Inter-annual variability of the carbon dioxide oceanic sink south of Tasmania

We compiled a large data-set from 22 cruises spanning from 1991 to 2003, of the partial pressure of CO 2 (pCO 2 ) in surface waters over the continental shelf (CS) and adjacent open ocean (43 ◦ to 46 ◦ S; 145 ◦ to 150 ◦ E), south of Tasmania. Sea surface temperature (SST) anomalies (as intense as 2 ◦ C) are apparent in the subtropical zone 5 (STZ) and subAntarctic zone (SAZ). These SST anomalies also occur on the CS, and seem to be related to large-scale coupled atmosphere-ocean oscillations. Anomalies of pCO 2 normalized to a constant temperature are negatively related to SST anomalies. A depressed winter-time vertical input of dissolved inorganic carbon (DIC) during phases of positive SST anomalies, related to a poleward shift of westerly winds, and 10 a concomitant local decrease in wind stress are the likely cause of the negative relationship between pCO 2 and SST anomalies. The observed trend is an increase of the sink for atmospheric CO 2 associated with positive SST anomalies, although strongly modulated by inter-annual variability of wind speed. Assuming that phases of positive SST anomalies are indicative of the future evolution of regional ocean biogeochem- 15 istry under global warming, we show using a purely observational based approach that some provinces of the Southern Ocean could provide a potential negative feedback on increasing atmospheric CO 2 .

a concomitant local decrease in wind stress are the likely cause of the negative relationship between pCO 2 and SST anomalies. The observed trend is an increase of the sink for atmospheric CO 2 associated with positive SST anomalies, although strongly modulated by inter-annual variability of wind speed. Assuming that phases of positive SST anomalies are indicative of the future evolution of regional ocean biogeochemistry under global warming, we show using a purely observational based approach that some provinces of the Southern Ocean could provide a potential negative feedback on increasing atmospheric CO 2 .

Introduction
The ocean is a major and dynamic sink for anthropogenic CO 2 (e.g. Sabine et al. 2004) 20 playing an important role in the mitigation of climate change. The inclusion in climate models of potential feedbacks of air-sea CO 2 fluxes on the increase of atmospheric CO 2 is required to improve the reliability of the predictions of the future evolution of the global carbon cycle and climate change. The collection of partial pressure of CO 2 (pCO 2 ) data for surface waters during the last 30 years has allowed the investigation backs in the carbon cycle related to changes in oceanic physical and biogeochemical processes.
The investigation of long term trends in surface water CO 2 requires the description of the inter-annual variability of pCO 2 . The main drivers of the inter-annual variability of surface pCO 2 described to date are large-scale atmosphere-ocean coupled climate 10 oscillations including the El Niño Southern Oscillation (ENSO) for the equatorial and subtropical Pacific Ocean (Feely et al., 2002;Dore et al., 2003;Brix et al., 2005), and the North Atlantic Oscillation for the North Atlantic Ocean (Gruber et al., 2002). In the Southern Ocean, atmosphere-ocean coupled climate oscillations such as the Southern Annular Mode (SAM) (Wetzel et al. 2005;Le Quéré et al., 2007;Lenton 15 and Matear, 2007;Lovenduski et al., 2007) and ENSO (Verdy et al., 2007) have been identified using biogeochemical ocean general circulation models as major drivers of inter-annual variability of pCO 2 . These atmosphere-ocean coupled-climate oscillations can alter mixed layer dynamics with an effect on biogeochemical cycles (Le Quéré et al., 2000, 2003Lovenduski and Gruber, 2005). The detection of changes in bio- 20 geochemical cycling in response to inter-annual changes in mixed layer properties can provide insights into the marine biogeochemical response to future climate changes, namely increases in sea surface temperature (SST) and stratification that are predicted by coupled climate models (e.g. Le Quéré et al., 2000Quéré et al., , 2002Quéré et al., , 2003. This has been addressed through modelling (e.g. Le Quéré et al., 2000Quéré et al., , 2003Quéré et al., , 2007Lenton and Matear, Due to the relative scarcity of pCO 2 field data in the Southern Ocean, the interannual variability of air-sea CO 2 fluxes has seldom been investigated from a purely observational based approach. The impact of warm anomalies on air-sea CO 2 fluxes in the Southern Ocean has to some extent been investigated using cruise-to-cruise comparisons (Jabaud-Jan et al., 2005;Brévière et al., 2006). These two studies show 10 warm anomalies lead to significant, but opposing, effects on air-sea CO 2 fluxes in different regions of the high latitude Southern Ocean. This could be due to the spatially heterogeneous control of export production through either light or nutrient limitation in the Southern Ocean (Le Quéré et al., 2000, 2003, with some modulation by thermodynamic effects of SST change on pCO 2 . 15 The aim of the present work is to investigate inter-annual variations of pCO 2 in the surface waters of the continental shelf ( Table 1). We analyzed how surface pCO 2 and air-sea CO 2 exchange varied in response to warm and cool anomalies, and discuss how this might be indicative of the future feedback of surface water warming on air-sea CO 2 exchange in the sub-polar region of the Southern Ocean. Introduction in the North Atlantic (Körtzinger et al., 2000), and the results showed that the pCO 2 data were consistent within ±1 µatm. ULg and LOCEAN/IPSL systems were intercalibrated during the OISO3 cruise (21 December-28 December 1998) in the central Indian sector of the Southern Ocean, and the results showed that the pCO 2 data were consistent within ±6 µatm. ULg and CSIRO systems were inter-calibrated during the 10 AA0301 cruise (11 October-27 October 2003) in eastern Indian sector of the Southern Ocean, and the results showed that the pCO 2 data were consistent within ±5 µatm.
Since 82% of the data were obtained by one single group (CSIRO ; Table 1), we assume the uncertainty of the whole data-set to be better than ±3 µatm. All data were converted to pCO 2 in wet air at 1 atm. During the time-span of the data-set (from 1991 to 2003), 15 atmospheric pCO 2 increased by 20 µatm (1.7 µatm yr −1 ), so data were referenced to 1997, the middle of the time series, according to: where pCO 2sea1997 is the pCO 2 in seawater referenced to 1997, pCO 2sea199i is the pCO 2 in seawater from a given year, pCO 2air1997 is the atmospheric pCO 2 in 1997, and 20 pCO 2ai r199i is the atmospheric pCO 2 for the same given year. Atmospheric pCO 2 data from the Cape Grim station (40.7 • S, 144.7 • E; Tasmania) were obtained from the Cooperative Air Sampling Network of the National Oceanic and Atmospheric Administration/Earth System Research Laboratory/Global Monitoring Division (http://www.cmdl.noaa.gov/). Hereafter, pCO 2 refers to pCO 2sea1997 . 25 Data were collocated with bathymetry based on the Smith and Sandwell (1997)  EGU and averaged by sorting for depths <300 m, and data at depths >1000 m were considered as open ocean. The sub-Tropical Front (STF) was identified from gradients of sea surface salinity (SSS) and SST (e.g. Belkin and Gordon 1996). The STF separates warmer and saltier sub-tropical zone (STZ) waters from cooler and fresher sub-Antarctic zone (SAZ) waters.

3 Results and discussion
3.1 Climatological SST and pCO 2 seasonal cycles Climatological seasonal cycles of SST, pCO 2 and pCO 2 normalized to a temperature of 14 • C (pCO 2 @14 • C, using the algorithms of Copin-Montégut, 1988, 1989 were obtained by fitting monthly averages with a wave function in the form of: where y is either SST, pCO 2 or pCO 2 @14using the algorithms ofC, x is time (julian days) and a, b, c, and d are fitted constants. In the CS, STZ and SAZ, under-saturation of CO 2 is observed throughout the year (Figs. 2,3,4), showing these regions are perennial sinks for atmospheric CO 2 . In 15 the 3 regions, similar climatological pCO 2 and pCO 2 @14 • C seasonal trends are observed in timing and amplitude: values decrease from late September to late February (austral spring-summer) as net biological uptake removes dissolved inorganic carbon (DIC) from surface waters. From March to September (austral fall-winter), pCO 2 and pCO 2 @14 • C values increase in relation to destratification and mixing of surface wa-20 ters with DIC rich deeper waters (Goyet et al., 1991;Poisson et al., 1993;Metzl et al., 1991Metzl et al., , 1995Metzl et al., , 1998Metzl et al., , 1999. The amplitude of the seasonal cycle of pCO 2 was lower than the one of pCO 2 @14 • C because warming of surface waters during spring and summer leads to a thermodynamic increase of pCO 2 , that opposes a decrease due to net biological carbon uptake. The thermodynamic effect of SST change on the seasonal Introduction EGU amplitude of pCO 2 is similar in the CS (51 µatm), the STZ (50 µatm), and the SAZ (47 µatm), because SST amplitude is similar in the 3 regions (3.7 • C, 3.6 • C and 3.3 • C, respectively).
The SSS values for all three water masses do not show a distinct seasonal signal like SST, which is strongly influenced by seasonal heating and cooling of surface waters.

5
The water mass on the Tasmanian CS is a mixture of STZ and SAZ waters (Harris et al., 1987(Harris et al., , 1991. This is apparent in our data-set, as the average SSS value in the CS (35.02±0.17) lies between those of the STZ (35.19±0.15) and the SAZ (34.73±0.15). The SSS values in the CS show more scatter than in the other two regions (Figs. 2,3,4), suggesting a variable degree of mixing between the STZ and SAZ water masses.

10
For a two end-member mixing model, the water mass on the CS is on average composed of 64% STZ water, consistent the annual average of SST in the CS (13.8±1.3 • C) being similar to the STZ (13.4±1.3 • C) and distinctly different from the SAZ (11.5±1.2 • C) annual averages.
3.2 Inter-annual SST and pCO 2 variations 15 Monthly anomalies of SST, pCO 2 , and pCO 2 @14 • C were computed as the difference between observations and averaged monthly values for all the cruise data. SST anomalies of up to 2 • C are observed in the CS, STZ and SAZ, and similar trends of pCO 2 @14 • C are observed in the three regions (Figs. 2,3,4). SST values that are below the climatology tend to coincide with less saline waters suggesting a greater con-20 tribution of SAZ waters. Warm anomalies are typically associated with saltier waters implying a greater contribution of subtropical waters and consistent with observations of increases in warming and salinity for the region (Rintoul and Sokolov, 2001;Rintoul and Trull, 2001;Morrow et al., 2007 1 ).
A comparison of the monthly SST anomalies with anomalies for pCO 2 and BGD 4,2007 Inter-annual variability of the CO 2 sink south of Tasmania EGU pCO 2 @14 • C show consistent trends. Values of pCO 2 @14 • C associated with positive SST anomalies (>0.5 • C) generally lie below the climatological pCO 2 @14 • C cycle. Values associated with significant negative SST anomalies (<0.5 • C) generally lie above the climatological pCO 2 @14 • C cycle (Figs. 2,3,4). The trends for pCO 2 anomalies are not as clear. In the STZ, positive pCO 2 and pCO 2 @14 • C anomalies correspond 5 to negative SST anomalies during spring (October-December). However, during the summer (January-March) the pCO 2 anomalies in the STZ show the opposite behavior of pCO 2 @14 • C. For summer, the pCO 2 values associated with positive SST anomalies generally occur above the climatological pCO 2 cycle, and values associated with significant negative SST anomalies tend to lie below the climatological pCO 2 cycle.

10
The SST anomalies in the STZ, SAZ and CS are roughly consistent in timing and amplitude (Fig. 5), suggesting a similar driver of these anomalies in the three water masses. Large-scale coupled atmosphere-ocean oscillations likely to drive such large SST anomalies simultaneously in the mid-latitude band of the Southern Ocean are ENSO through atmospheric bridges (Li, 2000;Verdy et al., 2006;Morrow et al., 2007 1 ), 15 the subtropical dipole pattern (Behera andYamagata, 2001), or SAM (Hall andVisbeck, 2002;Lovenduski and Gruber, 2005;Morrow et al., 2007 1 ). SAM is the principal mode of atmospheric forcing and climate variability in the Southern Ocean with potential to significantly impact on biogeochemical carbon cycling and air-sea CO 2 fluxes (Le Quéré et al., 2007;Lenton and Matear, 2007;Lovenduski et al., 20 2007;Verdy et al., 2007). An increase in the SAM index leads to synchronous SST anomalies in the STZ and SAZ (Lovenduski and Gruber 2005;Lovenduski et al. 2007), as we observed in our dataset (Fig. 5). The increased SAM index is associated with a poleward shift of westerly winds with a local decrease in wind stress (Hartmann and Lo, 1998;Thompson and Wallace, 2000). This causes an increase in Ekman convergence 25 in the SAZ, with deepening isopycnals, surface warming in the mid-latitudes, and a southward expansion of subtropical waters (Cai et al., 2005;Roemmich et al., 2007). Near Tasmania, surface warming and an increased outflow of warm and salty subtropical waters from the South Tasman Sea have been observed and linked to a polewards BGD 4,2007 Inter-annual variability of the CO 2 sink south of Tasmania EGU shift in the wind stress over the Southern Ocean (Rintoul and Sokolov, 2001;Morrow et al., 2007 1 ;Ridgway, 2007).
A reduction in available nutrients and reduced DIC input through either the expansion of subtropical waters or less vertical mixing, could explain the observed relationship between pCO 2 @14 • C and SST anomalies for all 3 regions (Figs. 6, 7). Surface waters of the Tasman Sea have relatively low macronutrient concentrations (Condie and Dunn 2006) and low pCO 2 values (Takahashi et al. 2002), compared to subAntarctic waters to the south of the STF. The surface warming and decreased wind stress associated with a positive SAM index will also produce a deepening of isopycnal surfaces (Roemmich et al., 2007) and result in more stratified surface mixed layers (Chaigneau et al., 2004), thus reducing the potential to entrain subsurface waters with high nutrient and DIC concentrations into the surface mixed layer over winter (Le Quéré et al., 2000, 2003. The net effect of altered vertical mixing and the southerly expansion of the subtropical waters should be to produce warm SST anomalies and low pCO 2 @14 • C values in the mid-latitude waters. Conversely, during periods when the SAM index 15 is reduced, there is likely to be a greater supply of nutrients and DIC to the surface layer through vertical mixing and a retreat of subtropical waters to the north. Under these conditions the surface waters are expected to contain a greater component of relatively high nutrient and DIC waters of the SAZ, producing cool SST anomalies and high pCO 2 @14 • C anomalies. 20 The response of the region to the SAM could also explain different slopes between pCO 2 @14 • C anomalies and SST anomalies observed during spring-summer and fallwinter periods (Figs. 6, 7). Negative SST anomalies would be associated with enhanced nutrient and DIC inputs. Primary production in the region is partly limited by nutrient availability (Boyd et al., 2001). The greater nutrient availability associated with 25 negative SST anomalies would lead to enhanced primary production during spring and summer, and a decrease of the slope (becomes less negative) between pCO 2 @14 • C and SST anomalies for the fall-winter to spring-summer periods (Fig. 7). The overall negative relationship between pCO 2 @14 • C anomalies and SST anomalies ( Fig. 6) BGD 4,2007 Inter-annual variability of the CO 2 sink south of Tasmania

EGU
shows that the enhanced primary production associated with negative SST anomalies does not overcome the enhanced winter inputs of DIC. The pCO 2 anomalies in the 3 regions show more scatter than pCO 2 @14 • C anomalies versus SST anomalies (Fig. 6). This is due to the thermodynamic effect of temperature change that leads to a decrease of the positive pCO 2 anomalies associated with 5 negative SST anomalies, and conversely to an increase of negative pCO 2 anomalies associated with positive SST anomalies (Fig. 7). In some extreme cases there is a reversal of the direction of the pCO 2 anomalies, particularly in the STZ (Fig. 6).

Inter-annual air-sea CO 2 flux variations
Air-sea CO 2 fluxes were computed according to: where F is the air-sea CO 2 flux, k is the gas transfer velocity, α is the CO 2 solubility coefficient, and ∆pCO 2 is the air-sea pCO 2 gradient. We used the k-wind parameterization of Wanninkhof (1992), and the Weiss (1974) formulation of α as a function of SSS and SST. Atmospheric pCO 2 data from 1997 were expressed in wet air using the Inter-annual variability of the CO 2 sink south of Tasmania

EGU
Reyn SmithOIv2 SST grid nodes (n • 2 to n • 11, Fig. 1) were averaged, and used to compute monthly SST anomalies for the STZ and SAZ. This assumes that SST anomalies in the STZ and SAZ are synchronous and similar in amplitude and direction as shown in Fig. 5. Also, the position of the STF is highly variable in time (Sokolov and Rintoul, 2002) and it is not possible to arbitrarily associate the Reyn SmithOIv2 SST grid nodes 5 to either STZ or SAZ water masses. The pCO 2 @14 • C anomalies were computed from the Reyn SmithOIv2 monthly SST anomalies using the linear relationships shown in Fig. 6. These pCO 2 @14 • C anomalies were added to the climatological pCO 2 @14 • C cycles shown in Figs. 2, 3 and 4. The Reyn SmithOIv2 monthly SST anomalies were added to the climatological SST cycles 10 shown in Figs. 2, 3 and 4. The pCO 2 values were then computed from pCO 2 @14 • C and SST (both including the respective anomalies), using the algorithms of Copin-Montégut (1988;1989). The daily air-sea CO 2 fluxes were calculated using Eq. (3), and the NCEP u 10 data.
For the whole 1982-2005 period, the CS, STZ and SAZ act annually as 15 sinks for atmospheric CO 2 at the rate of, respectively, −6.4±0.7, −6.8±0.5, and −5.7±0.5 mmol m −2 d −1 . Significant potential inter-annual variability of annual F is apparent in the 3 regions (Fig. 8). Consistent with pCO 2 , negative annual F anomalies (stronger sink for atmospheric CO 2 ) are associated with positive annual SST anomalies, and conversely positive annual F anomalies (weaker sink for atmospheric CO 2 ) Introduction  (Table 2) lead to a decrease of the 1982-2005 average annual sink for atmospheric CO 2 of 4, 3, and 18%, for the CS, STZ and SAZ, respectively. The annual F anomalies computed from the 1982-2005 mean of daily u 10 data show a consistent negative relationship with the annual SST anomalies for all three regions (Fig. 8). This suggests the scatter in the annual F anomalies computed from daily 5 u 10 data versus the annual SST anomalies is to a large extent related to inter-annual variability in u 10 . In particular, the annual F anomalies in 1985, 1989, 1999 and 2001 are larger compared to other years for the 3 regions, leading to positive annual F anomalies in the CS and the STZ during 1985STZ during , 1989STZ during and 1999, and to virtually neutral annual F anomalies in the SAZ during 1985 and1989. 10 The F anomalies computed for the fall-winter and spring-summer periods using the 1982-2005 mean of daily u 10 ( Fig. 9) show that the annual F anomalies (Fig. 8) are mainly driven by the fluxes during the fall-winter period. This is related to the thermodynamic effect of temperature change that tends reduce the magnitude of pCO 2 anomalies (Fig. 7), and due to SST anomalies being more marked during the periods 15 of stratification (spring-summer, Figs. 2, 3, 4). Higher wind speeds during the fall-winter period (not shown) are also responsible for the larger contribution of F anomalies during this period. These trends are reflected in the F anomalies computed for the fallwinter and spring-summer periods using the daily u 10 , but with more scatter due to inter-annual variability of u 10 (Fig. 9). The small annual F anomalies in 1985, 1989, 20 1999and 2001) are due to the spring-summer F anomalies (Fig. 9). Figure 10 shows that average u 10 during the spring-summer period of 1985, 1989, 1999 and 2001 were lower than other years. This is most likely related to a poleward shift of westerly winds related to changes in the SAM which causes a local decrease in wind stress and positive SST anomalies (Hartmann and Lo, 1998) and reduced air-sea fluxes of CO 2 .  Quéré et al., 2002Quéré et al., , 2003. Based on correlations of remotely sensed chlorophyll-a and SST, Le Quéré et al. (2002) suggested that an increase of SST would lead to an overall increase in primary production in the Southern Ocean. This was attributed to stronger light limitation compared to nutrient limitation for most regions of the Southern Ocean and supported by the simulation of propagating warm anomalies using a biogeochem-5 ical ocean general circulation model (Le Quéré et al., 2003). Bopp et al. (2001) have also suggested that export production in the Southern Ocean will increase with projected warming of surface waters. The comparison of pCO 2 @14 • C and SST anomalies during fall-winter and spring-summer periods (Fig. 6) suggests that in our study region an increase of SST leads to a decrease of export production (Fig. 7). This is in 10 agreement with evidence for nutrient rather than of light limitation of primary production for the sub-polar waters near Tasmania (Boyd et al., 2001).

EGU
Besides modifying export production, climate changes associated with warming of surface waters will also change air-sea CO 2 fluxes due to changes of the input of CO 2 from deeper layers. Indeed, numerical models of CO 2 dynamics and air-sea CO 2 ex-15 change in the Southern Ocean highlight the very significant role of vertical mixing at seasonal (Louanchi et al., 1996;Metzl et al., 1999Metzl et al., , 2006 and inter-annual (Louanchi and Hoppema, 2000;Verdy et al., 2007) time scales. Changes in vertical mixing in the Southern Ocean can have an impact on the intensity of the air-sea CO 2 fluxes and can potentially modulate the atmospheric CO 2 content by changing the natural biogeo-20 chemical CO 2 cycle. Indeed, to explain the decrease of atmospheric CO 2 during the last glacial period, some hypothesis are based on a decrease in the ventilation of deep DIC rich water in relation to an increase of stratification (Toggweiler, 1999;Sigman and Boyle, 2000), due to a northward shift of the westerly winds (Toggweiler et al., 2006) or due to an increase of salinity in deeper waters (Watson and Naveira Garabato, 2006). 25 The decrease of vertical inputs of DIC combined with constant or enhanced export production could have increased the biological pump (in regions of light limitation for primary production) and enhanced the atmospheric CO 2 sink (Toggweiler, 1999;Sigman and Boyle, 2000;Toggweiler et al., 2006;Watson and Naveira Garabato, 2006). 4,2007 Inter-annual variability of the CO 2 sink south of Tasmania

EGU
This clearly illustrates that a modification of the natural biogeochemical cycle of CO 2 in the Southern Ocean can lead to strong feedbacks on the atmospheric CO 2 content. Our study region is small but includes a portion of two important waters masses: the STZ and SAZ. The results do indicate that a decrease of vertical input of DIC during winter associated to climate changes associated with positive SST anomalies could 5 lead to an increase of the sink for atmospheric CO 2 , although strongly modulated by inter-annual variability in wind speed. This suggests that climate change associated to warming of surface waters of some regions of the Southern Ocean, would lead to a negative feedback on increasing atmospheric CO 2 .

10
We compiled data obtained along 40 transects during 22 cruises carried out between 1991 and 2003, on the CS and adjacent open oceanic waters south of Tasmania. This allowed us to analyze how surface pCO 2 and air-sea CO 2 exchange vary in relation to warm and cool anomalies. Strong SST anomalies up to 2 • C were observed in the STZ and the SAZ. As the waters on the CS are a mixture of STZ and SAZ, these SST 15 anomalies also propagate onto the CS. The consistency in timing and amplitude of SST anomalies in the STZ and SAZ can only be attributed to a large scale coupled atmosphere-ocean oscillation.
Overall, positive SST anomalies are associated with negative pCO 2 @14 • C anomalies, and negative SST anomalies with positive pCO 2 @14 • C anomalies, in the CS, STZ 20 and SAZ. This seems to be related to a depressed input of DIC during the fall-winter period, during the phases of positive SST anomalies, in relation to a poleward shift of the westerly winds, and a local decrease in wind stress.
The potential effect of SST anomalies on air-sea CO 2 exchange were investigated using a 23 years consistent time series of SST and wind speed. The general trend is 25 an increase in the sink for atmospheric CO 2 associated with positive SST anomalies. This increase in the sink for atmospheric CO 2 is mainly due to the fluxes during the BGD 4, 2007 Inter-annual variability of the CO 2 sink south of Tasmania EGU fall-winter period. However, this general trend is strongly modulated by inter-annual variations of wind speed that affects the gas transfer velocity and the intensity of the air-sea CO 2 flux. Assuming that phases of positive SST anomalies are indicative of the future evolution of ocean biogeochemistry under global warming, we show based on a spatially restricted observational data-set, that some provinces of the Southern Ocean 5 could provide a potential negative feedback on increasing atmospheric CO 2 and associated climate change. The observations from our region are in agreement with recent modelling studies that show during positive phases of SAM a decrease of the CO 2 sink in high latitude areas of the Southern Ocean due to enhanced upwelling and an increase of the CO 2 sink in the low latitude areas of the Southern Ocean (Wetzel et A larger scale investigation in the Southern Ocean is required to quantify more rigorously potential feedbacks on the increase atmospheric CO 2 due to warming of surface waters. BGD 4,2007 Inter-annual variability of the CO 2 sink south of Tasmania  1986to 1994, Mar. Chem., 72(2-4), 103-114, 2000: Surface water carbon dioxide in the southwest Indian Sector of the Southern Ocean: A highly variable CO 2 source/sink region in summer, Mar. Chem., 35(1-2), 85-95, 1991. 5 Metzl, N., Poisson, A., Louanchi, F., Brunet, C., Schauer, B., and Brès, B.: Spatio-temporal distribution of air-sea fluxes of CO 2 in the Indian and Antarctic Ocean, Tellus B, 47(1-2), 56-69, 1995. Metzl, N., Louanchi, F., and Poisson, A.: Seasonal and interannual variations of sea surface carbon dioxide in the subtropical Indian Ocean, Mar. Chem., 60(1-2), 131-146, 1998. Inter-annual variability of the CO 2 sink south of Tasmania     . 6. Monthly anomalies of the partial pressure of CO 2 (pCO 2 ) and of the pCO 2 normalized to a temperature of 14 • C (pCO 2 @14 • C) plotted against the sea surface temperature (SST) monthly anomalies in the continental shelf (CS), the subtropical zone (STZ) and subAntarctic zone (SAZ) south of Tasmania, for the spring-summer period (grey squares) and the fall-winter period (black squares). Solid lines correspond to linear regression functions, and r 2 to the corresponding coefficient of determination. Symbols in brackets were excluded from the linear regressions. 4,2007 Inter-annual variability of the CO 2 sink south of Tasmania  Fig. 7. Conceptual frame relating the partial pressure of CO 2 (pCO 2 ) normalized to a temperature of 14 • C (pCO 2 @14 • C), and pCO 2 anomalies to sea surface temperature (SST) anomalies. 4,2007 Inter-annual variability of the CO 2 sink south of Tasmania    Fig. 9. Anomalies of air-sea CO 2 fluxes (F ) computed with daily wind speeds (u 10 ) and 1982-2005 mean of u 10 daily, plotted against the sea surface temperature (SST) anomalies in the continental shelf (CS), the subtropical zone (STZ) and subAntarctic zone (SAZ) south of Tasmania, for the spring-summer period (grey triangles) and the fall-winter period (black triangles), from 1982 to 2005. Spring-summer (fall-winter) period anomalies were computed as the difference between the spring-summer (fall-winter) period mean value and the average of spring-summer (fall-winter) period mean values of the whole data-set. 4,2007 Inter-annual variability of the CO 2 sink south of Tasmania  1981 1983 1985 1987 1989 1991 1993 1995 1997 1999