal variability of
the South Indian Ocean using the results of a global,
eddy-permitting, numerical simulation (Matano et al.,
2002; henceforth M2002). This analysis showed that at
seasonal time scales, the large-scale circulation in the
South Indian Ocean is modulated by the oscillation of
barotropic modes forced directly by the wind. The pul-
sation of these modes generates a seasonal variation of
the transport of the Agulhas Current, with a peak at the
transition between the austral winter and the austral
spring, and a trough at the transition between the austral
summer and the austral autumn. The modes described
by M2002 appear to be a robust characteristic of other
numerical simulations of the South Indian Ocean cir-
culation (e.g., Matano et al., 1998; Biastoch et al., 1999;
Fetter et al., submitted for publication). The existence of
these modes, however, has not been confirmed by ob-
servations, i.e., most of what we know about the sea-
sonal variations of the South Indian Ocean circulation
has been inferred from model results without a direct
verification from real data. It seems, therefore, of broad
interest to determine whether the variability structure
predicted by numerical models are corresponded by
observations. As a step in that direction in this article we
compare the results of the numerical simulation de-
scribed by M2002 with altimeter data. We show that
the structure of the seasonal variability deduced from
models is similar to that obtained from observations.
This finding indicates that regional processes, i.e., pro-
cesses occurring within sub-basins determined by the
bottom topography, control the seasonal variability of
the South Indian Ocean circulation.
2. Data and methods
2.1. Model
The model results were produced by the Parallel
Ocean Circulation Model (POCM) experiment (Tokma-
kian and Challenor, 1999). The model configuration
and numerical results have been described in M2002
therefore only a brief discussion will be included here.
The model solves the primitive equation in a Mercator
grid with an average horizontal grid spacing of 1/4
and 20 vertical levels. Our analysis is focused on the
experiment 4C, which was run for a 19-year period.
From 1979 to 1994 the model was forced with at-
mospheric fluxes derived from the reanalysis of the
European Center for Medium-Range Weather Forecast,
after that the forcing fluxes were replaced with data
from operational forecast experiments.
2.2. Altimeter data
T/P and ERS data, reprocessed by the NASA/GSFC
Pathfinder Ocean Altimeter group, are the basis for
the gridded fields of sea surface height. To determine
the 0spatial structure of the large-scale variability we use
T/P data which was interpolated using a loess smoother
filter with 8 in longitude, 2 in latitude and 25 days in
time with estimates at 3 days intervals on a one degree
grid. A detailed discussion of the data processing can be
Fig. 1. Bottom topography of the Tropical and the South Indian Ocean. Grey arrows mark the mean geostrophic circulation according to Stramma and
Lutjeharms (1997). The boxes in the stippled rectangles mark the approximate domains of the three modes of annual variability defined by Matano
et al. (2002).
316 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
found in Chelton et al. (1990). To reduce the steric
effect associated with the annual variation of the so-
lar radiation we subtracted the zonal average of the
SSHAs from the original fields. Since the interpola-
tion procedures used by Chelton et al. eliminates most
of the data in the Mozambique Channel and surround-
ing area we did some ancillary calculations using higher
resolution data from the Pathfinder mission.
To characterize the variability from the model and the
observations we calculated the standard deviations, an-
nual harmonics, Hovmller diagrams, and EOF modes
of the sea surface heights anomalies (SSHAs). To fa-
cilitate the comparison we used only data of the period
during which both data sets overlap i.e., from 1993
through 1998. Our study will focus in the Indian Ocean
region between 5N and 30S.
Fig. 2. Standard deviation [cm] from T/P data (panel a) and POCM_4C (panel b). The values were calculated from time series from the beginning of
1993 to the end of 1997. The shadow areas represent regions where the values are bigger than 3 cm. Panel c shows the time evolution of the basin
averaged sea level elevation from both time series.
317 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
3. Results
The M2002 analysis indicates that the seasonal var-
iabilities of the tropical and subtropical Indian Ocean
are connected by three barotropic modes. The first is lo-
cated in the eastern portion of the basin and reaches its
maximum amplitude in the Indonesian Passages. The
second mode is centered in the middle of the basin and
it is limited, in the zonal direction, by the bottom to-
pography. The third mode is restricted to the western
portion of the basin and encompasses two distinctive
regions. To the south it extends into the Mozambique
Channel and reaches up to the Agulhas Current. To the
north it extends up to the equator and encompasses
the South Equatorial Current. M2002 concluded that
while the first mode is largely driven by the seasonal
variations of the Indonesian throughflow, the other two
are forced by the wind stress forcing. These conclusions
were in general agreement with the earlier modeling
results of Biastoch et al. (1999) who also noted that
seasonal anomalies in the tropical circulation propa-
gate poleward through the Mozambique Channel and
influence the mass transport of the Agulhas Current.
The seasonal variations described from these models,
Fig. 3. Amplitudes (background color) and phases (white contours) of the annual harmonic from altimeter observations (upper panel) and the POCM-
4C experiment (bottom panel). The units of the amplitudes are in cm and the phases are in months. A color version of this figure could be found in the
online version of this article.
318 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
however, have not yet been verified by observations. As
a step in that direction in what follows we compare the
results of POCM with altimeter observations.
3.1. Standard deviations
To characterize the overall variability of both data
sets we computed the standard deviations of the SSHAs
(Fig. 2). There is a reasonable agreement between model
and observations although the former shows small-scale
structures that are absent in the coarsely sampled al-
timeter data (Fig. 2a and b). The largest differences are
observed near coastal regions (e.g., the Mozambique
Channel, the eastern Madagascar coast), which are out-
side of the resolution of the altimeter data. The region
between 4 and 18 S shows the largest variability in
SSHAs, with peaks of 14 cm near Indonesia and 11 cm
near 80E 10S. Most of the variance in both data
sets is contained in the annual and inter-annual periods
(Fig. 2c). The annual cycle explains 32% of the vari-
ance of the T/P data and 37% of POCM. The ampli-
tude of the annual fit of the basin averaged T/P data is
1.3+/0.15 cm with a maximum during the middle of
December and a minimum during July. The amplitude of
the model data is 1.1+/0.13 cm. The phase difference
between both data sets is 10 days.
Fig. 4. (a) Hovmller diagrams of the SSHA from model (top panel) and the altimeter data (middle panel) at 8S. The bottom panel shows the bottom
topography of the model at the same latitudinal strip. The units of SSHA are in cm. (b): Hovmller diagrams of the SSHA from model (top panel) and
the altimeter data (middle panel) at 12S. The bottompanel shows the bottomtopography of the model at the same latitudinal strip. The units of SSHA
are in cm.
319 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
3.2. Harmonic analysis
The amplitudes and phases of the annual harmon-
ics from both data sets are very similar (Fig. 3). The
amount of variance explained by the annual harmonic
is generally high, with a maximum of ~60% over the
central region and ~40% near the eastern and western
boundaries (not shown). Both data sets show amplitude
maxima (background color) in the northeastern and
northwestern portions of the basin, although the spatial
structures of the model show more mesoscale struc-
ture due to their higher spatial resolution. The model,
for example, shows two relative maximum in the north-
eastern region: one centered at ~90E and the other over
the Indonesian Passages, while the altimeter data shows
a single maximum that extends from ~80E to 120E.
The model maxima have been also reported in previous
observational studies, which argue that they are driven
by different dynamical processes: the maximum near
90W is driven by the local winds, while the one cen-
tered in the Indonesian Passages is driven by the inflow
from the Pacific Ocean (Woodberry et al., 1989; Inoue
and Welsh, 1993, Morrow and Birol, 1998; Gordon
et al., 1999). Model and observations also show high
amplitudes in the northwestern portion of the basin
(Fig. 3). The observations show an absolute maximum
near the northeastern tip of Madagascar and a secondary
maximum centered near 5S 48E. This maximum is
displaced to the northwest in the model and it is driv-
en by the seasonal variations of the South Equatorial
Countercurrent.
The distribution of phases (white contours), indicate
that the eastern and western regions are not connected
by annually propagating waves but are separated by a
region of vanishing amplitudes (in both data sets), and an
amphidromic line (Fig. 3). The most robust propagating
Fig. 4 (continued).
320 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
signal of both data sets is observed in the northeast-
ern portion of the basin. There, there is a southwest-
ward propagating signal that moves between 105 and
75E in a period of ~4 months. This signal was previ-
ously reported, and extensively described, by Perigaud
and Delecluse (1992) and Morrow and Birol (1998). The
model results indicate the existence of westward prop-
agation in the western portion of the domain (west of
Fig. 6. Amplitude of the annual harmonic derived from the model (upper panel) and from the Pathfinder altimeter data (bottom). The amplitudes are
given in cm. The stars in the left panel mark the location of the data (both model and altimeter) that we used to calculate Fig. 6.
Fig. 5. Amplitudes (upper panel) and phases (lower panel) of the annual harmonic of the zonally averaged sea surface elevation from the altimeter
(grey dashed contours) and the model (black solid line). The bars indicate the error of the fit.
321 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
~75E), but this signal is not obvious in the observations.
The region that separates the eastern and western am-
plitude maxima follows the approximate course of the
Mid-Indian Ridge and the Ninety East Ridge (Fig. 1).
M2002 observed that these topographic features interrupt
the westward progression of the annual signal in POCM.
These signals appear to be particularly sensitive to the
presence of the Mid-Indian Ridge, the Chagos Archipe-
lago, and the Mascarene Plateau. M2002 argued that the
topographic effects enhance the contribution of the local
forcing (in detriment to remote forcing), and that the
seasonal adjustment of the South Indian Ocean is there-
fore better described in terms of standing modes rather
than in terms of propagating waves. According to this
argument the transport of the Agulhas Current, for ex-
ample, is largely unaffected by the seasonal changes of
the winds in the eastern portion of the Indian basin (east
of ~75E), or the seasonal variations of the Indonesian
Throughflow. These arguments are in agreement with the
earlier modeling results of Kindle (1991) who reported
that the predominantly barotropic, seasonal signal that
is generated in the eastern portion of the Indian basin is
reflected at the Mascarene Plateau.
The M2002 arguments were entirely based on the
POCM results. Therefore it is of interest to compare the
characteristics of the westward propagating signals de-
rived from POCM with those from the altimeter (Fig. 4a
and b). There is a general agreement between POCM
and observations although the amplitudes predicted by
the model are smaller than those calculated from the
altimeter. The phase speeds derived from POCM, how-
ever, are close to those found in T/P and for this range of
latitudes and to the values predicted by theory (Chelton
and Schlax, 1996). The topographic effects discussed
in M2002 are also observed in the altimeter data. At
8S, the Ninety East Ridge limits the westward progres-
sion of the signals generated near the eastern boundary
(Fig. 4a). New waves appear to be generated along the
eastern flank of the ridge (Fig. 4a). These waves appear
to be affected, farther west, by the Mascarene Plateau.
Fig. 7. Amplitudes and phases of the annual harmonic computed along the coastal points in Fig. 5. The black line correspond to the mode and the grey
line to the observations.
322 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
There are some noticeable differences between the mod-
el and observations in the region between 60E and
80E. The model shows a standing mode with a max-
imum near the western limit while the existence of such
a mode is not obvious in the observations. Both data sets
show similar amplitudes (~3 cm) to the west of the
Mascarene Plateau and the existence of a regional mode
between 60E and the African coast. The same general
characteristics are also observed at 12S (Fig. 4b). In
both cases the main discrepancies between the annual
fits to the model and the observations are located in the
central region. The modeled amplitudes of the annual fit
are also smaller in both cases.
To further quantify the fitness between observations
and model we calculated the annual harmonic of the
zonally averaged SSH for the model and the T/P ob-
servations (Fig. 5). As expected from the above dis-
cussion the amplitudes of both data sets show similar
latitudinal distributions (Fig. 5, upper panel). The largest
differences are observed in region between the equator
and 8S but even there the fitting error bars indicate that
these differences are not statistically significant. Model
and observations, however, show significant discrepan-
cies of their phases, with a maximum of 4 months, in the
latitudinal strip between the equator and 5S (Fig. 5,
lower panel). In that region however, the amplitudes are
relatively small. Farther south the sea level reaches a
maximum (in both model and observations) between
February and March and a minimum between August
and September. In the overall balance this calculation
shows that there is a high correlation between model and
observations outside from the tropics, the correlation
Fig. 8. (a)Amplitudes of the first EOF from the observations (top panel) and the model (middle panel). The bottom panel shows their corresponding
time series. The units of amplitudes are in cm. A color version of this figure could be found in the online version of this article.
323 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
coefficient of annual harmonics of the SSHs from model
and observations is approximately 0.8.
The above analysis indicates that the seasonal var-
iability of the South Indian Ocean circulation can be
explained in terms of basin modes that draw their energy
from the local winds and that are delimited by the bot-
tom topography. Those conclusions are in agreement
with the earlier findings of M2002 whom, in addition,
showed that the tropical variations of the western-
most mode propagates poleward into the Mozambique
Channel and, from there, into the Agulhas Current. It is
not possible to corroborate these model predictions from
the coarsely gridded T/P data used in this analysis (e.g.,
Fig. 3). To address these matters however, we re-
calculated the annual harmonic of the western region
using a data set with higher spatial resolution (Pathfin-
der) (Fig. 6). The new calculation indicates that the
spatial structures derived from the altimeter are qual-
itatively similar to those from the model although the
amplitudes are smaller and there is no evidence of the
intense western intensification that characterize the
western boundary region. To investigate the existence
of poleward propagating signals in the Mozambique we
computed the amplitudes and phases of the annual
harmonics at the nearshore locations marked in Fig. 6
(Fig. 7). The altimeter data used for this calculation
correspond to the first ocean grid point along the de-
scending tracks (~50 km from the coast), while the
location from the model data corresponds to the grid
point closest to the observations. There is a close cor-
respondence between model and observations. The larg-
est discrepancies are observed in the equatorial region
where the model appears to overestimate the amplitude
of the seasonal cycle. South of ~5S, however, there is a
Fig 8b. (b): Amplitudes of the second EOF from the observations (top panel) and the model (middle panel). The bottom panel shows their cor-
responding time series. The units of amplitudes are in cm. A color version of this figure could be found in the online version of this article.
324 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
remarkable agreement between the phases and ampli-
tudes from both data sets. The distribution of phases
indicates the possible existence of propagation between
10 and ~20S. This signal, however, is lost near 20S,
perhaps due to the existence of a sharp bending of the
coastline that diverts the flow farther offshore (e.g.,
M2002).
3.3. EOF analysis
To further evaluate the realism of the numerical
simulation we calculated the Empirical Orthogonal
Functions (EOFs) of the SSHA anomalies of both data
sets (Fig. 8). The first three modes account for 49% of
the total variance of the observations and 46% of the
model. The first mode, which accounts for 23% and
24% of the variances of the observations and model,
reflects more the large-scale structure of the equatorial
circulation than the mesoscale structures associated with
the subtropical gyre (Fig. 8a). The amplitudes and time
series associated with both data sets are qualitatively
similar although the model amplitudes are slightly
smaller than those of the observations. The largest
SSHA gradients of both EOFs are located in the north-
eastern portion of the basin. The time series associated
with this mode seems to reflect more inter-annual
variations than seasonal changes. Note, for example,
that both data sets capture the 1995 and 1998 El Nios.
In fact, the annual fit to the time series indicates that
annual cycle accounts for only 36% of the variance in
the T/P mode and 25% in the POCM mode. As a side
calculation we extracted the annual signal from both
Fig. 8c. (c): Amplitudes of the third EOF from the observations (top panel) and the model (middle panel). The bottom panel shows their cor-
responding time series. The units of amplitudes are in cm. A color version of this figure could be found in the online version of this article.
325 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
Fig. 9. (a) Amplitudes of the first EOF mode [dyne/cm

3 10

9] of the wind stress curl (upper panel) and the corresponding time series (lower
panel). Shadow areas correspond to values bigger than 10 (solid) and 10 (stippled). (b): Amplitudes of the second EOF mode [dyne/cm

3 10

9]
of the wind stress curl (upper panel) and the corresponding time series (lower panel). Shadow areas correspond to values bigger than 5 (solid) and 5
(stippled).
326 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
data sets and re-calculated the corresponding EOFs. The
results (not shown) are similar to those shown herein
except that the westernmost maximum disappears and
there is only one maximum at 73E8S. In this cal-
culation the spatial structures of model and observed
EOFs at the Indonesian Passages are more similar to
each other than those shown in Fig. 8a.
The second EOFs represent 15% and 14% of the
variances of the observations and the model (Fig. 8b).
The annual fits to these modes explain 72% and 84% of
their respective variances, which indicates that this par-
ticular mode contains most of the seasonal signal. Both
EOFs show similar spatial and temporal structures, al-
though the amplitudes of the model are smaller than
those of the observations. North of 20S, both data sets
show two regional maximum centered near 90E and
65E. The corresponding maxima in the observations
are displaced to the northwest of those in the model,
although they both are within the region of maximum
SSHAvariation found in the harmonic analysis (Fig. 3).
There are relatively minor differences in the spatial
structures of the model and observations over the eastern
equatorial region. The model, for example, shows a
maximum at 83E 5S that is absent in the observations.
Both data sets, however, show the well-defined structure
of the South Equatorial Current near 5S.
The third EOF mode explains 11% and 9% of the
observations and models variance (Fig. 8c). This mode
also represents the seasonal variability with annual fit-
tings that explain 86% and 68% of the variance of the
observations and model. The largest differences be-
tween model and observations are in the eastern side of
the basin. There, the SSHA maxima that extends to-
wards the northwest in the altimeter is absent in the
model. This discrepancy seems to be related to the dom-
inance of the seasonal variability in this region. In fact,
the similarity between the EOFs is greatly improved
when the annual signal is extracted from the respective
time series of both data sets.
For the sake of completeness we also calculated the
EOFs of the curl of the ECMWF wind stress, which is
used to force POCM (Fig. 9a and b). The spatial struc-
ture of the first mode is largely uniform with no major
regional differentiation (except for a local maximum at
the tip of Madagascar), and therefore can be thought as
representations of the large-scale (hemispheric) circula-
tion (Fig. 9a). Most of the information on the basin scale
circulation is captured by the second EOF (Fig. 9b),
which explains 11% of the total variance (the annual
fit to this mode explains 54% of its own variance). One
of the most interesting characteristics of this mode is
the relative maximum in the eastern side of the basin
(~90E 11S), which closely corresponds with the am-
plitude maxima observed in the EOF and annual har-
monic of the SSHAs (Figs. 7 and 3 respectively). It
seems, therefore, reasonable to conclude that the wind
stress forcing locally drives the seasonal variability of
this region.
4. Summary and discussion
The seasonal variability of the Indian Ocean cir-
culation is driven by the inflow from the Indonesian
Passages and by the local wind forcing. The influence
of the Indonesian throughflow, however, appears to be
confined to the easternmost portion of the basin, while
the influence of the wind stress forcing is important
everywhere. Our analysis indicates that seasonal
variations of tropical origin propagate to the subtropics
in the central portion of the basin and in the western
region. In the central region, there is strong evidence of
an annual wave that propagates southwestward
between ~105E and 75E in a period of ~4 months.
The connection between the tropics and subtropics
in the western region is less robust. Preliminary cal-
culations using Pathfinder data, however, appears to
confirm the propagation of seasonal variations of trop-
ical origin through the Mozambique Channel. These
calculations, however, did not show the connec-
tions between the tropical and the Agulhas Current
variability suggested by earlier modeling studies (e.g.,
Biastoch et al., 2001, M2002; Reason et al., 2003).
This result is, perhaps, not surprising since the region
south of the Mozambique Channel is characterized
by high eddy variability and strong recirculation cells
that obscure the location of the mean flow (e.g., De
Ruijter et al., 2002; Riderinkhof and de Ruijter et al.,
2003). The problem is compounded by the fact that
the annual signal in the model appears to be trapped
close to the coast where it is difficult to compare model
with observations (e.g., Ffield et al., 1997). The recent
analysis of Palastanga et al. (2006), however, has
shown evidence on the dynamical connections between
the mesoscale variability around Madagascar and the
large-scale variability of the Indian Ocean.
Thus, our analysis confirms the model-based hy-
pothesis that regional barotropic processes control the
seasonal variability of the South Indian Ocean circula-
tion. The importance of this finding lies in the fact that
very little is known about the low-frequency variability
of this region. To the best of our understanding, for
example, there are only three studies that tried to address
these matters with observations: Quartly and Srokosz
(1993); Ffield et al. (1997), and Matano et al. (1998).
327 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328
These articles, however, were largely focused on the
Agulhas retroflection region and not in the basin scale
circulation. Most of our understanding of the seasonal
variability of the South Indian Ocean has been surmised
from modeling studies. These studies have a limited
academic value until validated by observations. In this
article, however, we show that the structure of the sea-
sonal variability deduced from model simulations is
very similar to that obtained from satellite observations.
The reasonable good correspondence between model
and observations is partially attributed to the fact that
seasonal variations are controlled by barotropic mo-
tions which, due to their relatively large scale, are well
resolved by eddy-permitting simulations, and are large-
ly insensitive to the details of the model stratification
or the particulars of the parameterization of sub-grid
processes (e.g., bottom friction, horizontal and vertical
diffusion, etc), which are the Achilles tendon of most
numerical simulations.
Acknowledgments
This article benefited from the criticism of two re-
viewers. The good sense of humor of one of them is
particularly appreciated. This work was supported by
NSF grant OCE-0726994, NASA grant NAG512378
and JPL contract 1206714.
References
Biastoch, A., Reason, C.J.C., Lutjeharms, J.R.E., Boebel, O., 1999.
The importance of flow in the Mozambique Channel to seasonality
in the greater Agulhas Current System. Geophys. Res. Lett. 26,
33213324.
Chelton, D.B., Schlax, M.G., 1996. Global observations of oceanic
Rossby waves. Science 272, 234238.
Chelton, D.B., Schlax, M.G., Witter, D.L., Richman, J.G., 1990. Geo
altimeter observations of the surface circulation of the Southern
Ocean. J. Geophys. Res. 101, 1413114145.
De Ruijter, W.P.M., Ridderinkhof, H., Lutjeharms, J.R.E., Schouten,
M.W., Veth, C., 2002. Observations of the flow in the Mozam-
bique Channel. Geophys. Res. Lett. 29 (10), 1502. doi:10.1029/
2001GL013714.
De Ruijter, W.P.M., van Aken, H.M., Beier, E.J., Lutjeharms, J.R.E.,
Matano, R.P., Schouten, M.W., 2004. Eddies and dipoles around
South Madagascar: formation, pathways and large-scale impact.
Deep-Sea Res., Part 1 51, 383400.
Fetter, A.J., Lutjeharms, J.R.E., Matano, R.P., 2007. Atmospheric
driving forces for the Agulhas Current in the subtropics. Geophys.
Res. Lett. 4, L15605. doi:10.1029/2007GL030200.
Ffield, A., Toole, J., Wilson, D., 1997. Seasonal circulation in the
South Indian Ocean. Geophys. Res. Lett. 24, 27732776.
Gordon, A.L., 1985. Indian-Atlantic transfer of thermocline water at
the Agulhas retroflection. Science 227 (4690), 10301033.
Gordon, A.L., Susanto, R.D., Ffield, A., 1999. Throughflow within
Makasaar Strait. Geophys. Res. Lett. 26 (21), 33253328.
Inoue, M., Welsh, S.E., 1993. Modeling seasonal variability in the
wind-driven upper-layer circulation of the Indo-Pacific region.
J. Phys. Oceanogr. 23, 14111436.
Kindle, J.C., 1991. Topography effects on the seasonal circulation of
the South Indian Ocean. J. Phys. Oceanogr. 96, 1682716837.
Lutjeharms, J.R.E., 1996. The exchange of water between the South
Indian and the South Atlantic. In: Wefer, G., Berger, W.H., Siedler,
G., Webb, D. (Eds.), The South Atlantic: Present and Past Cir-
culation. Springer-Verlag, Berlin, pp. 152162.
Lutjeharms, J.R.E., 2007. The Agulhas Current. Springer-Verlag. (330 pp).
Matano, R.P., Simionato, C.G., de Ruijter, W.P., van Leeuween, P.J.,
Chelton, D.B., Schlax, M.G., 1998. Seasonal variability in the
Agulhas retroflection region. Geophys. Res. Lett. 25, 43614364.
Matano, R.P., Beier, E.J., Strub, P.T., Tokmakian, R., 2002. Large-
scale forcing of the Agulhas variability: the seasonal cycle. J. Phys.
Oceanogr. 32, 12281241.
Morrow, R., Birol, F., 1998. Variability in the South Indian Ocean
from altimeter. Forcing mechanisms for the Leeuwin Current.
J. Geophys. Res. 99, 2656526583.
Palastanga, V., van Leeuwen, P.J., de Ruijter, W.P.M., 2006. A link
between low-frequency mesoscale eddy variability around Mada-
gascar and the large-scale Indian Ocean variability. J. Geophys.
Res. 111, C09029. doi:10.1029/2005JC003081.
Perigaud, C., Delecluse, P., 1992. Annual sea level variations in the
southern Indian Ocean from Geosat and shallow-water simulation.
J. Geophys. Res. 97, 2016920178.
Quartly, G.D., Srokosz, M.A., 1993. Seasonal variations in the region
of the Agulhas Retroflection: studies with Geosat and Fram.
J. Phys. Oceanogr. 23, 21072124.
Quartly, G.D., Srokosz, M.A., 2004. Eddies inthe Southern Mozambique
Channel. Deep-Sea Res., Part 2 51, 6983.
Reason, C.J.C., Lutjeharms, J.R.E., Hermes, J., Biastoch, A., 2003.
Inter-ocean fluxes south of Africa in an eddy permitting model.
Deep-Sea Res., Part 2 50, 281298.
Ridderinkhof, H., de Ruijter, W.P.M., 2003. Moored current obser-
vations in the Mozambique Channel. Deep Sea Res., Part 2 50,
19331955.
Schouten, M.W., de Ruijter, W.P.M., van Leeuwen, P.J., 2002. Up-
stream control of the Agulhas Ring shedding. J. Geophys. Res. 107
(C8), 3109. doi:10.1029/2001JC00804.
Stramma, L., Lutjeharms, J.R.E., 1997. The flow field of the subtrop-
ical gyre of the South Indian Ocean. J. Geophys. Res. 102 (C3),
55135530.
Tokmakian, R., Challenor, P.G., 1999. On the joint estimation of model
and satellite sea surface height anomaly errors. Ocean Model. 1,
3952.
Woodberry, K.E., Luther, M.E., O'Brien, J.J., 1989. The wind-driven
seasonal circulation inthe southern tropical Indian Ocean. J. Geophys.
Res. 94, 1798518002.
328 R.P. Matano et al. / Journal of Marine Systems 74 (2008) 315328

You might also like