• Banks, H., , and R. Wood, 2002: Where to look for anthropogenic climate change in the ocean. J. Climate, 15 , 879891.

  • Banks, H. T., , and N. L. Bindoff, 2003: Comparison of observed temperature and salinity changes in the Indo-Pacific with results from the coupled climate model HadCM3: Processes and mechanisms. J. Climate, 16 , 156166.

    • Search Google Scholar
    • Export Citation
  • Banks, H. T., , R. A. Wood, , T. C. Johns, , and G. S. Jones, 2000: Are observed decadal changes in intermediate water a signature of anthropogenic climate change? Geophys. Res. Lett., 27 , 29612964.

    • Search Google Scholar
    • Export Citation
  • Banks, H. T., , R. Wood, , and J. Gregory, 2002: Changes to Indian Ocean subantarctic mode water in a coupled climate model as CO2 forcing increases. J. Phys. Oceanogr., 32 , 28162827.

    • Search Google Scholar
    • Export Citation
  • Belkin, I. M., , and A. L. Gordon, 1996: Southern Ocean fronts from the Greenwich meridian to Tasmania. J. Geophys. Res., 101 , 36753696.

    • Search Google Scholar
    • Export Citation
  • Bindoff, N. L., , and T. J. McDougall, 2000: Decadal changes along an Indian Ocean section at 32°S and their interpretation. J. Phys. Oceanogr., 30 , 12071222.

    • Search Google Scholar
    • Export Citation
  • Bryden, H. L., , E. L. McDonagh, , and B. A. King, 2003a: Changes in ocean water mass properties: Oscillations or trends? Science, 300 , 20862088.

    • Search Google Scholar
    • Export Citation
  • Bryden, H. L., and Coauthors, 2003b: RRS Charles Darwin Cruise 139, 01 Mar–15 Apr 2002; Trans-Indian hydrographic section across 32°S. Southampton Oceanography Centre Cruise Rep. 45, 122 pp.

  • Bryden, H. L., , L. M. Beal, , and L. M. Duncan, 2005: Structure and transport of the Agulhas Current and its temporal variability. J. Oceanogr., in press.

    • Search Google Scholar
    • Export Citation
  • Deacon, G. E. R., 1937: The hydrology of the Southern Ocean. Discovery Rep. 15, 124 pp.

  • Domingues, C. M., , S. Wijffels, , M. Tomczak, , and J. A. Church, 1999: Volume transports and structure of the Leeuwin Current at 22°S—WOCE ICM6. International WOCE Newsletter, No. 37, WOCE International Project Office, Southampton, United Kingdom, 36–40.

  • Donohue, K., , and J. Toole, 2003: A near-synoptic survey of the Southwest Indian Ocean. Deep-Sea Res., 50B , 18931931.

  • Gordon, A. L., , J. R. E. Lutjeharms, , and M. L. Grundlingh, 1987: Stratification and circulation at the Agulhas Retroflection. Deep-Sea Res., 34 , 565599.

    • Search Google Scholar
    • Export Citation
  • Jenkins, W. J., 1980: Tritium and 3He in the Sargasso Sea. J. Mar. Res., 38 , 533569.

  • Joos, F., , G-K. Plattner, , T. F. Stocker, , A. Körtzinger, , and D. W. R. Wallace, 2003: Trends in marine dissolved oxygen: Implications for ocean circulation changes and the carbon budget. Eos, Trans. Amer. Geophys. Union, 84 , 197201.

    • Search Google Scholar
    • Export Citation
  • Josey, S. A., , E. C. Kent, , and P. K. Taylor, 1998: The Southampton Oceanography Centre (SOC) Ocean–Atmosphere Heat, Momentum, and Freshwater Flux Atlas. Southampton Oceanography Centre Rep. 6, 30 pp.

  • Marsh, R., , A. J. G. Nurser, , A. P. Megann, , and A. L. New, 2000: Water mass transformation in the Southern Ocean of a global isopycnal coordinate GCM. J. Phys. Oceanogr., 30 , 10131045.

    • Search Google Scholar
    • Export Citation
  • McCartney, M. S., 1977: Subantarctic Mode Water. Deep-Sea Res., 24 , (Suppl.). 103119.

  • McCartney, M. S., 1982: The subtropical recirculation of Mode Waters. J. Mar. Res., 40 , (Suppl.). 427464.

  • Rintoul, S. R., , and M. H. England, 2002: Ekman transport dominates local air–sea fluxes in driving variability of Subantarctic Mode Water. J. Phys. Oceanogr., 32 , 13081321.

    • Search Google Scholar
    • Export Citation
  • Roemmich, D., 1983: Optimal estimation of hydrographic station data and derived fields. J. Phys. Oceanogr., 13 , 15441549.

  • Stramma, L., , and J. R. E. Lutjeharms, 1997: The flow field of the subtropical gyre of the South Indian Ocean. J. Geophys. Res., 102 , 55135530.

    • Search Google Scholar
    • Export Citation
  • Talley, L. D., , and M. O. Baringer, 1997: Preliminary results from WOCE hydrographic sections at 80°E and 32°S in the Central Indian Ocean. Geophys. Res. Lett., 24 , 27892792.

    • Search Google Scholar
    • Export Citation
  • Toole, J. M., , and M. E. Raymer, 1985: Heat and fresh water budgets of the Indian Ocean—Revisited. Deep-Sea Res., 32 , 917928.

  • Toole, J. M., , and B. A. Warren, 1993: A hydrographic section across the subtropical South Indian Ocean. Deep-Sea Res., 40A , 19732019.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Positions of the observations used in this study. CTD stations from RRS Charles Darwin occupation (Mar 2002, dark blue circles), R/V Knorr occupation (Jun 1995, red triangles; Mar–Apr 1995, red upside-down triangles), and Darwin occupation (Nov 1987, green stars). Bottle stations from R/V Atlantis II occupation (Jul 1965, black diamonds) and RRS Discovery occupation (Apr 1936, light blue triangles). Depth contours of 0, 2000, and 4000 m are shown, and depths shallower than 2000 m are shaded.

  • View in gallery

    (a) The θ, (b) S, and (c) AOU sections for 2002 data. AOU is the difference between the saturated and observed dissolved oxygen concentrations.

  • View in gallery

    (a) Potential vorticity contoured at 2.5, 5.0, 7.5, and 10 × 10−11 (m s)−1 for the 20-dbar CTD data from the 2002 occupation of the section: PV = |f|/ρ.dρ/dz, where f is the Coriolis parameter, ρ is density, and z is depth. Shaded areas represent all PV values less than 5.0 × 10−11 (m s)−1. (b) The gray contours are the 9°, 10°, 11°, 12°, and 13°C potential temperature isotherms. The bold black line in (a) and (b) shows the position of the PV minimum between 200 and 800 dbar at each station.

  • View in gallery

    (a) The θ, (b) S, (c) PV, and (d) percentage of oxygen saturation at the PV minimum between 200 and 800 dbar for each 2002 (*) and 1987 (ˆ) station. Only data between 200 and 800 dbar are analyzed to ensure that the identified PV minimum is associated with SAMW. Stations shallower than 700 dbar are omitted as these did not necessarily sample the PV minimum associated with SAMW.

  • View in gallery

    The (θS) plots for the thermocline at (a) the western and (b) the eastern end of the sections. The lines represent the average (on neutral density surfaces) θ and S properties for the 1987 (green), 1995 (red), and 2002 (dark blue) occupations. All of the 1936 (light blue triangles) and 1965 (diamonds) bottle data that fall into the respective longitude ranges are plotted. The contours are lines of constant neutral density.

  • View in gallery

    Salinity changes from 1987 to 2002 (2002 minus 1987). Vertical salinity profiles are interpolated onto potential temperature levels in tenths of a degree and then horizontally onto a half-degree longitude grid. No smoothing is necessary. The salinity changes are contoured with an interval of 0.02 (Salinity has no dimensions, but one unit of practical salinity is essentially equivalent to one part per thousand of dissolved salt.) Shading denotes increasing salinity on temperature levels; no shading denotes freshening. The bold contour indicates no change. The stars indicate the potential temperature of the PV minimum associated with SAMW in 2002 as shown in Fig. 4a. The temperature range from 18°C to 4°C is approximately equivalent to depths of 50 to 1200 m.

  • View in gallery

    Changes between 1987 and 2002 θS curves on neutral density surfaces (gray lines) and for the minimum change (black lines). Dashed lines are one std dev of uncertainty on the difference. Changes are 2002 minus 1987, so a positive change in θ (S) represents a warming (increase in salinity). (a) Changes in θ (dθ) and (b) changes in S (dS) are shown for the western part of the section (40°–70°E) in (a) and (c), (d) for the eastern end of the section (80°E to Australia). Values of neutral density are shown on the right-hand side of each of the plots.

  • View in gallery

    Minimum distance changes for the western end of the section between 40° and 70°E. Differences between 1987 and 2002 (red), 1987 and 1995 (green), and 1995 and 2002 (dark blue) θS curves. Dashed lines are one standard deviation uncertainty on the difference. Differences between 1987 θS curve and 1965 bottle data (black diamonds) and 1987 θS curve and 1936 bottle data (light blue triangles). Horizontal lines represent one std dev of uncertainty on the difference. Changes are for the later year minus the earlier year, so a positive change in θ (S) represents a warming (increase in salinity) over that time period: (a) changes in θ () and (b) changes in S (dS). Values of neutral density are shown on the right-hand side of each of the plots.

  • View in gallery

    Same as in Fig. 8, but for isopycnal changes (on neutral density surfaces) for the eastern end of the section between 80°E and Australia.

  • View in gallery

    Differences of (a) potential temperature () and (b) salinity (dS) between 1987 and 2002 on average pressure surfaces between 40° and 70°E. Differences are 2002 minus 1987, so a positive change in θ (S) represents a warming (increase in salinity). The dotted lines represent the differences between the thermoclines if they are offset by +20 and –20 dbar.

  • View in gallery

    Colored blocks indicate the position of the ventilation sites in a general circulation model for thermocline mode water on portions of the 32°S section indicated by the solid black line (a) between 40° and 70°E and (b) east of 70°E. The color coding represents the temperature of the mode water at the section.

  • View in gallery

    Dissolved oxygen changes from 1987 to 2002 (2002 minus 1987) in μmol kg−1. Vertical dissolved oxygen profiles are interpolated onto potential temperature levels in tenths of a degree and then horizontally onto a half-degree longitude grid. The oxygen changes are contoured with an interval of 5 μmol kg−1. Shading denotes increasing oxygen on temperature levels; no shading denotes decreasing oxygen concentrations. The bold contour indicates no change. The stars indicate the potential temperature of the PV minimum associated with SAMW in 2002 as shown in Fig. 4a. The temperature range from 18°C to 4°C is approximately equivalent to depths of 50 to 1200 m.

  • View in gallery

    (a) Age since thermocline waters were most recently in contact with the sea surface vs potential temperature for individual bottles. Age is calculated from helium and tritium data collected on the 1995 WOCE cruises. Nonlinear mixing means that this He–Tr age is a lower bound to the actual age spectrum. I5W data (shown in blue) were collected between 40° and 50°E; I5E data (shown in red) were collected east of 85°E. (b) OUR calculated from the AOU (difference between saturated and measured oxygen concentration) divided by the He–Tr age. I5W data (shown in blue) were collected between 40° and 50°E; I5E data (shown in red) were collected east of 85°E. The green dots are OUR calculated for the potential temperature of each of the 1987 and 2002 bottle data between 3° and 17°C and deeper than 100 dbar. These OUR values are calculated using a linear regression of the 1995 OURs over 2°C, centered at the potential temperature of the bottle.

  • View in gallery

    Changes in age (since the water was last ventilated) between 2002 and 1987, averaged in 1°C potential temperature bins. Age is derived from the OUR–potential temperature relationship for 1995 data and AOU for 1987 and 2002 data. (a) At the western end of the section between 40° and 70°E and (b) at the eastern end of the section, east of 80°E. The dotted lines show one std dev of uncertainty in the difference between the water mass ages. This uncertainty reflects the variability of the oxygen concentrations in each temperature bin.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 162 162 22
PDF Downloads 79 79 7

Decadal Changes in the South Indian Ocean Thermocline

View More View Less
  • 1 Southampton Oceanography Centre, Southampton, Hampshire, United Kingdom
© Get Permissions
Full access

Abstract

A significant change in properties of the thermocline is observed across the whole Indian Ocean 32°S section between 1987 and 2002. This change represents a reversal of the pre-1987 freshening and decreasing oxygen concentrations of the upper thermocline that had been interpreted as a fingerprint of anthropogenic climate change. The thermocline at the western end of the section (40°–70°E) is occupied by a single variety of mode water with a potential temperature of around 13°C. The thermocline at the eastern end of the 32°S section is occupied by mode waters with a range of properties cooling from ∼11°C at 80°E to ∼9°C near the Australian coast. The change in θS properties between 1987 and 2002 is zonally coherent east of 80°E, with a maximum change on isopycnals at 11.6°C. Ages derived from helium–tritium data imply that the mode waters at all longitudes take about the same time to reach 32°S from their respective ventilation sites. Dissolved oxygen concentration changes imply that all of the mode water reached the section ∼20% faster in 2002 than in 1987.

Corresponding author address: Dr. Elaine L. McDonagh, Southampton Oceanography Centre, Empress Dock, Southampton SO14 3ZH, United Kingdom. Email: elm@soc.soton.ac.uk

Abstract

A significant change in properties of the thermocline is observed across the whole Indian Ocean 32°S section between 1987 and 2002. This change represents a reversal of the pre-1987 freshening and decreasing oxygen concentrations of the upper thermocline that had been interpreted as a fingerprint of anthropogenic climate change. The thermocline at the western end of the section (40°–70°E) is occupied by a single variety of mode water with a potential temperature of around 13°C. The thermocline at the eastern end of the 32°S section is occupied by mode waters with a range of properties cooling from ∼11°C at 80°E to ∼9°C near the Australian coast. The change in θS properties between 1987 and 2002 is zonally coherent east of 80°E, with a maximum change on isopycnals at 11.6°C. Ages derived from helium–tritium data imply that the mode waters at all longitudes take about the same time to reach 32°S from their respective ventilation sites. Dissolved oxygen concentration changes imply that all of the mode water reached the section ∼20% faster in 2002 than in 1987.

Corresponding author address: Dr. Elaine L. McDonagh, Southampton Oceanography Centre, Empress Dock, Southampton SO14 3ZH, United Kingdom. Email: elm@soc.soton.ac.uk

1. Introduction

The south Indian Ocean thermocline water is ventilated to the south of the subtropical gyre, subducts and moves northward. At 32°S, this water is found between surface water (seasonally mixed and approximately isohaline) and intermediate water (salinity minimum) and lies between approximately 150 and 1150 dbar. Although potential temperature (θ) and salinity (S) vary monotonically with depth, there exist regions in the thermocline, occupied by (θ, S) modes, where vertical gradients are relatively small. These mode waters, observed in all of the Southern Hemisphere subtropical gyres, were named Subantarctic Mode Water (SAMW) by McCartney (1977, 1982).

Mode waters originate as wintertime deep mixed layers and as such hold information about the integrated surface fluxes of heat and freshwater. This information is useful because the rapidly varying surface flux fields are hard to quantify, particularly in the data-sparse Southern Ocean where these mode waters form. The properties of the mode waters may also be altered by changes in the advection of water into the formation region. Rintoul and England (2002) found that the observed variability in SAMW properties south of Tasmania was primarily forced by changes in the amount of equatorward Ekman flux of cool, fresh Antarctic surface water across the subantarctic front. Thus, changes in the surface flux fields and advection of water into the formation region cause the properties of the deep winter mixed layer to vary, which ultimately causes variability in the temperature and salinity properties of the subtropical thermocline.

In the Third Hadley Centre Coupled Ocean–Atmosphere General Circulation Model (HadCM3), Banks and Wood (2002) found that subsurface temperature and salinity variability had a higher signal-to-noise ratio as a climate change indicator than one-off observations of heat transport or indicators of circulation such as thermohaline overturning and the Indonesian throughflow. The southern boundary of the Indian Ocean is one of six geographical regions identified by Banks and Wood (2002) as having a particularly high signal-to-noise ratio for detecting climate change. In a study of observations in the south Indian Ocean subtropical gyre, Bindoff and McDougall (2000) found that in the 25 yr prior to 1987 the thermocline waters of the Indian Ocean at 32°S had cooled and freshened on isopycnals. These observed changes in the Indian Ocean SAMW are most similar to the changes in HadCM3 when the model is driven by anthropogenic forcing (Banks et al. 2000). Banks and Bindoff (2003) classify this cooling and freshening on isopycnals in midlatitudes as the fingerprint of anthropogenic forcing.

Given that the climate models have only recently identified the southern Indian Ocean thermocline as a good place to identify climate change, it is fortuitous that the relatively remote 32°S section in the Indian Ocean has been entirely occupied on four occasions over a 66-yr period, with an additional partial reoccupation during the World Ocean Circulation Experiment (WOCE; Fig. 1). The first transindian section in 1936 was made as part of the RRS Discovery expeditions (Deacon 1937). The second aboard R/V Atlantis II in 1965 was part of the Indian Ocean Expedition (Toole and Raymer 1985). The third in 1987 was fitted into RRS Charles Darwin’s inaugural round-the-world voyage (Toole and Warren 1993). WOCE reoccupied the western (Donohue and Toole 2003) and eastern (Talley and Baringer 1997) parts of the section aboard R/V Knorr in 1995. In March–April 2002, we reoccupied the 32°S transindian section aboard Darwin with a principal goal to measure the meridional overturning circulation across this southern boundary of the Indian Ocean. Here we examine the evolution of water mass properties from 1936 to 2002.

The changes in thermocline properties at the western end of the section were considered by Bryden et al. (2003a). That study showed that the thermocline mode water had become saltier on potential temperature surfaces between 1987 and 2002, reversing the pre-1987 freshening trend that had previously been noted. By 2002 the mode water west of 80°E had reverted to nearly the same properties observed in 1936. Here we will extend that analysis to the entire section and consider a variety of techniques to quantify the variability.

Interpreting the differences in thermocline properties between occupations is subtle. We use several techniques to quantify the differences in order to learn as much as possible from the data without presupposing the physical mechanisms behind the change. Changes in properties of the potential vorticity (PV) minimum between 1987 and 2002 are used to ascertain possible mechanisms that force the mode water variability. Differences on potential temperature surfaces between 1987 and 2002 highlight the spatial pattern of change. Changes have zonal coherence over the eastern and western parts of the section, and this justifies examining the zonally averaged differences between occupations.

There are many ways of looking at the differences between θS curves. Although the difference between average θS curves can be difficult to interpret in terms of forcing mechanisms, such differences are statistically well defined. We use the minimum difference between θS curves as well as the change on isopycnals to quantify the difference between the thermocline properties of the five occupations.

We quantify the change in the heat and salt content of the thermocline by introducing the average pressure profile for each of the 1987 and 2002 datasets. Particle tracking in a general circulation model and historical data are used to ascertain where the thermocline mode waters ventilate. A combination of these results and changes in the properties of the PV minimum allows us to diagnose the relative importance of different forcing mechanisms for the changes between the 1987 and 2002 thermoclines and where the forcing changes occurred.

2. Data

The south Indian Ocean’s western boundary current, the Agulhas Current, penetrates to 1000 m at its inshore edge and deepens to more than 2000 m 50 km offshore (Bryden et al. 2005). Data from the Agulhas Current are excluded from this analysis in order to focus on the recently ventilated thermocline. The eastern boundary current, the Leeuwin Current, also flows southward and is 30 km wide but penetrates to only 200 m (Domingues et al. 1999). We do not exclude data at the eastern end of the section because the shallow Leeuwin Current was not considered to have a significant effect on the northward-flowing thermocline.

Temperature and salinity measurements were made on each of the 1987, 1995, and 2002 cruises using a CTD. Data were interpolated onto a 20-dbar grid using a linear regression over 20 dbar. The 2002 θ and S data for the top 1500 dbar are shown in Figs. 2a,b. Only bottle data exist for the 1936 and 1965 sections: these temperature and salinity values are not interpolated, and the individual bottle values are used in this study. The positions of the CTD and bottle stations used here are shown in Fig. 1.

At some longitudes, the sections are separated by up to 2° of latitude. B. A. King and E. L. McDonagh (2004, unpublished manuscript, hereafter KMD) analyzed CTD data from floats that spanned the distance between the 1987 and 2002 sections. They found that latitudinal variation of salinity properties on θ surfaces is small compared to the changes between the 1987 and 2002 data. Therefore, the meridional separation of the sections should not affect our results significantly and is not considered further.

Bottle oxygen data from the 1987 occupation were interpolated onto the same 20-dbar grid as the CTD data using optimal interpolation routines based on the algorithm of Roemmich (1983). The oxygen measurements (calibrated to bottle samples) from the Seabird sensor on the 2002 occupation were considered to be of such a high quality (Bryden et al. 2003b) that they were used in preference to the bottle data to generate a 20-dbar dataset using the same method described above for the CTD θ and S data. Apparent oxygen utilization (AOU) is the difference between saturation and observed concentrations of dissolved oxygen. A section of AOU for the 2002 data is shown in Fig. 2c. Bottle oxygen data from the 1987, 1995 (I5W and I5E), and 2002 cruises and helium–tritium data for the I5W and I5E sections are also used.

3. The south Indian Ocean thermocline

In the introduction, we described the main thermocline as being above the salinity minimum associated with Antarctic Intermediate Water (AAIW), apparent in Fig. 2b at around 1000 dbar and below the seasonally mixed layer. We formalize this by looking at the stability parameter Rρ. Rρ = αdθ/βdS, where α and β are the thermal expansion and haline contraction coefficient of seawater, respectively. In the thermocline, Rρ > 1. Below the base of the thermocline (or equivalently the salinity minimum), Rρ becomes negative as the salinity stratification changes sign. At the top of the main thermocline, Rρ increases as the salinity stratification reduces in the seasonal thermocline. Identifying the top of the main thermocline is somewhat subjective, but we identify it as the depth where Rρ began to increase rapidly—for the 2002 data, this occurs when Rρ ≈ 5.5. In the 2002 data, the base of the thermocline lies, on average, at 1093 ± 109 dbar or 4.8° ± 0.2°C, and the top of the thermocline lies at 120 ± 56 dbar or 17.0° ± 0.3°C. Uncertainties are one standard deviation based on 115 stations in the 2002 section.

4. Properties of the thermocline mode water

a. Properties of the potential vorticity minimum

The mode waters that occupy the thermocline are not immediately obvious in plots of θ and S because of the monotonic nature of these variables in the thermocline (Figs. 2a,b). However, if one looks at potential vorticity (PV = |f|/ρ.dρ/dz, where f is the Coriolis parameter, ρ is density, and z is depth), the mode waters stand out as a band of low values (shaded region in Fig. 3a) that lie mostly between 300 and 700 dbar. There is zonal variation in the properties of the PV minimum. From the western end of the section to approximately 75°E, the PV minimum rises from about 600 to 300 dbar approximately following the 13°C isotherm (Fig. 3b). East of 75°E, the depth of the PV minimum is more constant at about 500 dbar. However, this isobar crosses through the bowl-shaped thermocline of the subtropical gyre, and the θ of the PV minimum decreases from approximately 11°C at 75°E to 9°C at the eastern end of the section. To more accurately examine zonal variation in the properties of thermocline mode water, we look at the properties of the PV minimum in the 1987 and 2002 data (Fig. 4). Although there are differences between the properties of the PV minimum in 1987 and 2002, the gross structure of properties with longitude is the same. Here we will examine this gross structure.

Excluding outliers, the θ and S properties of the PV minimum broadly fit into two groups. Between 40° and 70°E, there is a mode at ∼13°C. East of 80°E, the mode is cooler (11°–9°C) and fresher (Figs. 4a,b). Between 70° and 80°E, there is a transition between the two modes. West of 40°E, but outside of the southward flowing Agulhas Current, each of these types of SAMW appears to exist together with another mode at about 17°C. The 17°C mode, described as Subtropical Mode Water (STMW), was noted in the 1987 section (Toole and Warren 1993) and is also apparent in the 2002 data (Fig. 4a) but was not observed in the 1995 dataset (Donohue and Toole 2003). This STMW could be the result of winter mixing in the Agulhas retroflection zone (Gordon et al. 1987). If this is the case, then the mix of mode waters west of 40°E is recirculated, containing the ∼10° and 13°C modes carried south in the Agulhas Current and the locally formed 17°C mode. Therefore, we consider that the water west of 40°E does not represent the recently ventilated thermocline but is recirculating. This is consistent with the higher PV west of 40°E (Fig. 4c) and the lower oxygen saturation (other than in the locally formed 17°C mode; Fig. 4d), suggesting it has been longer since these waters ventilated.

b. Changes in the properties of the potential vorticity minimum

The properties of the PV minimum (shown in Fig. 4) are interpolated onto a 0.5° longitude grid and the average value of the difference (2002 minus 1987) between the two occupations is found. The averaging and differencing was done over the two longitude bands described in the previous section, 40°–70°E and 80°E–Australia. The resulting changes on PV minimum surfaces (and the associated uncertainties) for θ, S, and neutral density are shown in Table 1.

Between 40° and 70°E, the mode water of θ ≈ 13°C has exhibited a net cooling between 1987 and 2002 of 0.442° ± 0.107°C, decreased in salinity by 0.029 ± 0.016, and significantly (greater than two standard deviations) increased in density by 0.072 ± 0.011. East of 80°E, θ and S increase significantly by 0.366° ± 0.133°C and 0.078 ± 0.018, respectively, between 1987 and 2002. At the eastern end of the section, the changes in θ and S are density compensating with an insignificant density change of –0.011 ± 0.010 for the mode waters.

5. Quantifying changes in thermocline properties

Mesoscale variability and the associated heave of isopycnals dominated averages along a section (or differences between occupations) when the averaging (differencing) was done on pressure surfaces. For this reason, the average θ and S (Fig. 5b) were found on density surfaces, and this preserved the tight θS relationship that is seen when all of the CTD data from an occupation are overlaid. The differences between occupations are now examined.

a. Changes on potential temperature surfaces

An analysis of the zonal structure of the changes on θ surfaces was undertaken on the 1987 and 2002 CTD data. Vertical salinity profiles were interpolated onto potential temperature levels in tenths of a degree and then were horizontally interpolated onto a half-degree longitude grid. The difference between these fields is shown in Fig. 6. The difference is 2002 values minus 1987 values, so a positive (negative) difference represents a net increase in salinity (freshening) during that time period. At the western end of the section between 40° and 70°E, there is a zonally coherent increase in salinity above the 10°C isotherm between 1987 and 2002. This signal is centered around 13°C, the temperature of the mode water at the western end of the section, and represents a reversal of the pre-1987 trend reported by Bindoff and McDougall (2000).

East of 80°E, the upper thermocline also increased in salinity between 1987 and 2002 (Fig. 6). This sense of salinity change in the upper thermocline reverses the pre-1987 trend (as at the western end of the section). The reversal to lower thermocline freshening occurs at around 9°C (rather than 10°C at the western end of the section). There is a high level of zonal coherence in the increasing salinity signal above 9°C centered around 11°–12°C implying that the maximum change occurs at a single temperature (that of the warmest resident mode water), rather than following the temperature of the PV minimum, which changes from ∼11°C at 80°E to ∼9°C at the eastern end of the section (∼115°E). The zonal coherence of the differences observed at each end of the section (Fig. 6) justifies the use of differences between average properties in the subsequent sections.

b. Minimum and isopycnal differences between θ–S relationships

Two methods are used to quantify differences between θS properties. The first is the minimum distance between θS curves. We use 1987 as a baseline and look in a direction perpendicular to the θS relationship (when the θ and S axes are scaled by α and β, respectively). Formally this is the minimum distance √{(αdθ)2 + (βdS)2} from, for example, the 2002 θS curve. The second method is the differences between θS properties on isopycnals.

We initially look at both minimum distance and isopycnal changes to quantify the differences between thermoclines in 1987 and 2002 at the eastern and western ends of the section (Fig. 7). This allows us to briefly evaluate the two methods and also compare our results with previous studies such as Bindoff and McDougall (2000). The magnitude of the change in both θ and S is substantially smaller for the minimum distance values than for the isopycnal variations. Also, whilst the minimum distance salinity changes in the same sense as the isopycnal variations, the minimum distance dθ, which must be perpendicular to the slope of the 1987 θS curve, generally is of the opposite sign to the isopycnal dθ. This change in the sign of the temperature difference highlights the caution that should be used if one attempts to interpret the differences between θS curves in terms of buoyancy changes. We prefer to think of the differences between θS curves as a statistically well-defined way of looking at the changes in the properties of the thermocline.

Larger error bars for both methods in Figs. 7c,d than in Figs. 7a,b are a reflection of the less tight θS relationship at the eastern end of the section than at the western end. Higher variability in θS properties at the eastern end of the section than the western end was also noted by KMD, who studied Argo float data in conjunction with some of these CTD data.

Figure 7b shows a 0.093 increase (1987 to 2002) in salinity on isopycnal surfaces for the upper thermocline in the western segment centered at 12.9°C (γn = 26.66). This change almost entirely reverses the –0.12 pre-1987 freshening that Bindoff and McDougall (2000) observed on γn = 26.75 for the whole section. On the eastern part of the section, the maximum increase in salinity on isopycnals of 0.069 between 1987 and 2002 is observed at 11.6°C. Therefore, for both the eastern and western ends of the section, the upper-thermocline cooling and freshening trend observed by Bindoff and McDougall (2000) pre-1987 was reversed between 1987 and 2002. The lower-thermocline pre-1987 trend of freshening on isopycnals persists between 1987 and 2002 with a minimum value of –0.031 at 7.3°C on the western part of the section and –0.035 at 7.2°C on the eastern part of the section.

As the changes in the mode water properties are not isopycnal between 40° and 70°E (Table 1), simply looking at changes on isopycnals serves to exaggerate the differences since the isopycnals and the θS curve are nearly parallel (Banks et al. 2002). Thus we prefer to look at the minimum change differences. The density-compensating changes in the mode water east of 80°E between 1987 and 2002 (see small Δγn in Table 1) suggest that it may be more appropriate to consider changes on density surfaces for this portion of the section.

Tables 2 and 3 show the average changes at the western (minimum distance) and the eastern (isopycnal) ends of the section, respectively, over 1°C ranges that are representative of the maximum change in the upper and lower thermocline. At the far western end of the section, the AAIW layer is invaded by relatively high salinity Red Sea water (RSW) from the north. Figure 2b shows an erosion of the AAIW salinity minimum associated with higher concentrations of RSW particularly west of 50°E. As the 1995 section only samples west of 50°E, it has a much higher proportion of RSW in the intermediate water layer than either the 1987 or 2002 sections. This propagates through to the average θS curves (Fig. 5a) where the 1995 data have a relatively high salinity around the temperature of the intermediate water salinity minimum. We estimate that the effect of the RSW is confined to waters colder than approximately 7.7°C so we use the changes at that temperature to quantify the 1987–95 and 1995–2002 changes in the lower thermocline at the western end of the section.

Including the 1995 data in the analysis gives us information about how consistent the change has been since 1987. For the western part of the section, the upper thermocline increase in salinity and the lower thermocline freshening (as well as the associated temperature changes) occurred at about the same rate between 1987 and 1995 as between 1995 and 2002. This is evidenced by the green (1987–95 change) and dark blue (1995–2002 change) curves overlying one another in Fig. 8 as the two epochs that they represent cover approximately the same time period. If we look at the upper thermocline differences east of 80°E and include the 1995 data, then we see a different pattern. Figure 9 and Table 3 show the isopycnal changes between 1987 and 1995 (green curve, dθ = 0.243°C and dS = 0.061; Table 3) and between 1995 and 2002 (blue curve, = 0.017°C and dS = 0.004; Table 3). There are no significant changes in θS properties of the upper-thermocline mode waters in the eastern Indian Ocean thermocline between 1995 and 2002. The lower-thermocline changes imply a consistent net rate of freshening and cooling between 1987 and 1995, and 1995 and 2002. Furthermore the rate of change of the lower-thermocline water (if one uses the same differencing method) at around 7°C is the same for each of the post-1987 epochs in the western and eastern south Indian Ocean.

Figures 8 and 9 also show the pre-1987 changes. Both of these figures highlight the reversal in upper-thermocline property changes post-1987. These figures show the same sequence of changes at the eastern and western ends of the section for 1936–65 and 1965–87. In the upper thermocline, the 1936–87 changes broadly overlie the 1965–87 changes for both potential temperature and salinity for the western (Fig. 8, Table 2) and eastern (Fig. 9, Table 3) ends of the section. This implies that the pre-1987 changes occurred between 1965 and 1987 and that the θS properties of the 1936 and 1965 upper thermocline were similar for the whole section. The pre-1987 change of the lower thermocline is evident in the 1936–87 differences. This sense of change persists post-1987 and is centered at approximately 7°C. At around 7°C, the 1965–87 changes are close to zero at both the eastern and western portions of the section. This implies that θS properties of the 1965 and 1987 lower thermocline were similar and that the pre-1987 lower-thermocline changes occurred between 1936 and 1965.

We have chosen to quantify the differences between θS curves using different methods for the two subsections. Note however that the same conclusions on the sequence of changes in salinity would be drawn if isopycnal, minimum difference, or indeed changes on potential temperature surfaces were considered.

c. Changes in the heat and salt content

The different ways of defining the property changes allow us to quantify the difference between tight θS relationships for each of the years. However, this neglects changes in the stratification or equivalently changes along the θS relationship. Here we attempt to quantify the changes in the heat and salt content by including stratification. It is inappropriate to perform this analysis anywhere other than where the sections exactly overlie one another because although θS properties do not vary significantly over the meridional separation of the sections east of 80°E, variation in the depths of θ and S surfaces are likely to be significant. We use the 1987 and 2002 data in the western part of the 32°S section for this analysis because the thermoclines occupy almost exactly the same pressure range (to within 10 dbar) in the two datasets. Based on the stability parameter (; described in section 3), the 1987 and 2002 thermoclines lie, on average, between 220 and 1190 dbar between 40° and 70°E.

Differencing the thermocline properties on pressure surfaces is dominated by mesoscale heave. Therefore, we difference the thermocline properties (potential temperature, salinity, and pressure) after averaging on density surfaces. The average potential temperature and salinity profiles are then compared on pressure surfaces using the average pressure profiles. The standard error of the estimate of the mean pressure for each density surface (28 stations in 1987; 41 stations in 2002) is in the range of 8–17 dbar over most of the thermocline. The average 1987 θ and S profiles are subtracted from the average 2002 θ and S profiles, respectively (Fig. 10). The dotted lines in Fig. 10 show the change in θ and S if one of the profiles is offset by +20 and –20 dbar. This represents the uncertainty in matching the average thermocline properties on average depth.

Over the range of pressures occupied by the 13°C mode water (maximum depth of the 12°C isotherm is ∼600 dbar and minimum depth of the 14°C isotherm is approximately 200 dbar; Fig. 2a), the thermocline exhibits cooling with an extreme value of about −0.33°C at 360 dbar (Fig. 10a). Over this pressure range, the salinity difference stays very close to zero (Fig. 10b). Below the mode water and, on average, above the base of the upper thermocline at 10°C (750 dbar; Fig. 2a), the change in temperature properties reverses to warming, and the change in salinity becomes positive and up to 0.041 at 609 dbar. In the lower thermocline, the salinity change reduces to near zero, but the temperature change remains positive reaching a broad maximum of about 0.33°C between 900 and 1100 dbar.

Changes in heat and salt content have many potential sources of contamination and uncertainty, but they do show the same sense of change as the changes in properties of the PV minimum at the western end of the section (Table 1): cooling in mode water temperature with no significant change for salinity. The average change in temperature between 200 and 600 dbar between 1987 and 2002 is –0.20°C. In section 8 of this paper, we estimate a travel time of 7 yr from the ventilation region (between 40° and 45°S) to 32°S. The implied speed estimate can be combined with the 400-dbar layer thickness to give a volume flux of 2 m3 s−1 for each meter of longitude. If the volume flux is the same in 1987 and 2002, then an increase in surface heat flux of 3 W m−2 out of the ocean is required in the ventilation region. The consequences of a change in the volume flux of the mode water are discussed at the end of section 8.

6. Sources of mode water to the south Indian Ocean thermocline

To identify where waters observed in the thermocline along 32°S were last in the mixed layer, we looked at particle trajectories in the Global Isopycnic Model (GIM; Marsh et al. 2000). The section was split into two lengths to reflect the structure of the mode water observed on the section (section 4). The western part of the section was defined to be between 40° and 70°E and the eastern part of the section was definted to be between 70°E and Australia, consistent with the above analysis. In the model, the thermocline layer with the highest throughflow of water was identified for each part of the section. This was a warmer layer at the western end of the section (mean layer temperature = 16.35°C) than at the eastern end of the section (mean layer temperature = 15.14°C). Therefore, the model reproduces the observed mode water in its gross structure though the absolute temperatures are warmer in the model by 3°–5°C.

The model particle trajectories are then used to back-trace these “mode water” layers to where they were last ventilated (Figs. 11a,b). The color coding represents the temperature of the water within the respective mode water layers when they are at the 32°S section. Figure 11 shows the mode water at the western end of the section being ventilated over a more limited longitude range than that at the eastern part of the section. The mode waters are progressively cooler as one shifts east through the ventilation sites, and a larger range of mode water temperatures is observed at the eastern end of the section than at the western end. All of the ventilation occurs within the latitude range 35°–50°S, with the majority of ventilation sites in the zone 40°–45°S.

Belkin and Gordon (1996) primarily use the southern boundary of the Subantarctic Mode Water thermostad as a definition of the subantarctic front in the eastern part of the south Indian Ocean. In historic data, they observe SAMW forming at 10°–11°C at 110°E and 9°–10°C at 134°E. In the western part of the south Indian Ocean, they find thermostads forming north of the Agulhas Front at 17°–18°C at 20°E and 12°–14°C at 70°E. Therefore, the thermocline mode waters formed in the western south Indian Ocean (and observed at the western end of the 32°S section) are subtropical rather than subantarctic.

The pattern of mode water properties in the south Indian Ocean thermocline observed in Fig. 4 and described in section 4a concur with the observations used by McCartney (1982) to define this water mass. He deduced that there were two main regions of mode water formation in the Indian Ocean sector, and, consistent with the global pattern, these occurred on progressively cooler (denser) isotherms (isopycnals) toward the east. The 13°C mode water formation is centered at around 40°S, 70°E with the formation site for mode water around 10°C centered at approximately 45°S, 110°E. These mode waters are then subducted and move northward in the subtropical gyre occupying the western and eastern portions of the thermocline, respectively. The separation of the modes is further maintained by the existence of an anticyclonic subgyre in the southwest Indian Ocean, the eastern limit of which is at approximately 70°E between 30° and 40°S (Stramma and Lutjeharms 1997). This subgyre contains the 13°C mode water.

Changes on potential temperature surfaces (section 5a) imply that the maximum change occurs for only one element (11°–12°C) of the mode water in the eastern part of the section between 1987 and 2002. The isopycnal changes in the θS properties identify the maximum change at 11.6°C, the warmest mode water observed at the eastern end of the 32°S section (section 5b). According to the particle trajectory analysis in GIM, the warmest mode water observed on the eastern part of the section forms at approximately 90°E. This implies that the surface forcing change that produced the maximum upper-thermocline property change between 1987 and 2002 occurs at ∼90°E. There have been changes in the other mode water properties between 1987 and 2002 in the eastern south Indian thermocline, but the change at around 11.6°C is the largest in magnitude and crucially reverses the pre-1987 trend.

Formation of mode water in the western Indian Ocean appears, from both the studies of the historic data and the particle tracking, to be confined to a single formation site centered at approximately 70°E, 40°S. This implies that changes in forcing that cause property changes in the mode water are also confined to this formation site. In the next section, we interpret the changes in the mode water that have been observed in terms of changes in forcing in the formation region.

7. Interpreting changes in terms of buoyancy fluxes in the formation region

When one considers the changes in θS properties between 1987 and 2002, the eastern and western ends of the section look somewhat similar. The upper thermocline increases in salinity on θ surfaces, with a maximum change at the temperature of the mode water. Also, the lower thermocline freshens throughout. However, when one includes stratification into the analysis, the eastern and western ends of the section look less similar. The mode water at the western end of the section exhibits significant cooling and insignificant freshening with a significant increase in density (Table 1; section 5c). At the eastern end of the section, the mode water warms and increases in salinity with an insignificant change in density (Table 1).

Following the arguments of Rintoul and England (2002), a density-compensating change in the properties of a mode water is unlikely to be caused by air–sea fluxes, but rather by the advection of water with different θS properties but similar density into the formation site of the mode. Rintoul and England find that the predominantly isopycnal year-to-year variability in properties of the 9°C SAMW near 145°E is dominated by variations in Ekman transports. They argue that the local air–sea fluxes are unlikely to combine in such a way as to make them density compensating. Also, the air–sea flux anomalies required, particularly in the case of freshwater, are unrealistically large. Although the magnitude of the changes here are comparable to those observed by Rintoul and England (2002), the argument about them being implausibly large is less compelling because the change happens over 15 yr rather than from one year to the next as in the Rintoul and England study. However, the density-compensating nature of the changes makes it more plausible that the changes observed in the 11°–12°C mode water between 1987 and 2002 at the eastern end of the section are forced by a relative decrease in the cool, fresh surface waters of the subantarctic zone (Antarctic Surface Water driven northward by the wind into the formation site of the mode water.

At the western end of the section, the changes in the mode water properties (i.e., properties of the PV minimum) are dominated by the changes in temperature. This is consistent with the change in temperature on average pressure surfaces derived in section 5c over this longitude range (cooling and near-zero change in salinity over the pressure range occupied by the mode water). It is unlikely that the advection of a water mass into the formation region would cause no change in salinity, so it is most plausible that these water mass changes are caused by changes in air–sea fluxes, dominated by changes in the net heat flux.

8. Changes in oxygen concentration

We difference the 2002 and 1987 oxygen concentrations on potential temperature surfaces and examine the difference on potential temperature surfaces to exclude the effect of temperature on the equilibrium concentration of dissolved oxygen. This 1987–2002 difference shows an increase in oxygen concentrations through the majority of the thermocline (Fig. 12), but the maximum change lies consistently about 1°–2°C cooler than the temperature of the local mode water or PV minimum. This increase in oxygen concentrations in the upper thermocline reverses the pre-1987 trend observed by Bindoff and McDougall (2000). The lower thermocline is also observed to increase in oxygen concentration, continuing the pre-1987 trend. The maximum change in oxygen concentration of ∼10 μmol kg−1 in the upper thermocline is typically ∼30%–50% of the AOU for those water masses (Fig. 2c). AOU is the difference between the saturation concentration and the measured concentration of dissolved oxygen.

As dissolved oxygen is not a conservative tracer, the differences between two sections that do not exactly overlie one another will include a signal in addition to the decadal change. This is the case for the region east of 80°E where the 1987 and 2002 sections do not overlie. However, we do not see a change where the sections cross over and swap their meridional position at approximately 105°E, so we retain the differences east of 80°E but note that these are more uncertain than the portion where the sections exactly overlie one another.

The increase in thermocline oxygen concentrations reverses what has been considered to be a global trend in upper-ocean oxygen concentrations. In a review of changing dissolved oxygen concentrations, Joos et al. (2003) concluded that the dominant mechanism underlying dissolved oxygen inventory changes is circulation change, while biological production has a minor effect on the O2 inventory. Indeed, Bindoff and McDougall (2000) interpreted the pre-1987 decrease in upper-thermocline oxygen concentrations as a slowing down of the subtropical gyre circulation in the south Indian Ocean. Therefore, the trend that we observe may imply that the subtropical gyre circulation has sped up between 1987 and 2002. Moreover, the 1987–2002 trend in oxygen concentrations is not consistent with Joos et al.’s conclusion that the changes (specifically the decrease) in dissolved oxygen concentrations are a sign of the large-scale reorganization of the ocean circulation in response to anthropogenic forcing.

If biological production has a minor effect on the O2, inventory then it is reasonable to assume that the oxygen utilization rate (OUR) is reasonably constant as a function of potential temperature over time. The overall OUR is equal to the amount of oxygen used since the water mass was ventilated (AOU) divided by the time since that water was ventilated. We calculate the OUR from the 1995 data that included oxygen concentrations and water mass ages deduced from an analysis of the helium and tritium concentrations. The ages were calculated based on the methodology described in Jenkins (1980) using an updated solubility equilibrium equation of –18.6–0.0867T ‰ (W. J. Jenkins 2002, personal communication), where T is the temperature of the water in degrees Celsius. The structure of the ages with temperature at the western end of the section (blue symbols in Fig. 13a) is different from that at the eastern end (red symbols in Fig. 13a), particularly in the thermocline at around 9°–13°C. At the eastern end of the section, any water in that temperature range is younger than water of the same temperature at the western end of the section. Also, at the eastern end of the section, age changes less rapidly with potential temperature in the upper thermocline. The result of this is that the cooler mode waters at the eastern end of the section (∼9°–11°C) are a very similar age (∼5–7 yr) to the 13°C mode water at the western end of the section. However, when the oxygen utilization rate is calculated (AOU/age), there is no difference between the eastern and western datasets (Fig. 13b), and for the following calculations we treat these rates as a single dataset.

We estimate an OUR for each of the 1987 and 2002 bottle data using the relationship between potential temperature and OUR shown in Fig. 13b for the 1995 data. We take the potential temperature of the 1987 or 2002 bottle and assign an OUR to the 1987 and 2002 bottle data using a linear regression through the 1995 data over 2°C centered at the potential temperature of the 1987 or 2002 bottle. The green dots in Fig. 13b show all of the OURs calculated in this way. These OURs are then combined with the AOUs calculated for all of the 1987 and 2002 bottle data to derive an age (AOU/OUR) for every bottle with an oxygen concentration. These ages are binned into 1°C bins and differenced (2002 minus 1987; Fig. 14). To limit extreme values of OUR, we exclude all 1995 data shallower than 100 dbar and less than 1 yr old. We also focus our calculations for the change of age on bottle data for 1987 and 2002 that have potential temperatures of between 3° and 17°C and that are deeper than 100 dbar. The differing age structure between the western and eastern ends shown in Fig. 13a for the 1995 data is also apparent in the 1987 and 2002 derived ages (not shown here). We used the 2002 bottle data for this part of the analysis but could equally have used the CTDO data.

Negative values of Δage in Fig. 14 imply that for those temperatures water has been delivered from the ventilation site to 32°S faster in 2002 than in 1987. At the western end of the section, the maximum magnitude Δage of –2 yr is observed between 11° and 12°C where the average age is estimated to be 9–10 yr (Fig. 13a). This is 1°–2°C cooler than the local 13°C mode water. If we assume the ventilation site has not moved, the change in age implies that the subtropical gyre has sped up by 20%–25% (given ages in Fig. 13a). This overlaps with a maximum speedup of around 20% for the younger 13°C mode water. At the eastern end of the section, the Δage signal at around the temperature of the mode water is smaller (although in the same sense) as at the western end of the section with a maximum value of ∼−0.7 yr. However, as the water at a potential temperature at the eastern end of the section is consistently younger than water at the same temperature at the western end of the section, these smaller Δage values still imply a speedup of this part of the gyre of 10%–15%. The dotted lines in Figs. 14a,b show one standard deviation of uncertainty in the difference between the water mass ages. This uncertainty reflects the variability of the oxygen concentrations in each temperature bin. We could reasonably add an additional factor of two uncertainties based on the spread of OURs at any potential temperature in the thermocline (Fig. 13b).

What is the effect on mode water potential temperature if the ventilation rate is increased by 20%? If the volume throughput in the ventilation region is increased but the net heat flux is unchanged, then the effect of the heat flux on this larger volume of water will be 20% less. Rintoul and England (2002) suggest that the annual average heat flux is a 15 W m−2 gain by the ocean south of Tasmania based on the Southampton Oceanography Centre (SOC) climatology (Josey et al. 1998). We know from our calculations in section 5c that a reduction of 3 W m−2 (20% of 15 W m−2) in the heat flux to the mode water effects a temperature decrease of 0.2°C. Therefore, the temperature changes observed in the mode water in Fig. 10a can entirely be accounted for by a 20% increase in the speed of the gyre. However, the coincidence of these numbers is likely an accident. The annual average heating in the ventilation region that serves the western end of our section (where the mode water is up to 2°C warmer than farther east) may be larger than that estimated by Rintoul and England. Therefore, to sustain the 20% increase in ventilation rate, with only a 0.2°C cooling, might actually require a net increase in the amount of heat gained by the ocean.

9. Summary and conclusions

We observe a significant increase in the salinity of the upper thermocline θ–S properties along 32°S in the Indian Ocean between 1987 and 2002. This reversal of the pre-1987 freshening of the upper thermocline occurs across the whole section. This is evident regardless of which of the methods described in section 5 is used to estimate the changes in θS properties from 1987 to 2002. Changes in the pre-1987 thermocline are observed to occur in the same sequence along the whole section: the upper thermocline changes occurred primarily between 1965 and 1987, and the lower thermocline changes occurred between 1936 and 1965.

The thermocline at the western end of the section is occupied by a single variety of mode water with a potential temperature of around 13°C that is ventilated at a site near 40°S, 70°E. This mode water forms north of the Agulhas front so it is strictly subtropical rather than subantarctic in origin. The heat and freshwater contents of the water column show that the depth range associated with the mode water exhibits a significant temperature decrease of –0.2°C between 1987 and 2002, with little or no salinity change, which makes it plausible that the variability in this mode is forced by variations in air–sea heat flux in the ventilation region. This temperature decrease could be forced by a 3 W m−2 reduction in the ocean heat gain from the atmosphere at the ventilation site. It would almost certainly be impossible to measure a 3 W m−2 change in surface heat flux, and that emphasizes the sensitivity of subsurface property changes to surface forcing.

The thermocline at the eastern end of the 32°S section is occupied by mode water with a range of properties that cool from ∼11°C at 80°E to ∼9°C near the Australian coast. These mode waters form over a range of ventilation sites, with the coolest of these waters forming farthest east. The change in θS between 1987 and 2002 is a maximum at 11.6°C. This change is zonally coherent east of 80°E. Therefore, the reversal in salinity change is dominated by the variability of the warmest mode water on the eastern part of the section thought to ventilate near 90°E.

The post-1987 increase in salinity in the upper thermocline east of 80°E occurs primarily between 1987 and 1995 (as noted by Talley and Baringer 1997) with insignificant changes in the θS properties between 1995 and 2002. As the mode water ages imply that all of the mode water on the section is about the same age, then it seems that the forcing of the variability of the mode water at the eastern end of the section is decoupled from that at the western end of the section where upper-thermocline changes occur consistently from 1987 to 1995 to 2002. Density-compensating variation in the mode water properties between 1987 and 2002 at the eastern end of the section and the likelihood that these are caused by changes in equatorward Ekman flux are further evidence for decoupled forcing of mode water variability at the eastern and western ends of the section.

Ages derived from helium and tritium data imply that the mode waters all along the section take about the same time to reach the 32°S section from their respective ventilation sites. Increases in dissolved oxygen concentration from 1987 to 2002 imply that the mode waters all along the section are reaching the section ∼20% faster in 2002 than in 1987. This speeding up of the subtropical gyre reverses the pre-1987 trend observed by Bindoff and McDougall (2000) and is contrary to the implication that the pre-1987 changes in oxygen concentrations imply a large-scale reorganization of the ocean circulation in response to anthropogenic forcing (Joos et al. 2003).

Acknowledgments

We thank Lisa Woolgar for her help with the helium–tritium age analysis. The optimal interpolation routines were kindly provided by Lynne Talley. Support was provided by the Natural Environment Research Council under Grant NER/A/S/2000/00438 (E. L. M. and H. L. B.) and under the core strategic research projects Ocean Variability and Climate (B. A. K., S. A. C, and R. M.) and Biophysical Interactions and Controls over Export Production (R. J. S). Comments by two anonymous reviewers helped clarify the analysis and conclusions.

REFERENCES

  • Banks, H., , and R. Wood, 2002: Where to look for anthropogenic climate change in the ocean. J. Climate, 15 , 879891.

  • Banks, H. T., , and N. L. Bindoff, 2003: Comparison of observed temperature and salinity changes in the Indo-Pacific with results from the coupled climate model HadCM3: Processes and mechanisms. J. Climate, 16 , 156166.

    • Search Google Scholar
    • Export Citation
  • Banks, H. T., , R. A. Wood, , T. C. Johns, , and G. S. Jones, 2000: Are observed decadal changes in intermediate water a signature of anthropogenic climate change? Geophys. Res. Lett., 27 , 29612964.

    • Search Google Scholar
    • Export Citation
  • Banks, H. T., , R. Wood, , and J. Gregory, 2002: Changes to Indian Ocean subantarctic mode water in a coupled climate model as CO2 forcing increases. J. Phys. Oceanogr., 32 , 28162827.

    • Search Google Scholar
    • Export Citation
  • Belkin, I. M., , and A. L. Gordon, 1996: Southern Ocean fronts from the Greenwich meridian to Tasmania. J. Geophys. Res., 101 , 36753696.

    • Search Google Scholar
    • Export Citation
  • Bindoff, N. L., , and T. J. McDougall, 2000: Decadal changes along an Indian Ocean section at 32°S and their interpretation. J. Phys. Oceanogr., 30 , 12071222.

    • Search Google Scholar
    • Export Citation
  • Bryden, H. L., , E. L. McDonagh, , and B. A. King, 2003a: Changes in ocean water mass properties: Oscillations or trends? Science, 300 , 20862088.

    • Search Google Scholar
    • Export Citation
  • Bryden, H. L., and Coauthors, 2003b: RRS Charles Darwin Cruise 139, 01 Mar–15 Apr 2002; Trans-Indian hydrographic section across 32°S. Southampton Oceanography Centre Cruise Rep. 45, 122 pp.

  • Bryden, H. L., , L. M. Beal, , and L. M. Duncan, 2005: Structure and transport of the Agulhas Current and its temporal variability. J. Oceanogr., in press.

    • Search Google Scholar
    • Export Citation
  • Deacon, G. E. R., 1937: The hydrology of the Southern Ocean. Discovery Rep. 15, 124 pp.

  • Domingues, C. M., , S. Wijffels, , M. Tomczak, , and J. A. Church, 1999: Volume transports and structure of the Leeuwin Current at 22°S—WOCE ICM6. International WOCE Newsletter, No. 37, WOCE International Project Office, Southampton, United Kingdom, 36–40.

  • Donohue, K., , and J. Toole, 2003: A near-synoptic survey of the Southwest Indian Ocean. Deep-Sea Res., 50B , 18931931.

  • Gordon, A. L., , J. R. E. Lutjeharms, , and M. L. Grundlingh, 1987: Stratification and circulation at the Agulhas Retroflection. Deep-Sea Res., 34 , 565599.

    • Search Google Scholar
    • Export Citation
  • Jenkins, W. J., 1980: Tritium and 3He in the Sargasso Sea. J. Mar. Res., 38 , 533569.

  • Joos, F., , G-K. Plattner, , T. F. Stocker, , A. Körtzinger, , and D. W. R. Wallace, 2003: Trends in marine dissolved oxygen: Implications for ocean circulation changes and the carbon budget. Eos, Trans. Amer. Geophys. Union, 84 , 197201.

    • Search Google Scholar
    • Export Citation
  • Josey, S. A., , E. C. Kent, , and P. K. Taylor, 1998: The Southampton Oceanography Centre (SOC) Ocean–Atmosphere Heat, Momentum, and Freshwater Flux Atlas. Southampton Oceanography Centre Rep. 6, 30 pp.

  • Marsh, R., , A. J. G. Nurser, , A. P. Megann, , and A. L. New, 2000: Water mass transformation in the Southern Ocean of a global isopycnal coordinate GCM. J. Phys. Oceanogr., 30 , 10131045.

    • Search Google Scholar
    • Export Citation
  • McCartney, M. S., 1977: Subantarctic Mode Water. Deep-Sea Res., 24 , (Suppl.). 103119.

  • McCartney, M. S., 1982: The subtropical recirculation of Mode Waters. J. Mar. Res., 40 , (Suppl.). 427464.

  • Rintoul, S. R., , and M. H. England, 2002: Ekman transport dominates local air–sea fluxes in driving variability of Subantarctic Mode Water. J. Phys. Oceanogr., 32 , 13081321.

    • Search Google Scholar
    • Export Citation
  • Roemmich, D., 1983: Optimal estimation of hydrographic station data and derived fields. J. Phys. Oceanogr., 13 , 15441549.

  • Stramma, L., , and J. R. E. Lutjeharms, 1997: The flow field of the subtropical gyre of the South Indian Ocean. J. Geophys. Res., 102 , 55135530.

    • Search Google Scholar
    • Export Citation
  • Talley, L. D., , and M. O. Baringer, 1997: Preliminary results from WOCE hydrographic sections at 80°E and 32°S in the Central Indian Ocean. Geophys. Res. Lett., 24 , 27892792.

    • Search Google Scholar
    • Export Citation
  • Toole, J. M., , and M. E. Raymer, 1985: Heat and fresh water budgets of the Indian Ocean—Revisited. Deep-Sea Res., 32 , 917928.

  • Toole, J. M., , and B. A. Warren, 1993: A hydrographic section across the subtropical South Indian Ocean. Deep-Sea Res., 40A , 19732019.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Positions of the observations used in this study. CTD stations from RRS Charles Darwin occupation (Mar 2002, dark blue circles), R/V Knorr occupation (Jun 1995, red triangles; Mar–Apr 1995, red upside-down triangles), and Darwin occupation (Nov 1987, green stars). Bottle stations from R/V Atlantis II occupation (Jul 1965, black diamonds) and RRS Discovery occupation (Apr 1936, light blue triangles). Depth contours of 0, 2000, and 4000 m are shown, and depths shallower than 2000 m are shaded.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 2.
Fig. 2.

(a) The θ, (b) S, and (c) AOU sections for 2002 data. AOU is the difference between the saturated and observed dissolved oxygen concentrations.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 3.
Fig. 3.

(a) Potential vorticity contoured at 2.5, 5.0, 7.5, and 10 × 10−11 (m s)−1 for the 20-dbar CTD data from the 2002 occupation of the section: PV = |f|/ρ.dρ/dz, where f is the Coriolis parameter, ρ is density, and z is depth. Shaded areas represent all PV values less than 5.0 × 10−11 (m s)−1. (b) The gray contours are the 9°, 10°, 11°, 12°, and 13°C potential temperature isotherms. The bold black line in (a) and (b) shows the position of the PV minimum between 200 and 800 dbar at each station.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 4.
Fig. 4.

(a) The θ, (b) S, (c) PV, and (d) percentage of oxygen saturation at the PV minimum between 200 and 800 dbar for each 2002 (*) and 1987 (ˆ) station. Only data between 200 and 800 dbar are analyzed to ensure that the identified PV minimum is associated with SAMW. Stations shallower than 700 dbar are omitted as these did not necessarily sample the PV minimum associated with SAMW.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 5.
Fig. 5.

The (θS) plots for the thermocline at (a) the western and (b) the eastern end of the sections. The lines represent the average (on neutral density surfaces) θ and S properties for the 1987 (green), 1995 (red), and 2002 (dark blue) occupations. All of the 1936 (light blue triangles) and 1965 (diamonds) bottle data that fall into the respective longitude ranges are plotted. The contours are lines of constant neutral density.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 6.
Fig. 6.

Salinity changes from 1987 to 2002 (2002 minus 1987). Vertical salinity profiles are interpolated onto potential temperature levels in tenths of a degree and then horizontally onto a half-degree longitude grid. No smoothing is necessary. The salinity changes are contoured with an interval of 0.02 (Salinity has no dimensions, but one unit of practical salinity is essentially equivalent to one part per thousand of dissolved salt.) Shading denotes increasing salinity on temperature levels; no shading denotes freshening. The bold contour indicates no change. The stars indicate the potential temperature of the PV minimum associated with SAMW in 2002 as shown in Fig. 4a. The temperature range from 18°C to 4°C is approximately equivalent to depths of 50 to 1200 m.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 7.
Fig. 7.

Changes between 1987 and 2002 θS curves on neutral density surfaces (gray lines) and for the minimum change (black lines). Dashed lines are one std dev of uncertainty on the difference. Changes are 2002 minus 1987, so a positive change in θ (S) represents a warming (increase in salinity). (a) Changes in θ (dθ) and (b) changes in S (dS) are shown for the western part of the section (40°–70°E) in (a) and (c), (d) for the eastern end of the section (80°E to Australia). Values of neutral density are shown on the right-hand side of each of the plots.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 8.
Fig. 8.

Minimum distance changes for the western end of the section between 40° and 70°E. Differences between 1987 and 2002 (red), 1987 and 1995 (green), and 1995 and 2002 (dark blue) θS curves. Dashed lines are one standard deviation uncertainty on the difference. Differences between 1987 θS curve and 1965 bottle data (black diamonds) and 1987 θS curve and 1936 bottle data (light blue triangles). Horizontal lines represent one std dev of uncertainty on the difference. Changes are for the later year minus the earlier year, so a positive change in θ (S) represents a warming (increase in salinity) over that time period: (a) changes in θ () and (b) changes in S (dS). Values of neutral density are shown on the right-hand side of each of the plots.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 9.
Fig. 9.

Same as in Fig. 8, but for isopycnal changes (on neutral density surfaces) for the eastern end of the section between 80°E and Australia.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 10.
Fig. 10.

Differences of (a) potential temperature () and (b) salinity (dS) between 1987 and 2002 on average pressure surfaces between 40° and 70°E. Differences are 2002 minus 1987, so a positive change in θ (S) represents a warming (increase in salinity). The dotted lines represent the differences between the thermoclines if they are offset by +20 and –20 dbar.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 11.
Fig. 11.

Colored blocks indicate the position of the ventilation sites in a general circulation model for thermocline mode water on portions of the 32°S section indicated by the solid black line (a) between 40° and 70°E and (b) east of 70°E. The color coding represents the temperature of the mode water at the section.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 12.
Fig. 12.

Dissolved oxygen changes from 1987 to 2002 (2002 minus 1987) in μmol kg−1. Vertical dissolved oxygen profiles are interpolated onto potential temperature levels in tenths of a degree and then horizontally onto a half-degree longitude grid. The oxygen changes are contoured with an interval of 5 μmol kg−1. Shading denotes increasing oxygen on temperature levels; no shading denotes decreasing oxygen concentrations. The bold contour indicates no change. The stars indicate the potential temperature of the PV minimum associated with SAMW in 2002 as shown in Fig. 4a. The temperature range from 18°C to 4°C is approximately equivalent to depths of 50 to 1200 m.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 13.
Fig. 13.

(a) Age since thermocline waters were most recently in contact with the sea surface vs potential temperature for individual bottles. Age is calculated from helium and tritium data collected on the 1995 WOCE cruises. Nonlinear mixing means that this He–Tr age is a lower bound to the actual age spectrum. I5W data (shown in blue) were collected between 40° and 50°E; I5E data (shown in red) were collected east of 85°E. (b) OUR calculated from the AOU (difference between saturated and measured oxygen concentration) divided by the He–Tr age. I5W data (shown in blue) were collected between 40° and 50°E; I5E data (shown in red) were collected east of 85°E. The green dots are OUR calculated for the potential temperature of each of the 1987 and 2002 bottle data between 3° and 17°C and deeper than 100 dbar. These OUR values are calculated using a linear regression of the 1995 OURs over 2°C, centered at the potential temperature of the bottle.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Fig. 14.
Fig. 14.

Changes in age (since the water was last ventilated) between 2002 and 1987, averaged in 1°C potential temperature bins. Age is derived from the OUR–potential temperature relationship for 1995 data and AOU for 1987 and 2002 data. (a) At the western end of the section between 40° and 70°E and (b) at the eastern end of the section, east of 80°E. The dotted lines show one std dev of uncertainty in the difference between the water mass ages. This uncertainty reflects the variability of the oxygen concentrations in each temperature bin.

Citation: Journal of Climate 18, 10; 10.1175/JCLI3350.1

Table 1.

Changes in the potential temperature (Δθ), salinity (ΔS), and neutral density (Δγn) of the PV minimum between 1987 and 2002. Changes are 2002 minus 1987, so a positive (negative) change represents an increase (decrease) in the property. Changes are calculated over two longitude bands: 40°–70°E and 80°–114°E. The potential temperature of the PV minimum is ∼13°C between 40° and 70°E and reduces from ∼11°C at 80°E to 9°C at 114°E. One std dev of uncertainty is also given. Only data between 200 and 800 dbar were analyzed to ensure that the PV minimum is associated with SAMW. Stations shallower than 700 dbar are omitted as these might not sample the PV minimum associated with SAMW.

Table 1.
Table 2.

Minimum distance changes between occupations on the western segment of the section between 40° and 70°E.

Table 2.
Table 3.

Isopycnal (on a neutral density surface) changes between occupations on the eastern segment of the section between 80°E and Australia.

Table 3.
Save