1
The Indian Ocean HydroBase:
A high-quality climatological dataset for the Indian Ocean
Taiyo Kobayashi* and Toshio Suga
Institute of Observational Research for Global Change
Japan Agency for Marine-Earth Science and Technology
Submitted Progress in Oceanography on September 3, 2004
Revised on March 24, 2005
Accepted on July 22, 2005
Corresponding author:
Taiyo Kobayashi
Institute of Observational Research for Global Change
Japan Agency for Marine-Earth Science and Technology
2-15, Natsushima-cho, Yokosuka-city
Kanagawa, 237-0061, Japan
Tel: +81-46-867-9842
Fax: +81-46-867-9835
E-mail: [email protected]
2
Abstract.
The present study developed a high-quality climatological dataset for the Indian Ocean – the
Indian Ocean HydroBase (IOHB) – from a combined dataset including the World Ocean Database 1998
version 2 (WOD98v2). Methods are similar to those used by previous studies for other oceans. Japanese
data for the IOHB originated from the Japanese datasets MIRC (Marine Information Research Center)
Ocean Dataset 2001 and Far Seas Collection; these datasets contain more Japanese observations than
WOD98v2.
Water mass properties in the IOHB climatology are consistent with previous studies.
Seasonal patterns of properties near the sea surface are well reproduced, and deep-layer properties are
consistent with the Reid-Mantyla climatology that is derived from high-quality observations.
The
isopycnal climatology of the IOHB differs from the World Ocean Atlas 2001 (WOA01) along the fronts
associated with the Antarctic Circumpolar Current (ACC).
The WOA01 shows a warm and saline
intermediate water intrusion from South Africa to the east along the northern edge of the front. Such an
intrusion is absent in IOHB where less saline intermediate water extends continuously northward from
the southern ocean. The WOA01 shows a continuous belt of low potential vorticity along the ACC. This
feature is less distinct in the IOHB climatology and in the Reid-Mantyla climatology. The IOHB consists
of a 1° × 1° gridded climatology and the datasets of raw and quality-controlled hydrographic stations. The
latter is valuable for quality control of the Argo float salinity data as climatological reference. These
datasets are available freely via the Internet.
Key words: Indian Ocean; climatological dataset, water mass, circulation, quality control
3
Contents
1
Introduction
2
Source datasets
3
Quality control procedures
3.1
Preparation of the source datasets
3.1.1 World Ocean Database 1998 version 2
3.1.2 MIRC (Marine Information Research Center) Ocean Dataset 2001
3.1.3 Far Seas Collection
3.2
Removal of duplicates
3.3
Removal of suspicious data by visual inspection
3.3.1 Suspicious CTD stations
3.3.2 Suspicious stations with only sea surface data
4
3.4
Removal of suspicious nutrient data
3.5
Statistical checks
3.6
Visual checks
Quality control results
4.1
Horizontal data distributions
4.2
Seasonal data distributions
4.3
Vertical data distributions
4.4
Temporal distribution of hydrographic data
5
Characteristics in the near-surface layer at the peaks of the Monsoon
6
Characteristics on isopycnal surfaces
6.1
Subantarctic Mode Water (26.7-σ0)
6.2
Intermediate Waters (27.3-σ0)
4
6.3
Deep waters (36.92-σ2)
6.4
Bottom waters (45.90-σ4)
7
Comparison with WOA
8
An application of the IOHB to quality control of Argo float salinity
9
Summary
Acknowledgements
References
5
1. Introduction
Efforts to construct gridded climatological datasets from hydrographic data have been ongoing
since 1980. Undoubtedly, the most famous climatological atlases are the series of World Ocean Atlases
(WOAs; Levitus, 1982; National Oceanographic Data Center [NODC], 1994; Conkright, Levitus, O'Brien,
Boyer, Antonov, & Stephens, 1998; Conkright, Locarnini, Garcia, O’Brien, Boyer, Stephens, et al., 2002),
which are produced from a huge number of historical data with objective interpolation on depth surfaces.
They are widely used: in climate and ocean modeling studies: as an “observed” ocean for comparison with
model results and as restoring reference. Recently, a Special Analysis Center (SAC) in the World Ocean
Circulation Experiment Hydrographic Program (WHP) produced another global ocean climatology
(Gouretski & Jancke, 1998, 1999).
In addition to these global climatologies, “HydroBase” climatological datasets have been
produced for the North Atlantic (Lozier, Owens, & Curry, 1995) and the North Pacific (Macdonald, Suga,
& Curry, 2001). HydroBase climatologies have several advantages over the climatologies mentioned
above. All hydrographic data are carefully quality checked by visual inspection at least once; visual
checking screens suspicious data more strictly than automatic quality control procedures used in editing
WOAs (e.g., Conkright et al., 2002) and SAC climatologies (e.g., Gouretski & Jancke, 1999). HydroBase
climatologies are derived from averages along isopycnal surfaces. This averaging strategy precludes the
formation of artificial water masses near frontal regions (Lozier et al., 1995; Macdonald et al., 2001). Note
that the SAC climatology has also been constructed from averages along neutral surfaces. In addition to
the gridded climatology the HydroBase includes the raw and quality-controlled station data. These are
available for water mass analyses. Furthermore, custom-made gridded climatologies can be built from
any subset of the station data with attached tools to produce, for example, a regional climatology on a
mesh smaller than 1° × 1°. Curry, Dickson, & Yashayaev (2003) demonstrated sophisticated features of
the HydroBase in an examination of the decadal-scale change in the freshwater balance of the Atlantic
6
Ocean.
The present study describes a HydroBase climatology for the Indian Ocean: the Indian Ocean
HydroBase (IOHB).
The Indian Ocean is characterized by large seasonal variations driven by the
Monsoon winds (Schott & McCreary, 2001). Various researchers have also identified significant
interannual variability in the Indian Ocean, such as Indian Ocean Dipole Mode (Webster, Moore,
Loschnigg, & Leben, 1999; Saji, Goswami, Vinayachandran, & Yamagata, 1999).
For the global
thermohaline circulation the Indian Ocean is a terminus for deep water (Mantyla & Reid, 1995), but also
carries the return path of surface waters, part of the Warm Water Route (Gordon, 1986). The IOHB, with
its gridded climatology and individual station data, is available for general ocean studies related to the
above mentioned features. Its quality-controlled salinity data enables a detailed examination of water
masses such as Antarctic Intermediate Water (AAIW) and North Atlantic Deep Water (NADW). Also the
seasonal and interannual changes of freshwater/salt budgets and associated transports and fluxes can be
clarified by the IOHB more closely than the preceding study by Rao & Sivakumar (2003).
Many hydrographic variables are autonomously measured by profiling floats deployed in the
world oceans under the Argo Project (The Argo Science Team, 2001). Because sensors drift in time it is
important to evaluate the measurement quality in profiling float observations. Several such methods
have been developed (Feng & Wijffels, 2001; Kobayashi, Ichikawa, Takatsuki, Suga, Iwasaka, Ando, et al.,
2002; Wong, Johnson, & Owens, 2003), all using comparisons with local climatological datasets. The
quality control depends crucially on a reference dataset (Kobayashi & Minato, 2005a) and float
observations can therefore benefit from sophisticated datasets such as HydroBase. At present a reference
dataset derived from the IOHB is used in the quality control operation of the Argo float data at the Japan
Agency for Marine–Earth Science and Technology (JAMSTEC) from January 2005.
The purpose of this manuscript is twofold: to describe fundamental features of the IOHB,
especially details in the quality control procedure and characteristics of the quality-controlled
7
hydrographic data.
And, to present climatological properties on isopycnal surfaces and to identify
differences between the IOHB and the WOA climatologies. The emphasis for the latter is to demonstrate
some of superior features of IOHB rather than to present new knowledge for the Indian Ocean. Section 2
describes the source datasets of IOHB. Details of the quality control procedure are explained in section 3.
Section 4 presents IOHB statistics such as horizontal data distributions.
Seasonal near-surface
properties affected by the Monsoon are presented in section 5. Section 6 includes IOHB isopycnal
climatologies for levels below the thermocline. Section 7 discusses the differences between the IOHB and
WOA to clarify IOHB characteristics. In section 8 some results from quality control of the Argo float data
are introduced as an application of the IOHB. Finally section 9 contains conclusions.
2. Source datasets
Generally, the value of a climatological dataset depends on the amount and quality of individual
data. The IOHB is enhanced by inclusion of as many hydrographic data as possible. The largest
hydrographic dataset is the latest version of the World Ocean Database (WOD) edited by NODC.
However, the Marine Information Research Center (MIRC) in Japan reported that many of the Japan
Oceanographic Data Center (JODC) data have not yet been reported to NODC (MIRC, 2000; pers. comm.).
In addition, the National Research Institute of Far Seas (NRIFS) in Japan has edited another dataset
that also includes data not yet reported to JODC or NODC (K. Mizuno, 2001; pers. comm.).
The IOHB incorporates hydrographic data within the Indian Ocean between 20–150°E and the
Red Sea, the Persian Gulf, and the Indonesian seas south of the Equator (Figure 1). It was produced from
the following four datasets.
(1) World Ocean Database 1998 version 2 (WOD98v2; Conkright, Levitus, O'Brien, Boyer, Stephens,
Johnson, et al., 1999)
(2) MIRC Ocean Dataset 2001 (MODS2001; MIRC, 2001)
8
(3) Far Seas Collection (FSC; NRIFS, 1999)
(4) WHP-CTD data (Diggs, Kappa, Kinkade, & Swift, 2002)
The WOD98v2 dataset, the latest in the WOD series (as of June 2001), was published in January
2000. Hydrographic data in WOD98v2 are separately stored in two categories: bottle sampling and
conductivity–temperature–depth (CTD) profiles. A few lower-resolution CTD stations are included in the
former category. Quality control for the higher-resolution CTD data will be completed at a later date and
they are not used in the subsequent study. The statistical quality control of HydroBase (Curry, 1996) is
designed for profiles with low vertical resolution such as those derived from bottle sampling.
The MODS2001 dataset compiled by MIRC is composed of all hydrographic data reported to
JODC. MODS2001 stores more Japanese data than the WOD98v2 dataset does (MIRC, 2000; pers.
comm.).
The FSC consists of Japanese hydrographic data obtained after 1966. Most of these were
recorded on training ships of Japanese fisheries high schools.
Some of these data have been
quality-controlled (Mizuno, 1995). The FSC data include temperature profiles extending to depths of
500–1000 m; salinity data are limited to sea surface measurements before 1988. Some post-1989 data
include profiles with both, temperature and salinity.
The WHP-CTD data include all CTD profiles obtained by WHP (Diggs et al., 2002). Quality
control procedures for WHP-CTD data differ from the other datasets: all suspicious data (including
oxygen) are removed by visual inspection.
3 Quality control procedures
Quality control for the IOHB is similar to that used by Lozier et al. (1995) and Macdonald et al.
(2001). As described below, source datasets (WOD98v2, MODS2001, and FSC) are prepared (section 3.1)
and then combined into one raw dataset of the IOHB after removing duplicate stations (section 3.2).
9
Suspicious data that may distort water mass statistics are removed by visual inspection (section 3.3).
Statistical checks based on local water mass analysis (section 3.5) are then implemented. A visual check
then screens out suspicious data that survived previous checks (section 3.6). In addition, suspicious
nutrient data are discarded by visual inspection before the statistical checks (section 3.4). Figure 2
summarizes the quality control procedures.
3.1 Preparation of source datasets
Data that fail quality checks in source datasets (described later) are removed. The remaining
temperature/salinity profiles are converted to HydroBase format because density is necessary for
subsequent quality controls (Lozier et al., 1995; Curry, 1996). All dissolved oxygen data before 1960 are
removed because measuring methods changed in the mid-1950s. Bottom depths at stations are derived
from world ocean elevation data ETOPO5; stations with negative sea-depth are likely in the wrong
location and are thus discarded. Stations at depths shallower than 200 m are defined as near the coast
where water mass distributions are more complex; quality control for such stations will be done later.
Additionally, the following procedures are applied to each source dataset, considering its characteristics.
3.1.1 WOD98v2
Data that fail yearly, seasonal, and monthly statistical checks are discarded based on NODC
quality flags. There are 292 new ship codes that are not defined in the NODC list. The new ship codes use
characters (#, @, %, =, -, etc.) that are not usually used. Japanese data are identified and do not proceed to
quality control checks.
3.1.2 MODS2001
Non-Japanese data are discarded from the dataset. Suspicious data marked by quality flags
10
from MIRC are excluded. Such data either failed a statistical check or contained two or more erroneous
observations in one profile. Data from MODS2001 have four-digit ship codes because of perfect ship
identification by MIRC (MIRC, 2001).
3.1.3 FSC
Old data with very low salinity (< 25) are removed.
At such stations, chlorine might be
mistakenly recorded as salinity. Stations including two or more casts are discarded because the header
information including measurement location and date cannot be assigned to the profiles. Data from the
FSC have three-digit ship codes. Cruise and station codes can be retrieved from the data management
code, because the codes are recorded in free format in the original data. Data can thus be traced back to
the original observations.
3.2 Removal of duplicates
Duplicated stations must be discarded before quality controls are applied to the combined raw
dataset (Conkright et al., 1999; MIRC, 2001). The following criteria are used to identify duplicate stations
during IOHB production.
・ Station locations are within 0.02° (1') in latitude and longitude.
・ Observation times are similar within one day.
The possible confusion between local time and
Universal Time is considered.
・ Station data are similar: within 0.01 for temperature and salinity and within 0.05 for dissolved
oxygen and nutrients.
If duplicated stations are found, the station with more observations and more variables is retained within
the dataset. First this procedure is applied to each raw dataset. Then an inter-dataset comparison
between MODS2001 and FSC is done. In the self-duplication check, 971 stations were removed from
11
WOD98v2, 64 from MODS2001 and 2096 from FSC. In the second stage 2185 stations contained in both,
MODS2001 and FSC, were discarded.
At the end of this step, one combined raw dataset is prepared (Figure 3a, Table 1).
3.3 Removal of suspicious data by visual inspection
Suspicious data that may distort the quality control statistics were removed by visual inspection
prior to the statistical check. Most of the removed data were found in groups measured during particular
cruises and which are inconsistent with the main data cluster. These could be classified into the following
two categories.
•
The water mass structure is quite different from that in near-by stations, probably due to errors in
station location (red stations in Figure 4).
•
Profiles have a large bias, especially in salinity (blue stations in Figure 4).
In these cases we discarded the whole profile from the dataset. All together 3636 stations were removed;
containing about 59,000 temperature-salinity pairs and about 15,000 oxygen data points. The following
sections describe additional station removal to prevent possible distortion of water mass statistics.
3.3.1 Suspicious CTD stations
The statistical check in HydroBase deals with each observation of temperature and salinity with
even weight (Curry, 1996). Because CTD data have a high vertical resolution, they have a large influence
on the statistics of water mass structures, even if CTD stations are only in the minority (Figure 5). Thus,
CTD data are checked by stricter criteria than bottle sampling data. At this stage, 233 CTD stations were
discarded from the dataset.
3.3.2 Suspicious stations with only sea surface data
12
The MODS2001 and FSC datasets include more than 20,000 stations with only sea surface
measurements; these stations are concentrated in the eastern Indian Ocean. Variability in these sea
surface data is much larger than the variability in data profile (Figure 6); data quality is inferior, probably
because they were taken from bucket samples (K. Mizuno, 2001; pers. comm.). More than 1300 sea
surface stations with values more than 2.5 standard deviations away from the mean of the 5°× 5° region
were discarded.
3.4 Removal of suspicious nutrient data
Some nutrient data have structures quite different from those at nearby stations (Figure 7).
Such suspicious nutrient data are removed by visual inspection before the statistical checks. Suspicious
data are found in all nutrient observations, especially in nitrates observed by ships of the Union of Soviet
Socialist Republics (USSR). A total of 799 profiles for phosphate, 674 for silicate, 428 for nitrate, and 159
for nitrite were removed in this procedure.
3.5 The statistical check
A statistical check, which discards data outside a defined range of the local climatology, is the
core of the quality control for the IOHB. For this check, data are divided into about 300 sub-regions that
consider the spatial variation of water mass structures (θ-S and θ-O2 relationships). The sub-regions
were chosen to obtain relationships as tight as possible while retaining sufficient numbers of data for the
statistics. In the mid-ocean, the quality control sub-regions cover larger areas compared to coastal
regions, where the increased number of observations allows smaller grids to be used (Figure 8). Data
distribution is also sparse in deeper layers, and observed spatial variability in the water mass is reduced.
Larger geographic bins are used for quality control in layers below 1000 m. In general the IOHB
geographic bins cover larger areas than other HydroBase (Lozier et al., 1995; Macdonald et al., 2001),
13
reflecting the sparse observations in the Indian Ocean.
Data in each sub-region are projected in the vertical onto isopycnal surfaces. The θ-S and θ-O2
relationships in the sub-regions are defined by connecting the property averages calculated in each
vertical bin (Figure 9).
Standard deviations of salinity as a function of potential temperature are
calculated for each vertical bin; θ-S data that fall outside a 2.3 standard deviation criterion are removed
from the dataset. Entire profiles are discarded if more than half of the observations in the profile are
removed. A similar check is done for the dissolved oxygen, but here a 3 standard deviation criterion is
applied. The above statistical check is applied twice in each sub-region. Data removed have no apparent
geographic bias, which would reflect unsuitable sub-regions. At this stage, 1917 stations were discarded;
about 40,000 temperature–salinity pairs and 21,000 oxygen data points were removed.
3.6 The visual check
The final quality control check is a visual scanning of property distributions on several isopycnal
surfaces. This check visually and subjectively screens out suspicious data that survive the above quality
control procedures. The check was performed on non-smoothed 1° × 1° isopycnal maps of averages and
standard deviations calculated from the data located within each grid. Pressure, potential temperature,
salinity, oxygen, and potential vorticity (PV) are visually checked on 18 isopycnal surfaces. In most cases,
such suspicious data were accompanied by “bull’s-eyes” on several properties on an isopycnal surface.
Visual scanning removed 45 stations and 12 oxygen profiles. When doubtful data are removed, the entire
statistics check is re-run from the beginning.
After the whole procedure for quality control was carried out, the IOHB dataset for the
individual station data was completed (Figure 3b).
4. Quality control results
14
4.1 Horizontal data distributions
Figure 10 shows the distribution of the quality-controlled stations. As in other oceans, many
observations are near the coast, in particularly west of Australia, east of Africa, and in the Arabian Sea.
There are few observations far from the continents, especially in the Antarctic Circumpolar Current
(ACC).
Most of the observations in the Indian Ocean were taken from ships from Japan (originated from
MODS2001 and the FSC), the USSR (NODC country code 90), or “unknown” countries (NODC country
code 99). Most of the Japanese data were taken by fisheries training ships; data are concentrated in the
eastern Indian Ocean, which is a fishing ground for tuna and squid (K. Mizuno, 2001; pers. comm.). There
are more than 20,000 Japanese stations (Table 2), but only 2250 of these include subsurface observations.
Most of the Japanese training ships for fisheries measured salinity only at the sea surface (section 2).
Ships from the USSR collected the largest part of hydrographic data in the Indian Ocean. About
25% of all stations in the IOHB (Table 2) were Soviet observations. When sea surface stations are
excluded, the percentage increases to 34%. Soviet measurements are distributed widely in the Indian
Ocean. Most of the hydrographic data around the ACC are obtained by ships from the USSR.
About 14% of the stations in the IOHB are of unknown nationality. That percentage increases to
about 19% when sea-surface-measurements-only stations are excluded.
Most of the stations from
unknown countries are in the western Indian Ocean, especially in the Arabian Sea. Worldwide, the
United States is a major data contributor. In the Indian Ocean, however, the United States supplies only
about 8% of all the station data. The relative contribution of hydrographic data from the USSR and
unknown countries is much higher in the Indian Ocean (sea Table 2) than in the WOD98 statistics
(Levitus, Boyer, Conkright, O'Brien, Antonov, Stephens, et al., 1998).
An unusual feature in the Indian Ocean is that there are more than 20 locations with larger data
concentration. For example at one location, more than 100 profiles were taken in three consecutive
15
months. Those resulted mostly from work by the USSR or unknown countries.
Figure 11 shows the distributions of stations containing nutrient data. There are 22,143 oxygen,
17,419 phosphate, 13,801 silicate, 6231 nitrate, and 4567 nitrite stations. The nutrient distributions
resemble those of temperature and salinity, although nitrate stations are very sparse around the ACC
except south of Australia. All stations from unknown countries have no oxygen/nutrients data, and
stations from the FSC have no nutrient data except for one cruise with oxygen data.
Figure 12 shows the distribution of the data removal ratio during the entire quality control.
Over most areas, less than 5% of the stations or measurement layers are removed. Geographical bias for
quality control is not large. More than 20% of stations/layers are removed in several regions; suspicious
CTD data are removed (see section 3.3.1) in the Bay of Bengal and east of Madagascar, and questionable
stations with probably erroneous location record (see section 3.3) are discarded in the central Indian
Ocean.
4.2 Seasonal data distribution
Observations taken in the ocean are affected by sea state. The number of observations generally
decreases in winter due to bad weather (e.g., Figure 14 in Macdonald et al. 2001). In the IOHB dataset,
the number of observations maximizes in February and is somewhat lower between September and
December (Figure 13a). In contrast, the seasonal changes of the spatial extent of hydrographic data are
relatively small (Figure 13b); observations in February and September have almost no geographical bias
except around Antarctica (Figure 14). The IOHB does have sufficient data to describe bi-monthly features
in the northern part of the Indian Ocean, especially in the near-surface layers (see section 5).
4.3 Vertical data distribution
Figure 15 shows the number of stations as a function of the depth bins of observation layers. In
16
WOD98v2 there are about 35,000 stations in the surface layer above 100 m. Station numbers decrease
strongly below 500 m and then gradually with depth. There are fewer observations at 125 m, 700 m and
900 m than just above and below those levels. In contrast, there are more measurements at 1000 m and
1500 m than in adjacent layers. Macdonald et al. (2001) showed a similar station distribution. About 20%
of the stations (excluding sea-surface-measurement-only stations) have observations below 1000 m. In
contrast, about 40% of stations in the North Pacific have observations below 1000 m (Macdonald et al.,
2001). There are many sea surface observations in MODS2001 and the FSC (see section 2). Except for
this feature, the vertical structure of the data distribution is similar to that of WOD98v2.
4.4 Temporal distribution of hydrographic data
Figure 16 shows the temporal distributions of hydrographic data in the IOHB. Systematic
studies of the Indian Ocean began with the International Indian Ocean Expedition (IIOE) in 1962–1965
(Wyrtki, 1971). Indian Ocean observations before the IIOE were very scarce. Most were restricted to
coastal regions: the Indonesian seas by the Netherlands (1929-30), Japan (1941–44), and around the
Mozambique Channel in the 1950s. Observations increased starting in the late 1950s, and 1000–3000
observations were taken each year after 1965. Intensive observations around the ACC began in the 1970s.
The number of Japanese observations increased rapidly at the end of the 1960s after which time
500–1500 profiles have been obtained each year. However, the Japanese contribution to the number of
observation layers is much smaller than that of WOD98v2, because many Japanese stations recorded only
sea surface data.
5. Characteristics in the near-surface layer at the peaks of Monsoon (Figure 17)
Current systems in the Indian Ocean change seasonally because of monsoonal forcing. Since the
IIOE, scientific efforts to understand this seasonal change have focused on several viewpoints, as
17
reviewed in Schott & McCreary (2001). This section shows that the IOHB reproduces features of the
seasonally changing surface layers.
Figure 17 shows surface layer features at the Northeast (January/February) and Southwest
(July/August) Monsoon peaks. Large seasonal variations related to the reversal of the Somali Current
system occur in the Arabian Sea. At the peak of the Southwest Monsoon, the Somali Current flows
northeastward along the coasts of Somalia and the Arabian Peninsula. The current is accompanied by a
shoaling of isopycnal surfaces (Figure 17h) caused by coastal upwelling (Schott, 1983); water cooler than
25°C reaches the sea surface (Rao & Sivakumar, 2000). During this season saline waters from the
Arabian Sea are carried toward the Bay of Bengal (Rao & Sivakumar, 2003).
A domed structure appears in the pycnocline during the Northeast Monsoon in the Arabian Sea
(Figure 17g). The structure extends southwestward and links with the permanent isopycnal dome around
5–10°S that is related to Ekman divergence at the sea surface (Schott & McCreary, 2001). The 24.5-σ0
surface outcrops in the northern Arabian Sea due to winter cooling; saline water spreads from the local
vertical maximum of salinity (Morrison, 1997). Except for this region, the sea surface temperature
exceeds 27°C (Figure 17a).
Relatively fresh water is found in the Bay of Bengal at the sea surface (Figures 17c and d)
because of river discharge and excess precipitation. These low salinity waters extend southeastward
along the Indonesian coast (Shenoi, Saji, & Almeida, 1999). During the Northeast Monsoon, the less
saline water extends further west to around 10°S. Some of the fresher water in the Bay of Bengal is
drawn into the Arabian Sea by the seasonal coastal current (Shetye, Gouveia, Shankar, Shenoi,
Vinayachandran, Sundar, et al., 1996).
Pycnocline properties are affected by warm and saline water in the Arabian Sea that extends
southward and then flows eastward along the Equator (Figures 17k and l). During boreal winter, part of
this saline water extends further south and can be traced along the African coast to north of Madagascar.
18
During boreal summer, in contrast, colder and fresher water from the eastern Indian Ocean intrudes
northward along the coast.
The IOHB can contribute to studies on seasonal variations in the Indian Ocean and inter-annual
modulations of them. Seasonal features of the IOHB are consistent with the results of previous studies
based on synoptic observations (Molinari, Olson, & Reverdin, 1990; Shenoi et al., 1999; Rao & Sivakumar,
2000, 2003) and agree with numerical model results (e.g., McCreary, Kundu, & Molinari, 1993). Peter &
Mizuno (2000) produced maps of dynamic height that were referenced to 400 dbar. The present results
(Figures 17e and f) resemble those except for the Arabian Sea during the Southwest Monsoon. Peter &
Mizuno (2000) showed an eastward flow along 15°N and a coastal return flow. Differences may be due to
the sparse observations and the shallow reference depth in their study. However, even a reference depth
of 1000 dbar in the present study does not reproduce the seasonal current systems completely: the
Monsoon influences the deep layer below 2000 dbar (Warren & Johnson, 1992). The 1000 dbar reference
depth is chosen in this study because of a lack of seasonal dynamic height data referred to 2000 dbar; this
lack of data makes it difficult to determine flow patterns. There are not enough deep profiles to construct
bimonthly maps with simple averaging.
6. Characteristics on isopycnal surfaces
This section demonstrates that the isopycnal climatology of the IOHB below the thermocline is
consistent with the features described in previous studies.
Figures 18–21 show the horizontal
distributions (on a 1°×1° grid) of water properties on the 26.7-σ0, 27.3-σ0, 36.92-σ2, and 45.90-σ4 isopycnal
surfaces, respectively. The surfaces describe typical water masses in the Indian Ocean; densities were
selected to be comparable with maps in previous studies (Mantyla & Reid, 1995; McCarthy & Talley, 1999;
Reid, 2003).
In all figures, we show features of the IOHB climatology using weighted averages
(smoothing); smoothing factors are Gaussian functions with e-folding scales of 2.5° zonally and 1.5°
19
meridionally. Such smoothing can also extrapolate into regions of sparse data; blank panels are filled in if
the sum of the weighting factors for calculating statistics is less than 3.
Property distributions on layers shallower than 2000 dbar may vary seasonally because of
monsoonal influences, especially in the North Indian Ocean (Mantyla & Reid, 1995; Schott & McCreary,
2001). Here we do not consider seasonal variability and show mean climatological features on 26.7-σ0 and
27.3-σ0. Thus, the total variability (standard deviations) of isopycnal water properties includes the
seasonal variability.
6.1 Subantarctic Mode Water (26.7-σ0; Figure 18)
The thick pycnostad at the southern edge of the subtropical gyre in the Indian Ocean
(McCartney, 1977) is known as Subantarctic Mode Water (SAMW). McCartney (1982) concluded that
SAMW in the Indian Ocean has two separate cores at 26.7-σ0 and 26.85-σ0.
Figure 18 shows the
horizontal distribution of the lighter SAMW. On this isopycnal surface a strong pressure gradient is
found along 40-45°S with variability exceeding 200 dbar. Hereafter, such this pressure gradient is simply
refereed to as “pressure front”, corresponding to a density front on depth surfaces. This sharp pressure
front is associated with the strong eastward South Indian Ocean Current (Stramma, 1992; Stramma &
Lutjeharms, 1997). The front is narrower with larger variability west of the Kerguelen Plateau. East of
the plateau, it is less distinct. Temperature and salinity fronts spread slightly further south of the
pressure front; their variabilities are also larger to the west than to the east.
The subtropical gyre is located north of the pressure front. Around 30–40°S the western part of
the isopycnal surface suddenly deepens around 60°E and 80°E. This is consistent with Stramma &
Lutjeharms (1997); a considerable part of the South Indian Ocean Current recirculates into the
subtropical gyre around 60°E and 80°E. This also shows up in the stream function pattern in the IOHB.
The low-PV water of the SAMW is widely distributed in the subtropical gyre, with properties of 11–12°C
20
in temperature, 35–35.2 in salinity, and 5–6 ml/l in oxygen. The youngest SAMW with the lowest PV and
highest oxygen is found at 80–110°E just north of the pressure front. The large pressure variability of the
isopycnal surface also occurs in this region, possibly reflecting winter outcropping at the surface. The
lighter SAMW is probably formed there and is then subducted northwestward.
Colder, fresher, and oxygen-depleted water north of the subtropical gyre extends westward along
10°S from the Indonesian seas. The water meets SAMW around 10–20°S where a distinct water mass
front forms. This water has high PV, stretching to the northwestern coasts of Australia. The high PV
water separates SAMW from water near the Equator.
Most water from the south returns southward along the coasts of Madagascar and South Africa,
but some flows north of Madagascar to the Equator. After that, oxygen distributions suggest that part of
this water flows eastward along the Equator. The isopycnal surface north of 10°S is almost flat; averaging
operations that disregard seasons may diminish the flow patterns. There are weak variations in several
properties in the Arabian Sea that probably signal seasonal changes in the current system. The Arabian
Sea is filled with warm, saline, and oxygen-poor water, which is formed by the excess evaporation and
enhanced biological activity (Olson, Hitchcock, Fine, & Warren, 1993). Features on the 26.85-σ0 surface
resemble those in Figure 18 except that denser SAMW is formed around 120°E.
6.2 Intermediate Waters (27.3-σ0; Figure 19)
This layer is located in the core layer of two major intermediate waters: the AAIW and Red Sea
Water (RSW). The AAIW is a salinity minimum from the southern ocean (Piola & Georgi, 1982; Fine,
1993), and the RSW is saline water extending from the Arabian Sea (Beal, Ffield, & Gordon, 2000) with
origins in the Red Sea (Murray & Johns, 1997).
The isopycnal surface deepens from 200 dbar to 800–1000 dbar between 55 and 40°S. This
pressure front supports the strong eastward ACC. Around 20–50°E another strong pressure front is seen
21
near 40°S, corresponding to the Agulhas Front (Belkin & Gordon, 1996). Pressure variability here
exceeds 200 dbar. The two fronts converge north of the Crozet Plateau, and it continues to north of the
Kerguelen Plateau. The united front then diverges to the east of that plateau. South of Africa, low PV
water spreads north of the pressure front along the ACC and stretches to the south of Australia along the
northern flank of the front. This band of low PV water is also present in results from McCarthy & Talley
(1999). South of the pressure front, the water mass is colder, fresher, and oxygen-richer. Its large
variability is probably due to seasonal outcropping.
The subtropical gyre is seen as a bowl-shaped structure north of the fronts in the South Indian
Ocean. Cold, fresh, and oxygen-rich AAIW extends northward from south of the ACC, filling the entire
southern subtropical gyre. A subtropical trough extends eastward to south of Australia. Westward
intermediate flow south of Australia is also suggested by Reid (2003) and some numerical studies
(Semtner & Chervin, 1992); such flow may connect the South Indian Ocean and South Pacific.
Northward-spreading AAIW intersects warmer, saltier and oxygen-poorer RSW. Water property
differences support a weak water mass front north of the subtropical gyre. This feature is especially
distinct in the oxygen distribution. The most saline RSW is found in the Arabian Sea. Strong variations
in temperature and salinity occur in the Gulf of Aden. These variations are probably caused by seasonal
variability in RSW spreading (Gamsakhurdiya, Meshchanov, & Shapiro, 1991; Murray & Johns, 1997).
Some of the RSW extends eastward to the Bay of Bengal. Oxygen in the Arabian Sea and the Bay of
Bengal is locally consumed by high primary productivity heightened by additional organic matter in river
discharges (Olson et al., 1993).
Some RSW moves southward along the African coast through the Mozambique Channel and
then reaches the southern tip of Africa (Gordon, Lutjeharms, & Gründlingh, 1987). Beal et al. (2000)
discussed its southward coastal flow based on the PV distribution. Similar features can be found in the
IOHB: the meridional gradient in PV field is smaller in the Mozambique Channel compared to east of
22
Madagascar.
6.3 Deep waters (36.92-σ2; Figure 20)
This isopycnal surface corresponds to the layer just above the salinity maximum core of saline
water from the NADW (Reid, 2003). A strong rise of the isopycnal surface is found near 45–60°S, pressure
values change from 500 to 2200 dbar. The pressure front corresponds to the deeper part of ACC. The
structure of pressure front is complex, and large variations are found in several zonal bands. The front on
this isopycnal is influenced by bottom topography. Just north of the front, a weak trough representing the
deeper part of the South Indian subtropical gyre is present. North of this trough, the isopycnal surface is
flat around 2000dbar extending into the Indian Ocean. The flatness suggests that only weak currents
exist (Mantyla & Reid, 1995), so stream functions on the 26.7-σ0 and 27.3-σ0 surfaces are referenced to
2000 dbar in this study.
The youngest (the most saline and oxygen-rich) NADW in the Indian Ocean is found south of
Africa in the Agulhas Basin, which is filled directly from the Atlantic. This young NADW continues
northward and spreads into the Mozambique Basin. It then flows eastward along the northern flank of
the Southeast Indian Ridge; its high salinity can be traced to south of Australia. Some of the saline water
intrudes into the Enderby Basin at the west of the Kerguelen Plateau. Some NADW may also flow into
the Central Indian Basin along the eastern flank of the Central Indian Ridge. This feature is more
distinct in the PV distribution; water with higher PV extends northwestward along the Central Indian
Ridge. Mantyla & Reid (1995) reported that a narrow extension of the NADW high salinity water extends
to the east of Madagascar. This feature is indistinct in this study, however, because of the lower resolution
of the climatology.
Another saline water mass is located in the Arabian Sea and the Bay of Bengal. This water has
also been identified in previous studies (Reid, 1981, 2003; Mantyla & Reid, 1995). Mantyla & Reid (1995)
23
suggested that the salty water forms through mixture with saltier water above. Temperature and salinity
variability in this layer is large.
Water properties on this isopycnal change south of the ACC where cold, fresh, and oxygen-rich
water occurs. Very cold water (below –1°C) spreads south of the water mass front along 60–65°S.
6.4 Bottom waters (45.90-σ4; Figure 21)
This isopycnal represents the densest layer on which water from the south can reach the
northernmost of the Indian Ocean (Mantyla & Reid, 1995). The surface is at 3000 dbar or deeper north of
45°S and northward intrusions of the southern water are strongly guided by bottom topography. The
source water with the highest salinity and richest oxygen is in the Agulhas Basin, which is filled directly
from the Atlantic. The saline water overflows to the east as the deep part of the ACC and its signal
gradually decreases. It has two branches extending northward; the western branch continues along the
eastern coasts of Madagascar and Africa to the Somali Basin. The water in the eastern branch flows into
the Central Indian Basin along the eastern flank of the Central Indian Ridge. Similar northward
extensions of high PV water are also found, although a lack of data obscures them.
The pressure front on this isopycnal surface that corresponds to the deepest part of the ACC is
visible around 45–55°S. Another pressure front extends to the near surface around Antarctica, where it
merges with a water mass front. Very cold, fresh, and oxygen-rich waters occur south of the water mass
front.
7. Comparison with the WOA
This section compares the IOHB with the World Ocean Atlas 2001 (WOA01, Conkright et al.,
2002). By examining the differences between these climatologies, the characteristics of each become
distinct.
24
Figures 22–24 show distributions of water properties from the WOA01 and differences from the
IOHB on the 26.7-σ0, 27.3-σ0, and 36.92-σ2 isopycnal surfaces, respectively. According to previous work by
Lozier et al. (1995) and Macdonald et al. (2001), it is expected that the two climatologies show large
differences around the pressure front corresponding to the ACC. This is indeed the case. Positive
differences of pressure exceeding 150 dbar are found near 40–80°E on the 26.7-σ0 surface (Figure 22). The
pressure front in the IOHB is shifted slightly southward relative to that in the WOA01. Negative
differences along the South Africa coast represent the stronger frontal structure of the Agulhas Current
in the IOHB. In Figure 23 the WOA01 reproduces the ACC on the 27.3-σ0 surface as a gentle pressure
front around 40–55°S probably because of strong smoothing. Thus the pressure differences show
negative/positive patches along the southern/northern side of the front. A negative patch is located near
40°S, 20–50°E, because of the less distinct double frontal structure in the WOA01.
Large differences between both climatologies also occur in the water mass distributions. The
WOA01 includes a strong extension of warm and saline water from South Africa on the 27.3-σ0 surface
(Figure 23); the water of the Agulhas Extension intrudes into the colder and less saline water along the
northern side of the ACC. A similar feature occurs on the 26.7-σ0 surface although its signal is fainter
(Figure 22). In contrast, IOHB climatology shows smoother features, gentle gradation of water masses in
this area (see Figure 19). Difference maps clarify the feature: a belt of colder (by more than 1.5°C) and
less saline (by more than 0.25) water extends across the Southern Ocean on the 27.3-σ0 surface (Figure 23).
Note that the meridional smoothing scale for the IOHB isopycnal climatology is 1.5°, much smaller than
the smoothing scale (~1000 km) used for the WOA01. That is, the large differences in the water mass
patterns along the ACC can not be directly attributed to the strength of smoothing.
Features on the 36.92-σ2 surface around the pressure front (Figure 24) differ from those on less
dense layers. Positive and negative patches (exceeding 300 dbar) are found on the pressure difference
map; negative differences dominate along the pressure front. In the distribution of properties on this
25
surface in the WOA01, an eastward NADW extension from the Agulhas Basin seems punctuated at 50°E
where differences exceed 0.1°C and 0.02; we may have a different impression of its distribution from the
IOHB (see Figure 20).
Differences at the pressure front are described on meridional sections along 65°E from the
WOA01 and IOHB (Figure 25). A saline wedge at the pressure front pinches off the subtropical part of the
AAIW in the WOA01. The wedge extends to the 36.7-σ2 surface. The AAIW in the WOA01 has a more
saline core than the AAIW captured by nearby individual observations (Figure 26).
The PV distribution shows another difference along the pressure front. A belt of lower PV is
apparent on the 27.3-σ0 surface in the WOA01 (see Figure 23). The PV belt is also present in the IOHB
(see Figure 19) and was also noted by McCarthy & Talley (1999); however, the WOA01 shows this feature
as much more pronounced. The PV distribution in the WOA01 differs from those in the Reid-Mantyla
dataset (Mantyla & Reid, 1995) and the IOHB. It seems, however, not to be a global feature; systematic
differences are not found in the North Pacific between the WOA (Levitus, 1982; Talley, 1988) and
HydroBase (Macdonald et al., 2001; Suga, Motoki, Aoki, & Macdonald, 2004).
The PV differences around the ACC are caused by changes in pressure (depth) of isopycnal
surfaces (see Figure 25). The isopycnal surfaces of the IOHB (above 27.3-σ0) spread at a greater depth
than those of the WOA01 around the ACC, and larger differences are found in the less dense layers. Thus,
the IOHB has thinner isopycnal layers than the WOA01 around the front, and the PV shows larger values
there.
It is difficult to ascertain the exact causes of such differences on isopycnal surfaces; because both
climatologies are different in many aspects. For example, the temperature climatology of WOA01 is based
on more observations as it includes those without salinity. Its annual mean is produced by averaging
monthly/seasonal mean fields of each property whereas ours is calculated by simple averaging. Among of
them, the most influential factor is considered to be the difference of their averaging (smoothing)
26
strategies, along depth surfaces in WOA01 and on isopycnals for the IOHB. This can be illustrated by a
comparison with another climatology constructed by averaging on depth surfaces from the same IOHB
dataset of the quality-controlled individual stations (Figure 27). The larger smoothing scale of 2.5° used in
both climatologies makes the differences caused by the averaging strategies more prominent. The depth
averaged climatology bears a general resemblance to the WOA01; the property differences around the
ACC front, especially in the less saline wedge, and the deepening isopycnal surfaces are almost the same
as those between the IOHB and the WOA01 (see Figure 25). That is, it is strongly suggested that basic
features of the differences between the IOHB and WOA01 can be attributed to their averaging
(smoothing) strategies, rather than to their differences in editing procedures, source data, and so on.
The previous study of Suga et al. (2004) explained that the WOA smoothing operation on depth
levels over water mass fronts results in artificially denser water due to the nonlinear nature of the
equation of state. As a result, isopycnal surfaces reproduced in WOA spread at shallower depths at such
fronts as the ACC front. The process is more effective when the water mass front has larger gradients in
temperature and salinity; generally such gradients are larger at shallower depth than at deeper depth.
This scenario can explain the differences between the IOHB and the WOA01 in pressure field for the
layers above 27.3-σ0 at the ACC very well.
The causes of the negative differences between the climatologies below 27.3-σ0 in pressure field
are not clear: we think that they are attributed to something artificial, which are probably produced by
interaction between smoothing operation and bottom topography (the Kerguelen Plateau, see Figure 1). It
is noted that almost similar features are produced in the isodepth averaged climatology (see Figure 27).
Features in SAMW also show large differences between the IOHB and WOA01 (see Figure 22).
SAMW in the IOHB has lower PV and seems to extend further north than SAMW in WOA01 (see Figure
18); in contrast, the water north of SAMW has much higher PV values in WOA01 and has a wider lateral
extent. The meridional view along 65°E (Figure 25) shows that SAMW in the WOA01 has a core at
27
26.5–26.6-σ0, which is less dense than the counterpart in the IOHB. Similar trends are found in other
meridional sections. Causes of the differences are not clear. Note that similar changes are found in PV
features of the isodepth averaged climatology of the IOHB (see Figure 27). SAMW in the WOA01 has
much warmer and saltier characteristics west of Australia. The IOHB shows a similar, albeit weaker,
feature. A large region of negative pressure differences occurs where SAMW forms.
On the 36.92-σ2 surface (Figure 24), differences also occur in the Arabian Sea and Bay of Bengal.
These partly exceed 0.2°C and 0.03, but there are only slight variations in the pressure field. Seasonal
variability, averaging procedures, and differences in the quality control may cause these differences.
Regions with large differences coincide well with regions with large standard deviation (see Figure 19).
8. An applications of the IOHB to quality control of Argo float salinity
In this section, we introduce some results of quality control of the Argo salinity data as an
application of the IOHB. Autonomous measurements by Argo floats need systematic evaluation of its data
quality, especially in salinity. Deep-layer climatological datasets are one of the best references for this.
The results of the Argo standard method (Wong et al., 2003) for the delayed-mode quality control (dQC)
critically depends on the reference datasets (Kobayashi & Minato, 2005a). Thus, the sophisticated dataset
of the IOHB is considered as one of key factors in float observations in the Indian Ocean.
The reference dataset for the Argo dQC consists of profiles suitable for application of the
Wong et al. (2003) method as follows: 1) Profiles must contain at least 15 observation levels to assure
adequate interpolation of reference salinity at selected temperatures. 2) Profiles must extend below
1000 dbar to select suitable salinity from profiles having temperature inversions (Kobayashi & Minato,
2005a). The threshold of 1000 dbar is below the depth of the temperature maximum in the South Indian
Ocean. Data from the Indonesian seas and the Red Sea are completely removed due to their different
water mass structures. These data are unsuitable for estimating climatological references with objective
28
mapping.
Figure 28 shows results of the standard Argo dQC with the IOHB reference. The calibrated
salinity values (the gray belts in the right panels) are used as reference to evaluate the float
measurements (the solid circles), and they provide the most preferable correction values, if float salinity
data need to be corrected. It is also important for the Argo dQC that the calibrated salinities agree with
the “true” values such as the CTD measurements at the float deployment (gray circles).
In the tropical and subtropical cases (Figures 28a and b), the optimal values of calibrated
salinities (white line; the center of gray belt) agree with the salinities independently obtained by the
shipboard CTDs, or true values. Also, calibration errors (half width of gray belts) are about 0.01 or less in
both cases. This shows that we can quality-control float salinity with accuracy of 0.01 or better, to achieve
the Argo target for salinity measurement accuracy (±0.01) in these regions.
In the subantarctic Indian Ocean (Figure 28c), the calibration results are inferior to the previous
cases. The calibrated salinity includes shipboard CTD within its errors; however, the deviations of its
optimal values from the “true” value are about 0.02, and the calibration errors are 0.02-0.03. This may
have two reasons: the historical data is sparse, and the vertically uniform structure of water mass makes
the application of the method of Wong et al. (2003) less successful (Kobayashi & Minato, 2005b). That is,
Argo dQC with the reference of the IOHB guarantees the quality of the Argo float salinity data with
accuracy within 0.01 in the tropical/subtropical Indian Ocean and about 0.02 even in the subantarctic
region. In the future the quality of salinity data will be improved by using the better reference dataset
derived from the IOHB.
9. Summary
A high-quality climatological dataset for the Indian Ocean, the Indian Ocean HydroBase
(IOHB), was produced by combining raw data from three existing datasets (WOD98v2, MODS2001, and
29
FSC) using quality control methods similar to those of Lozier et al. (1995) and Macdonald et al. (2001).
Raw datasets in the IOHB include Japanese data taken from the MODS2001 and FSC datasets and
non-Japanese data from the WOD98v2. These datasets include Japanese observations that are not
included in WOD98v2 (MIRC, 2000; pers. comm.; K. Mizuno, 2001; pers. comm.).
Water mass properties in the IOHB climatology are consistent with previous studies.
Seasonal patterns of properties near the sea surface are well reproduced, and property distributions in
the deep layers are consistent with those in the high-quality Reid-Mantyla dataset (Mantyla & Reid,
1995). Comparison of water properties between the IOHB and the WOA01 yields many differences near
the ACC, especially along the pressure front. The WOA01 shows warm saline water intruding into cold
fresh water along the north side of the ACC, so that the saline band seems to separate the subtropical
component of the AAIW from its southern part. This band of warm saline water is not present in the
IOHB, and AAIW extends continuously northward from south of the front. Small amounts of low PV
water is distributed along the northern flank of the ACC in the IOHB and the Reid-Mantyla dataset
(McCarthy & Talley, 1999), but its structure is highly emphasized in the WOA01; this difference is
explained by the systematic change of the isopycnal layer depth at the ACC pressure front.
The
differences between the two climatologies can be mainly attributed to those of their averaging strategies;
the features reproduced in the IOHB seem to be more reasonable than those in the WOA01.
The IOHB combines a gridded climatology with datasets of raw and quality-controlled
hydrographic stations.
All these are freely available from the websites of Argo JAMSTEC
(http://www.jamstec.go.jp/ARGO/J_ARGOe.html), and in the near future from the HydroBase2 official
website (http://www.whoi.edu/science/PO/hydrobase/).
Acknowledgements
We thank members of the Argo group at JAMSTEC for their help and encouragement.
30
Especially, we gratefully acknowledge the quality control help of Mr. H. Nakajima and the comments of Dr.
K. Mizuno on the Japanese observations in the Indian Ocean. Also, we thank to Ms. Y. Ichikawa and Ms.
E. Sugiyama for their programming help. We thank Dr. K. Hasunuma for his hearty encouragement.
Special thanks are also extended to Dr. A. M. Macdonald for her advice on quality control procedures and
to Dr. R. G. Curry for her kind support of publication of the IOHB datasets on the official HydroBase2
website. We also thank two reviewers for their valuable comments. The IOHB includes two sections of
WHP-CTD data (35MF62JADE_1 and 35MF71JADE_1); these data were observed through
Franco–Indonesian cooperation (Principle Investigators are Drs. M. Fieux and A. G. Ilahude). This
publication has used data from the international Argo Project (www.argo.ucsd.edu). Argo is a pilot project
of the Global Ocean and Global Climate Observing System. The data from Argo profiling floats are freely
available. This work was supported by “The ARGO Project – Advanced Ocean Observing System –” as one
of the Millennium Projects of the Japanese Government.
31
References
Beal, L. M., Ffield, A., & Gordon, A. L. (2000). Spreading of Red Sea overflow waters in the Indian Ocean.
Journal of Geophysical Research, 105, 8549-8564.
Belkin, I. M., & Gordon, A. L. (1996). Southern ocean fronts from the Greenwich meridian to Tasmania.
Journal of Geophysical Research, 101, 3675-3696.
Conkright, M. E., Levitus, S., O'Brien, T., Boyer, T., Antonov, J., & Stephens, C. (1998). World Ocean Atlas
1998 CD-ROM Data Set Documentation. National Oceanographic Data Center Internal Report 15, 16
pp., National Oceanographic and Atmospheric Administration, Silver Spring, Md.
Conkright, M. E., Levitus, S., O'Brien, T., Boyer, T. P., Stephens, C., Johnson, D., Baranova, O., Antonov,
J., Gelfeld, R., Rochester, J., & Forgy, C. (1999). World Ocean Database 1998 CD-ROM dataset
documentation version 2.0. National Oceanographic Data Center Internal Report 14, 113 pp., National
Oceanographic and Atmospheric Administration, Silver Spring, Md.
Conkright, M. E., Locarnini, R. A., Garcia, H. E., O’Brien, T. D., Boyer, T. P., Stephens, C., & Antonov, J.
I. (2002). World Ocean Atlas 2001: Objective Analyses, Data Statistics, and Figures, CD-ROM
Documentation. National Oceanographic Data Center, Silver Spring, MD, 17 pp.
Curry, R. (1996). HydroBase – A database of hydrographic stations and tools for climatological analysis.
Woods Hole Oceanographic Institution Technical Report WHOI96-01, 44 pp.
Curry, R., Dickson, B., & Yashayaev, I. (2003). A change in the freshwater balance of the Atlantic Ocean
over the past four decades. Nature, 426, 826-829.
Diggs, S., Kappa, J., Kinkade, D., & Swift, J. (2002). WOCE Version 3.0. Scripps Institution of
Oceanography, University of California San Diego.
Feng, M., & Wijffels, S. (2001). Results from a pilot Argo float program in the southeastern Indian Ocean.
The Scientific and Technical Workshop of the Data Buoy Cooperation Panel, 22-23 October 2001,
Perth, Australia.
32
Fine, R. A. (1993). Circulation of Antarctic Intermediate Water in the South Indian Ocean. Deep-Sea
Research I, 40, 2021-2042.
Gordon, A. L. (1986). Interocean exchange of thermocline water. Journal of Geophysical Research, 91,
5037-5046.
Gordon, A. L., Lutjeharms, J. R. E., & Gründlingh, M. L. (1987). Stratification and circulation at the
Agulhas Retroflection. Deep-Sea Research, 34, 565-599.
Gouretski, V. V., & Jancke, K. (1998). A new world ocean climatology: Optimal interpolation of historical
and WOCE hydrographic data on neutral surfaces. WOCE report 162/98, 40pp.
Gouretski, V. V., & Jancke, K. (1999). A description and quality assessment of the historical hydrographic
data for South Pacific Ocean. Journal of Atmospheric and Oceanic Technology, 16, 1791-1815.
Gamsakhurdiya, G. R., Meshchanov, S. L., & Shapiro, G. K. (1991). Seasonal variations in the distribution
of Red Sea Waters in the northwestern Indian Ocean. Oceanology, 31, 32-37.
Kobayashi, T., Ichikawa, Y., Takatsuki, Y., Suga, T., Iwasaka, N., Ando, K., Mizuno, K., Shikama, N., &
Takeuchi, K. (2002). Quality control of Argo data based on high quality climatological dataset
(HydroBase) I. ARGO Technical Report FY2001, Japan Marine Science and Technology Center, 36-48.
Kobayashi T., & Minato, S. (2005a). Importance of reference dataset improvements for Argo delayed-mode
quality control. Journal of Oceanography (in press).
Kobayashi, T., & Minato, S. (2005b). What observation scheme should we use for profiling floats to
achieve the Argo goal for salinity measurement accuracy? - Suggestions from software calibration -.
Journal of Atmospheric and Oceanic Technology (in press).
Levitus, S. (1982). Climatological Atlas of the World Ocean. National Oceanographic and Atmospheric
Administration, 173pp.
Levitus, S., Boyer, T. P., Conkright, M. E., O'Brien, T., Antonov, J., Stephens, C., Stathoplos, L., Johnson,
D., & Gelfeld, R. (1998). World Ocean Database 1998: Volume 1: Introduction. NOAA Atlas NESDIS 18,
33
U.S. Government Printing Office, Washington, D.C., 346pp.
Lozier, S. M., Owens, W. B., & Curry, R. G. (1995). The climatology of the North Atlantic. Progress in
Oceanography, 36, 1-44.
MaCartney, M. S. (1977). Subantarctic mode water. Deep-Sea Research, 24 (suppl.), 103-119.
MaCartney, M. S. (1982). The subtropical recirculation of mode waters. Journal of Marine Research, 40,
427-464.
Macdonald, A. M., Suga, T., & Curry, R. G. (2001). An isopycnally averaged North Pacific Climatology.
Journal of Atmospheric and Oceanic Technology, 18, 394-420.
Mantyla, A. W., & Reid, J. L. (1995). On the origin of deep and bottom waters of the Indian Ocean. Journal
of Geophysical Research, 100, 2417-2439.
Marine Information Research Center (2001). MIRC Ocean Dataset 2001 Documentation. MIRC Technical
Report, 1, 169pp. (in Japanese).
McCarthy, M. C., & Talley L. D. (1999). Three-dimensional isoneutral potential vorticity structure in the
Indian Ocean. Journal of Geophysical Research, 104, 13,251-13,267.
McCreary, J. P. Jr., Kundu, P. K., & Molinari, R. L. (1993). A numerical investigation of dynamics,
thermodynamics and mixed-layer processes in the Indian Ocean. Progress in Oceanography, 31,
181-244.
Mizuno, K. (1995). Quality control of temperature-depth data for the studies on basin-scale oceanographic
variability. Bulletin of the National Research Institute of Far Seas Fisheries, 32, 147-171.
Molinari, R. L., Olson, D., & Reverdin, G. (1990). Surface current distributions in the tropical Indian
Ocean derived from compilations of surface buoy trajectories. Journal of Geophysical Research, 95,
7217-7238.
Morrison, J. M. (1997). Inter-monsoonal changes in the T-S properties of the near-surface waters of the
northern Arabian Sea. Geophysical Research Letters, 24, 2553-2556.
34
Murray, S. P., & Johns, W. (1997). Direct observations of seasonal exchange through the Bab el Mandab
Strait. Geophysical Research Letters, 24, 2557-2560.
National Oceanographic Data Center (1994). World Ocean Atlas 1994, CD-ROM Data Set Documentation.
National Oceanographic Data Center Informal Report 13, 30 pp., National Oceanic and Atmospheric
Administration, Silver Spring, Md.
National
Research
Institute
of
Far
Seas
(1999).
Far
Seas
Collection.
online,
dataset,
http://pcocn4.enyo.affrc.go.jp/products.html.
Olson, D. B., Hitchcock, G. L., Fine, R. A., & Warren, B. A. (1993). Maintenance of the low-oxygen layer in
the central Arabian Sea. Deep-Sea Research II, 40, 673-685.
Peter, B. N., & Mizuno, K. (2000). Annual cycle of steric height in the Indian Ocean estimated from the
thermal field. Deep-Sea Research I, 47, 1351-1368.
Piola, A. R., & Georgi, D. T. (1982). Circulation properties of Antarctic Intermediate Water and
Subantarctic Mode Water. Deep-Sea Research, 29, 687-711.
Rao, R. R., & Sivakumar, R. (2000). Seasonal variability of near-surface thermal structure and heat
budget of the mixed layer of the tropical Indian Ocean from a new global ocean temperature
climatology. Journal of Geophysical Research, 105, 995-1015.
Rao, R. R., & Sivakumar, R. (2003). Seasonal variability of sea surface salinity and salt budget of the
mixed layer of the north Indian Ocean. Journal of Geophysical Research, 108 (C1), 3009,
doi:10.1029/2001JC000907.
Reid, J. L. (1981). On the mid-depth circulation of the world ocean. In B. A. Warren, & C. Wunsch (Eds.),
Evolution of Physical Oceanography (pp. 70-111). Cambridge, MA: MIT Press.
Reid, J. L. (2003). On the total geostrophic circulation of the Indian Ocean: flow patterns, tracers, and
transports. Progress in Oceanography, 56, 137-186.
Saji, N. H., Goswami, B. N., Vinayachandran, P. N., & Yamagata, T. (1999). A dipole in the tropical Indian
35
Ocean. Nature, 401, 360-363.
Semtner, A. J., Jr., & Chervin, R. M. (1992). Ocean general circulation from a global eddy-resolving model.
Journal of Geophysical Research, 97, 5493-5550.
Schott, F. (1983). Monsoon response of the Somali Current and associated upwelling. Progress in
Oceanography, 12, 357-381.
Schott, A. F., & McCreary, J. P. (2001). The monsoon circulation of the Indian Ocean. Progress in
Oceanography, 51, 1-123.
Shenoi, S. S. C., Saji, P. K., & Almeida, A. M. (1999). Near-surface circulation and kinetic energy in the
tropical Indian Ocean derive from Lagrangian drifters. Journal of Marine Research, 57, 885-907.
Shetye, S. R., Gouveia, A. D., Shankar, D., Shenoi, S. S. C., Vinayachandran, P. N., Sundar, D., Michael, G.
S., & Nampoothiri, G. (1996). Hydrography and circulation in the western Bay of Bengal during the
northeast monsoon. Journal of Geophysical Research, 101, 14,011-14,025.
Stramma, L. (1992). The South Indian Ocean Current. Journal of Physical Oceanography, 22, 421-430.
Stramma. L., & Lutjeharms, J. R. E. (1997). The flow field of the subtropical gyre of the South Indian
Ocean. Journal of Geophysical Research, 102, 5513-5530.
Suga, T., Motoki, K., Aoki, Y., & Macdonald, A. M. (2004). The North Pacific climatology of winter mixed
layer and mode waters. Journal of Physical Oceanography, 34, 3-22.
Talley, L. D. (1988). Potential vorticity distribution in the North Pacific. Journal of Physical
Oceanography, 18, 89-106.
The Argo Science Team: Roemmich, D., Boebel, O., Desaubies, Y., Freeland, H., Kim, K., King, B., LeTraon,
P.-Y., Molinari, R., Owens, W. B., Riser, S., Send, U., Takeuchi, K., & Wijffels, S. (2001). Argo: the
global array of profiling floats. In C. J. Koblinsky, & N. R. Smith (Eds.), Observing the Oceans in the
21st Century (pp. 248-258), Bureau of Meteorology, Melbourne, Australia: Godae Project Office.
Warren, B. A., & Johnson, G. C. (1992). Deep currents in the Arabian Sea in 1987. Marine Geology, 104,
36
279-288.
Webster, P. J., Moore, A. M., Loschnigg, J. P., & Leben, R. R. (1999). Coupled ocean-atmosphere dynamics
in the Indian Ocean during 1997-98. Nature, 401, 356-359.
Wong, A. P. S., Johnson, G. C., & Owens, W. B. (2003). Delayed-mode calibration of autonomous CTD
profiling float salinity data by θ-S climatology. Journal of Atmospheric and Oceanic Technology, 20,
308-318.
Wyrtki, K. (1971). Oceanographic atlas of the International Indian Ocean Expedition, 531pp., Washington,
D. C.: National Science Foundation.
37
Captions of Figures and Tables
Table 1: Number of stations and observation layers in the raw dataset of the Indian Ocean HydroBase
(IOHB).
Table 2: Number of hydrographic stations by country in the Indian Ocean (from the IOHB) and the World
Ocean (from World Ocean Dataset 1998; WOD98). Statistics for the WOD98 result from Ocean
Station Data (OSD) in Levitus et al. (1998).
Figure 1: The Indian Ocean including main bottom topography features (B = basin; P = plateau). The
Indian Ocean HydroBase (IOHB) includes hydrographic data in the area enclosed by the solid line.
The dashed line surrounds the area where suspicious sea surface stations in the MODS2001 and FSC
were removed (see section 3.3.2).
Figure 2: Quality control flowchart for the IOHB.
Figure 3: Individual station data in the area of 10-20°S, 110-120°E in (a) raw and (b) quality-controlled
datasets of the IOHB. Yellow and light blue represent Japanese observations from MODS2001 and the
FSC, respectively. Stations from WOD98v2 are plotted in black except for those observed by the USSR
and unknown countries, which are plotted in red and green, respectively. Dark blue in panel (b)
denotes WHP-CTD data.
Figure 4: An example of suspicious profiles removed from the raw dataset by visual inspection. Red and
blue represent the profiles obtained at 10-20°S, 70-80°E by particular cruises of unknown ships of
99-D in 1962 and 99=M in 1982, respectively.
Figure 5: An example of suspicious CTD stations (red) to be removed before the statistics check. Solid
circles and line segments represent the means and the slope of the linier fit θ-S curve (see section 3.5
in the details), respectively, and those of orange and light blue are calculated from the data sets with
and without suspicious CTD stations. The statistics of water mass from 195 stations (black and red) in
38
the area of 10-15°N, 90-95°E are distorted to shift toward higher salinity by only 5 red CTD stations
(measured by R/V Akademik Korolev of USSR in 1979) due to their higher vertical resolution.
Figure 6: The θ-S distributions of sea surface data in the area of 10–15°S, 110–115°E. Gray circles
represent data derived from stations with only sea surface data, and open circles represent sea surface
data from stations with subsurface measurements. Crosses show subsurface measurements in the
area.
Figure 7: An example of suspicious nutrient data removal. Structure of nitrate (µ mol kg-1) observed at
10–20°N, 60–70°E. The vertical axis is (a) depth (m), and (b) potential temperature (°C), respectively.
The data shown in gray (measured by R/V Akademik Vervadskiy in 1980 and R/V Akademik Korolev
in 1975) are excluded from the IOHB.
Figure 8: Geographic binning used for the statistical checks. Thick lines represent the boundaries of
geographical bins for layers below 1000 m.
Figure 9: Example of statistical checks in a geographic bin (similar to Figure 6 in Macdonald et al., 2001).
Within each isopycnal bin, solid squares represent the means, and line segments indicate the slope of
the linear fit θ-S curve at the location of the mean and of ±2.3 standard deviations from the means,
respectively. θ-S data outside 2.3-σ are eliminated.
Figure 10: Distributions of (a) all quality-controlled stations and (b) those exclusive of the sea surface
stations in the IOHB. Dark blue denotes WHP-CTD data. Yellow and blue represent Japanese
observations from MODS2001 and the FSC, respectively. Stations from WOD98v2 are plotted in black
except for those observed by the USSR and unknown countries, which are plotted in red and green,
respectively.
Figure 11 Distribution of all quality-controlled stations containing (a) oxygen (22,143 stations), (b)
phosphate (17,419), (c) silicate (13,801), and (d) nitrate (6231), respectively. Colors used for stations
are as in Figure 10. Oxygen data were statistically quality-controlled in a manner similar to that for
39
temperature–salinity pairs. Other data were checked visually.
Figure 12: Removal ratio of data in each 10°×10° grid through quality control procedure. Horizontal bars
in each grid show the ratios for the stations (top), temperature and salinity layers (middle), and
oxygen data (bottom). Oxygen data were removed through the removal of temperature and salinity
measurements that also include oxygen data.
Figure 13: Seasonal change in the number of (a) quality-controlled hydrographic stations and (b) 2.5°×2.5°
grids covered by at least one station in the IOHB. Subsurface means that the sea surface stations are
excluded. The full range of (b) is 1355; thus, the entire Indian Ocean is covered.
Figure 14: Distributions of the quality-controlled stations taken in (a) February and (b) September. Colors
used for the stations are as in Figure 10.
Figure 15: Number of raw and quality-controlled stations as a function of depth bins for (a) the entire
IOHB and for the following source datasets: (b) WOD98v2, (c) MODS2001, (d) FSC, and (e) WHP-CTD.
The number of the quality-controlled stations by each source dataset is shown on the right-hand side
of panel (a).
Figure 16: Number of (a) stations and (b) observation layers in the IOHB as a function of the year after
1920. Pre-1920 quality-controlled stations and observations are 86 and 372 from the WOD98v2,
respectively.
Figure 17: Distributions of (a) temperature, (c) salinity, (e) dynamic height (reference is 1000 dbar) at the
sea surface, and (g) pressure, (i) potential temperature, and (k) salinity on 24.5-σ0 isopycnal surface at
the peak of the Northeast Monsoon (January/February) and (b, d, f, h, j, and l) those at the peak of the
Southwest Monsoon (July/August).
Figure 18: Distributions of the averages and standard deviations for (a), (b) pressure; (c), (d) potential
temperature; (e), (f) salinity; (g), (h) dissolved oxygen; and averages for (i) potential vorticity and (j)
stream function (reference is 2000 dbar) on the 26.7-σ0 isopycnal surface in the IOHB.
40
Figure 19: As in Figure 18, except for the 27.3-σ0 isopycnal surface.
Figure 20: Distributions of the averages and standard deviations for (a), (b) pressure; (c), (d) potential
temperature; (e), (f) salinity; and averages for (g) dissolved oxygen and (h) potential vorticity on the
36.92-σ2 isopycnal surface in the IOHB.
Figure 21: As in Figure 20, except for the 45.90-σ4 isopycnal surface.
Figure 22: Distributions of (a) pressure, (c) potential temperature, (e) salinity, (g) dissolved oxygen, (h)
potential vorticity on the 26.7-σ0 isopycnal surface in the WOA01, and the differences between the
climatologies (IOHB – WOA01) for (b) pressure, (d) potential temperature, and (f) salinity.
Figure 23: As in Figure 22, except for the 27.3-σ0 isopycnal surface.
Figure 24: As in Figure 22, except for the 36.92-σ2 isopycnal surface.
Figure 25: Meridional distributions of (a), (b) salinity; (d), (e) salinity; and (g), (h) potential vorticity
against density (σ0 and σ2 for the upper and lower panel, respectively) along 65°E of (a, d, g) the IOHB
and (b, e, h) the WOA01. Panels (c), (f), and (i) show their difference between the climatologies (IOHB
– WOA01).
Figure 26: Comparison of θ-S structures between the individual observations (black) and the climatology
(gray) for the (a) IOHB and (b) WOA01 in the region of 35–45°S, 55–65°E.
Figure 27: Meridional distributions of (a), (b) salinity; (d), (e) salinity; and (g), (h) potential vorticity
against density (σ0 and σ2 for the upper and lower panel, respectively) along 65°E of the climatologies
constructed by the (a, d, g) isopycnal and (b, e, h) isodepth averagings from the same IOHB dataset of
quality-controlled individual stations. Both climatologies use the same smoothing parameter with
meridional scale of 2.5°.
Panels (c), (f), and (i) show their difference between the climatologies
(isopycnal average – isodepth average).
Figure 28: Salinity calibrations for the float data of (a) WMO ID 5900145 (2.6°C), (b) 1900310 (2.6°C), and
(c) 5900278 (2.4°C) by Wong et al. (2003) method with the reference dataset derived from the IOHB.
41
Left panels: Float measurement locations are shown by large gray diamonds except for the latest
position by large circle. Dots represent historical reference data. Right panels: Time series of
measured (line with solid circles) and calibrated salinities (gray belt: its center represented by while
line shows the optimal values, and half of its width represents calibration errors) interpolated to
each temperature. Solid line with vertical bars represents the estimated climatological salinity and
its mapping errors, respectively. Gray circle represents the “true” salinity measured by shipboard
CTD at the float deployment.
42
Table 1:
Dataset
Stations T/S obs.
WOD98v2
47396 759948
MODS2001
4994
28588
FSC
18542
24785
WHP-CTD
4067
Total
74999 813321
43
Table 2:
In d ia n O c e a n
W o rld O c e a n
In d ia n O c e a n H y d ro B a s e (q u a lity c o n tro lle d )
W O D 98
E x c lu s io n o f th e s e a
s u rfa c e s ta tio n s
C o u n try
A ll s ta tio n s
U n ite d S ta te s
5327
7 .7 %
5327
1 0 .5 %
272905
1 7 .9 %
17225
2 4 .8 %
17219
3 4 .0 %
231591
1 5 .2 %
9652
1 3 .9 %
9565
1 8 .9 %
32318
2 .1 %
Japan
20550
2 9 .6 %
2250
4 .4 %
204053
1 3 .4 %
O th e rs
16692
2 4 .0 %
16268
3 2 .1 %
785860
5 1 .5 %
T o ta l
69446
1 0 0 .0 %
50629
1 0 0 .0 %
1526727
1 0 0 .0 %
USSR
U nknow n
P ro file s in O S D