The 14C measurements in the Arabian Sea and the Bay of Bengal during the late 1990s offer a way to assess the temporal changes in the inventories of bomb-14C and its penetration into the ocean, in two decades since GEOSECS expeditions (1977–1978). The mean penetration depth of bomb radiocarbon during GEOSECS (1977–1978) was 270 m, which increased by ∼40% to 381 m in 1994–1998. The small changes in bomb-14C inventories, significant increase in the mean penetration depths and lowering of the surface Δ14C values in the northern Indian Ocean indicate the temporal variation of bomb-14C in two decades is mainly through downward transfer through mixing with deeper waters. The observed bomb-14C inventory in the northern Indian Ocean agrees with numerical model simulated values, except at the equatorial Indian Ocean. The high bomb-14C inventory at the equator can be attributed to lateral advection of 14C-enriched waters from the Pacific Ocean through the Indonesian archipelago. The air-sea CO2exchange rates in the northern Indian Ocean calculated from the bomb-14C inventories range from ∼7 mol m−2 yr−1 (in the northern Bay of Bengal) to 20 mol m−2 yr−1(in the equatorial Indian Ocean). Net sea-air flux of CO2 estimated for the northern Indian Ocean between 0° and 25°N is ∼104 ± 30 TgC yr−1. The Bay of Bengal is a net sink of atmospheric CO2 (∼−1 ± 0.4 TgC yr−1), while the Arabian Sea is a source of CO2 (∼69 ± 21 TgC yr−1).
 The naturally occurring radioactive isotope of carbon 14C, commonly known as radiocarbon, can be used as a tracer to study the pathways of carbon across various exchangeable carbon reservoirs (e.g., oceans, biosphere, soil organic matter), which interact with atmospheric CO2 reservoir on time scale of decades to centuries [Stuiver et al., 2000]. The inadvertent addition of bomb-14C in the environment between early 1950s and 1960s from above ground nuclear detonations can be used to address a host of problems pertaining to the global carbon cycle and ocean circulation. Since late 1950s global surface oceans showed a measurable increase in 14C concentration due to transfer of bomb-14C from the atmosphere to the oceans [Rafter and Ferguson, 1957]. 14C in oceans has long been employed as a useful tracer to study the processes of air-sea CO2 exchange and circulation at different time scales. Both natural (cosmogenic) and bomb produced (anthropogenic) 14C are suitable for these studies.
 The increase in oceanic inventory of bomb-14C depends on the local air-sea CO2exchange rate, diffusivity, vertical and lateral movement of seawaters. Since the bomb-14C concentration in the upper layers of the ocean changes on time scale of decades, it is an ideal tracer to study air-sea CO2 exchange processes and to trace decadal mixing processes in the thermocline [Quay and Stuiver, 1980; Broecker et al., 1985, 1995; Druffel, 1989]. Measurement of 14C in oceans is also important for validating ocean general circulation models that are primarily used for predicting oceanic CO2 uptake rates and used to simulate the oceanic 14C distribution [Toggweiler et al., 1989; Jain et al., 1995; Guilderson et al., 2000; Sweeney et al., 2007].
 The Arabian Sea and the Bay of Bengal in the northern Indian Ocean are characterized by contrasting oceanographic features, which result in distinct differences in their circulation pattern in the thermocline region and also in the magnitude and direction of the sea-air CO2fluxes. The Bay of Bengal receives large amount of fresh water throughout the year from seven major rivers, with peak input during the southwest monsoon season (June to September). Large inputs of fresh water causes formation of a low salinity and low-density layer at the surface Bay of Bengal, thereby creating a steep gradient of the isopycnal surfaces and reducing the vertical mixing. This, coupled with high biological productivity keeps the surface seawaterpCO2 in this oceanic region lower than that in the atmosphere, making it a sink of atmospheric CO2 [Kumar et al., 1996]. Whereas in the Arabian Sea the pCO2 of the surface waters is greater than that in the atmosphere due to strong wind induced upwelling of CO2 rich deep waters, thus making it a net source of CO2 to the atmosphere [Kumar et al., 1992; Sarma et al., 1998, 2000; Takahashi et al., 2009]. These contrasting physical oceanographic settings and the wind speed regimes of the two basins influence the regional ocean-atmosphere fluxes of CO2. In addition, these also influence the isotopic composition of carbon in the concerned reservoirs, i.e., both in the oceanic dissolved inorganic carbon (DIC) as well as in the atmospheric CO2 overlying this ocean. Complex and contrasting ocean circulation patterns in the Arabian Sea and the Bay of Bengal complicate the modeling of the distribution of tracers in this region. The first measurements of 14C in the water column of the Arabian Sea and the Bay of Bengal were done during 1977 and 1978, as a part of the international GEOSECS program [Stuiver and Östlund, 1983]. Based on reoccupation of three Arabian Sea GEOSECS stations, Bhushan et al.  described temporal changes in radiocarbon distribution about two decades after GEOSECS.
 To study the temporal changes in the distribution of radiocarbon in the Arabian Sea and the Bay of Bengal, three expeditions were made between 1997 and 1999 onboard FORV Sagar Sampada (Figure 1). During 1997 (cruise SS-152), three GEOSECS stations were reoccupied—one in the equatorial Indian Ocean and the other two in the central Bay of Bengal. In 1998 (cruise SS-164), GEOSECS stations in the Arabian Sea were reoccupied including one station in the Gulf of Aden near the Red Sea mouth. During 1999 (SS-172) stations in the northern and the southern Bay of Bengal and the Andaman Sea were occupied. Samples of seawater were collected in vertical profiles for measurements of14C in DIC. CO2 were also collected from the maritime air during these cruises for 14CO2 measurements. Here we report the observed changes of 14C in DIC of the northern Indian Ocean in two decades since the GEOSECS expedition. From the distributions of bomb-14C in the upper ocean, we have estimated net sea-air flux of CO2 for the northern Indian Ocean.
2. Material and Methods
2.1. Sampling of Seawater
 Seawater samples were collected using 30 lit PVC Go-Flo bottles (General Oceanics, Miami, Florida, USA) attached with a 12-position rosette fitted with eight bottles. Continuous profiles of temperature, salinity and dissolved oxygen were obtained using a CTD (conductivity-temperature-depth) profiler. From the raw CTD data, potential temperature (θ) and potential density (σθ) were derived using the software package SEASOFT (v. 4.232). The CTD temperature and salinity values are expressed following the conventions of ITS-90 (degree Celsius) and PSS-78 (Practical Salinity Scale), respectively. The dissolved oxygen and salinity values from CTD were calibrated by comparing the onboard measurements of O2 by Winkler's titration and salinity by Autosal salinometer in collected seawater samples. Water samples were collected from selected depths, from the surface (∼5 m) to a few hundred meters above the seafloor guided by the CTD profiles. Usually two samples were collected from each cast by tripping four 30 lit bottles at the desired depth, thus collecting about 120 lit of seawater for each depth. Aliquots of this sample were used for measurements of various parameters.
2.2. Analyses of Nutrients (Silicate, Total Nitrate and Phosphate)
 About 500 ml of seawater was used for the analysis of dissolved silicate, total nitrate (nitrate + nitrite) and phosphate [Strickland and Parsons, 1972]. The samples were kept refrigerated prior to analysis. Silicate and nitrate were measured onboard using a Techniconauto-analyzer. Phosphate was measured by spectrophotometry using aBeckman model 26 spectrophotometer. Based on repeat measurements of seawater samples, the precisions (±1σ) of silicate, total nitrate and phosphate measurements were 1.34%, 1.36% and 1.63%, respectively of the measured values. Accuracies of silicate and nitrate measurements were about ±5%, as checked by analyzing CSK standard solutions (Wako Chemical Industries Ltd., Japan).
2.3. Analysis of Total Dissolved Inorganic Carbon (ΣCO2)
 An aliquot of seawater sampled from each depth in the PVC Go-Flo bottles were collected in 125 ml Borosilicate glass bottles. Immediately after collection the samples were poisoned with 25 μl of saturated mercuric chloride solution to halt biological production. The bottles were sealed with ground glass stoppers (greased with Apeizon-N grease) and kept refrigerated at ∼2°C until analysis. ΣCO2 was measured in the samples using a UIC model 5012 CO2 coulometer (UIC Inc., Joliet, Illinois, USA) [Körtzinger et al., 1999]. Based on repeat analyses of seawater samples, typical precision of ΣCO2 measurements was ±3 μmol kg−1 (1σ). The accuracy of ΣCO2 measurements was checked from analyses of certified reference seawater (Batch#42, December, 1997, supplied by Prof. Andrew G. Dickson of Scripps Institute of Oceanography, USA). The mean ΣCO2 measured during this study for this standard seawater is 1983.2 ± 2.3 μmol kg−1 (1σ, n = 20), in good agreement with its certified value of 1985.1 ± 0.8 μmol kg−1.
2.4. Radiocarbon Analysis
14C in the seawater DIC samples were measured using conventional liquid scintillation counting (LSC) technique. Details of the method are discussed in Dutta . About 100 lit of seawater sample were taken for each 14C analysis. DIC was extracted onboard from the seawater soon after sample collection using a closed circulation system to minimize exchange and contamination with atmospheric CO2. In the laboratory the extracted DIC were converted to benzene for 14C assay, following the procedures outlined in Noakes et al.  and Gupta and Polach . Details of the LSC methods were described by Bhushan et al. . The 14C results are expressed as Δ14C permil (‰) as per the conventions outlined by Stuiver and Polach . The internal precision (1σ) of the Δ14C measurements based on modern samples (using 3 ml of benzene) was ±5‰.
3. Depth Profiles of 14C in DIC of the Northern Indian Ocean
 The depths sampled for 14C analysis are indicated in the θ-S plots ofFigure 2, which show the water masses sampled. The measured Δ14C profiles (filled circles with ±1σ errors) are shown in Figure 3. The reconstructed pre-nuclear Δ14C profiles for these stations are given as dashed lines (method described in Appendix A). Significant decrease in the Δ14C values were observed in the surface Bay of Bengal in two decades after the GEOSECS expedition of 1977–1978 (Figure 4). During 1997 and 1999, the Δ14C values of the surface waters (at ∼5 m) ranged from 57‰ in the southern Bay of Bengal (SS172–4041) to a minimum of 18‰ in the northern Bay of Bengal (SS172–4030). The mean surface Δ14C values for the Bay of Bengal stations between 9°N to 20°N were about 40‰; while Δ14C in the surface equatorial Indian Ocean (SS152–3846) was 55‰. These are lower by ∼60‰ to 80‰ than those measured during 1978, comparing the values at GEOSECS stations 445 and 448 (Table 1). The Δ14C decrease observed in the surface Arabian Sea was ∼30‰. During the GEOSECS expedition significant contrast was seen between the surface Δ14C values of the Arabian Sea (59 to 95‰) and the Bay of Bengal (107 to 117‰). This difference became insignificant after two decades, when the surface Arabian Sea recorded Δ14C values of 40 to 58‰ during 1994 and 1995, compared to 40 to 57‰ in the Andaman Sea and the southern Bay of Bengal during 1997 and 1999 (Figure 5) [Bhushan et al., 2000; Dutta et al., 2007]. The reduction in the surface Δ14C values can be partly attributed to the lowering of the atmospheric Δ14C from ∼300‰ during 1978 [Nydal and Lövseth, 1996; Chakraborty et al., 2008] to ∼90‰ in late 1990s [Bhushan et al., 1997; Dutta et al., 2006], and transfer of 14C-enriched waters from the surface to the deeper layers through vertical mixing. The later process appears to play a major role in the northern Indian Ocean as evidenced from the changes in the mean penetration depths of bomb-14C.
Table 1. Inventory and Mean Penetration Depths of Bomb-14C, and Bomb-14C Based Air-Sea CO2 Exchange Rates in the Northern Indian Oceana
 Lowest surface Δ14C in the Bay of Bengal measured at the station SS172–4030. This station is located closest to the Ganges-Brahmaputra delta, which receives large fresh water inputs both through river discharge as well as submarine groundwater discharge. The low Δ14C values measured at this station are most likely due to mixing with large amount of submarine groundwater discharge with old 14C-age.Ravi Prasad et al.  had reported 14C-ages ranging from 3040 BP to 9025 BP for groundwater samples collected from Orissa state, not far from the northeastern Bay of Bengal. In absence of detailed Δ14C measurements in the Ganges-Brahmaputra river system, it is difficult to assess the riverine contribution to depleted Δ14C values in the northern Bay of Bengal.
 In all stations of the Bay of Bengal a distinct maxima of Δ14C was seen at depths of 70∼100 m, with Δ14C values of ∼60 to 70‰. The subsurface maxima was developed due to gradual downward transfer of the surface bomb-14C transient to deeper levels, and subsequent equilibration of the surface water with atmospheric CO2 with progressively lower Δ14C values. This feature was not seen in the northern Bay of Bengal (SS172–4030) probably due to rapid replenishment of the surface waters resulting from voluminous fresh water discharge. Such subsurface maxima could not be resolved in the Arabian Sea Δ14C profiles obtained during 1994–1995 due to coarser sampling resolution. However, it is unlikely that such feature could develop in the Arabian Sea. The isopycnal gradient of the surface waters is much less in the Arabian Sea than in the Bay of Bengal, aiding vertical mixing of waters and hence more rapid transfer of 14C-enriched waters from the surface to deeper levels, thus preventing formation of any distinct subsurface Δ14C maxima in the Arabian Sea.
 From the GEOSECS measurements, no significant changes in the deep-water Δ14C values of the northern Indian Ocean stations are seen from the present study, except for the station SS164–4018 (reoccupation of GEOSECS 413), near the Gulf of Aden. The intermediate water between 1200 to 2200 m depths in this station has shown an increase of Δ14C by ∼40‰ from the values measured during 1977 [Bhushan et al., 2003]. This increase is most likely due to gradual sinking of more saline and dense Red Sea surface water (RSW), with higher Δ14C to deeper depths. The mixing proportion of RSW between 1200 and 2200m depths near the location of SS164–4018 is 5 to 20%, as seen from multi parametric water mass analyses in the northern Indian Ocean [Goyet et al., 1999]. In the deep waters at the equatorial Indian Ocean and the southern Bay of Bengal the measured Δ14C values below 2500 m are higher than the estimated pre-nuclear Δ14C profile (derived from silica) by about 20‰. Presence of Antarctic bottom water with higher silica content at these depths causes this deviation. The scatterplot of Δ14C versus silica for global ocean waters deeper than 1000 m [Broecker et al., 1995] shows significant deviation from the linearity at the high silica end caused by circum-polar waters. However, this feature is not seen in the southern Bay of Bengal station SS172–4041.
3.1. Inventories of Bomb-14C in the Northern Indian Ocean
 The inventories and mean penetration depths of bomb 14C were determined from the column-integrated bomb-14C, ΣCO2 and the surface 14C excess, following the procedures outlined in Broecker et al.  (Appendix B). The inventories for the GEOSECS stations and for the Arabian Sea stations occupied during 1994 and 1995 [Bhushan et al., 2000] are recalculated using the pre-nuclear surface ocean Δ14C values (Appendix A). The results are shown in Table 1. No significant changes are noticed in the recalculated values of bomb-14C inventories and mean penetration depths, from the values reported in Bhushan et al. .
 The snapshot distributions of bomb-14C in the northern Indian Ocean as observed during 1977–1978 (GEOSECS) [Stuiver and Östlund, 1983], during 1994–1995 [Somayajulu et al., 1999; Bhushan et al., 2000] and during 1997–1999 are shown in Figures 4 and 5. Between 1977 and 1978 and late 1990s the bomb-14C inventories has increased up to 10% on the average for most stations in the northern Indian Ocean, excluding the stations near the equatorial Indian Ocean, e.g., SS132–3273 (near GEOSECS 419) and SS152–3846 (reoccupation of GEOSECS 448), where bomb-14C inventory has increased by ∼80% (Table 2). The mean 14C penetration depths also have been more than doubled at these stations. In the equatorial Indian Ocean station SS152–3846, the largest increase of bomb-14C has been observed for depths between 200 m and 1000 m (Figure 3). Major portion of this bomb-14C were seen above 500 m. From the reoccupation of GEOSECS equatorial Indian Ocean stations (between 4°N and 6°S) during the French INDIGO expeditions of 1981–1987, it has been shown that bomb-14C inventory had increased to ∼90% and the penetration depth had increased to ∼80% [Bard et al., 1988, 1989]. This increase has been attributed to the westward flowing waters with high concentration of bomb tracers, due to the entry of 14C-enriched Pacific waters through the passages in the Indonesian archipelago, or mixing with the Mode water from the southern gyre of the Indian Ocean. Evidence of influx of Pacific water through the Indonesian archipelago has been demonstrated from GEOSECS surface tritium distribution [Fine, 1985], and high concentration of 90Sr in banded corals from the western Indian Ocean [Toggweiler and Trumbore, 1985]. Povinec et al.  observed that concentrations of tritium, 14C, 90Sr, and 129I in the surface waters of the Indian Ocean were similar to that of the northwestern Pacific Ocean, and attributed them to transport of Pacific Ocean water masses to the Indian Ocean through the Indonesian Seas. Such lateral transport processes must have been operative for the observed large increase in bomb-14C inventories at the stations SS132–3273 and SS152–3846. The inventory at SS172–4041 (near GEOSECS 445) which doesn't fall in the path of the throughflow water for its NW–SE trend, has not changed much since 1978, although it is at the same latitude (5°N) as SS132–3273. Lowest 14C inventory obtained for the station SS172–4030 (3.1 × 109 atoms cm−2), about 25% lower than other contemporary Bay of Bengal values.
Table 2. Changes in Bomb-14C Inventories in Two Decades Since GEOSECS
Stations: GEOSECS/PRL Reoccupied
Bomb-14C Inventory (×109 atoms cm−2)
PRL Expeditions (1994–1999)
Inventory Change in Two Decades
3.2. Comparison of Observed Changes of Bomb-14C Inventories With Modeled Values
Toggweiler et al. simulated the penetration of bomb-produced14C in the world ocean using the Geophysical Fluid Dynamics Laboratory's (GFDL) primitive equation ocean general circulation model. Changes in the inventories of bomb-14C have been simulated using this model, for the entire world ocean. In this study the comparison of changes in bomb-14C inventories are done between 1970s and late 1990s, the predicted changes between 1990 and 1981 should be considered as the lower limit. For the decade 1981 to 1990, the model predicts increase in bomb-14C inventories from 0 to 2 × 109 atoms cm−2for the Arabian Sea and the equatorial Indian Ocean, while no change has been predicted for the Bay of Bengal. The average bomb-14C inventory for the GEOSECS stations 416, 417 and 418 during 1977 was 5.9 × 109 atoms cm−2, which rose to 6.4 × 109 atoms cm−2during 1994 and 1995. Average bomb-14C inventory in the Bay of Bengal during 1978 was 4.1 × 109 atoms cm−2; while during 1997 and 1999 it rose to 4.4 × 109 atoms cm−2, consistent with the model prediction. Major discrepancy is seen at the equatorial Indian Ocean, where the inventory increased by 4.5 × 109 atoms cm−2 between 1978 and 1997, which is more than twice the model predicted rise between 1981 and 1990. Thus, the observed 14C values in the northern Indian Ocean obtained from this work and from Bhushan et al.  agrees with the GFDL model simulated values, except at the equatorial Indian Ocean. Peacock based on model based calculations projected roughly 30% increase in ocean bomb radiocarbon inventory between the mid-1970s and mid-1990s.
3.3. Changes in Penetration Depths of Bomb-14C in the Northern Indian Ocean
 The average value of the mean penetration depth during 1977–1978 for the five Arabian Sea GEOSECS stations (413, 416, 417, 418 and 419) was 270 m (Table 1). During 1994–1998, this has increased by ∼40%, to 381 m on the average (considering only the reoccupation or nearby stations). In the Bay of Bengal, the mean penetration depth has increased by ∼80%, from 168 m in 1978 (for the two GEOSECS stations 445 and 446), to 297 m during 1997 and 1999 (average of five stations). The relative increase of mean penetration depths in two decades is higher for the Bay of Bengal than that for the Arabian Sea, although the actual depth of penetration of bomb-14C is always higher in the Arabian Sea than in the Bay of Bengal. This could be due slower rate of penetration of bomb-14C in the Bay of Bengal during the GEOSECS times, when due to strong isopycnal gradient, bomb-14C was mainly accumulated within the top ∼200 m of water column. This also explains higher surface Δ14C in the Bay of Bengal than in the Arabian Sea during the GEOSECS. The downward transfer of bomb-14C would have become faster after it crossed the surface barrier of isopycnal gradient, thus causing relatively large increase in the penetration depths of bomb-14C after GEOSECS. However, since mean penetration depth is inversely proportional to the change in the surface ocean Δ14C from the pre-nuclear values (Appendix B), and the latter is steadily falling due to continuously decreasing atmospheric Δ14C, the observed changes in the mean penetration depths can also be explained from changes in the surface Δ14C values. From subsequent 14C measurements in the northern Indian Ocean, Dutta et al.  showed that the decline of Δ14C in the surface Bay of Bengal was just ∼4‰ in one decade between 1995 and 2006, as compared to ∼25‰ observed in two decades between 1978 and 1995. The smaller rates of surface ocean Δ14C decline in the recent decades can be attributed to much smaller decrease-rate of ∼5‰ yr−1 of the atmospheric Δ14C during late-1990s [Dutta et al., 2006], as compared to ∼12‰ yr−1 between 1978 and 1995 [Hua and Barbetti, 2004].
4. 14C-based Air-Sea CO2 Exchange Rates in the Northern Indian Ocean
 Air-sea exchange rates of CO2 have been calculated for all stations based on the bomb 14C inventories following the procedure of Stuiver , as described in Appendix B. The method is based on the assumption, that the net amount of bomb-14C gone into the ocean is proportional to the integrated atmosphere-ocean Δ14C gradient and the regional air-sea exchange rate of CO2. Another assumption of this procedure is that the changes in the bomb-14C inventory only result from air-sea exchange of CO2 and vertical diffusive mixing, and the invasion of CO2from air to sea is equal to its evasion out of the sea. However, this assumption may not be valid for some locations, e.g., SS172–4030 and SS152–3846 where lateral transport plays a significant role in deciding the bomb-14C inventory. To ascertain and quantify the component of bomb-14C contribution due to lateral transport, detailed measurements of 14C profiles are needed with better vertical and spatial coverage.
4.1. Distribution of Air-Sea CO2 Exchange Rates Over the Northern Indian Ocean
Bhushan et al. had reported the air-sea CO2 exchange rates for the Arabian Sea from the GEOSECS 14C measurements and from the measurements made during 1994 and 1995. In that study, tropospheric Δ14C history for the northern hemisphere was used, which has peak Δ14C of ∼1000‰ in 1964 [Bhushan et al., 1997]. As far as tropical Indian Ocean is concerned, it is more appropriate to use the tropospheric Δ14C trend for the Northern Hemisphere Zone 3 as discussed in Hua and Barbetti , where the peak Δ14C was ∼700‰ in 1965. Chakraborty et al.  reported peak tropospheric Δ14C of 708 ± 8‰ from tree rings of central India grown in 1965. In this study we have adopted a peak Δ14C of 700‰. In Bhushan et al.  the mixed layer Δ14C trend was derived from a Gulf of Kutch coral, which recorded peak Δ14C of 170‰ in 1968 [Chakraborty et al., 1994]. The mixed-layer Δ14C trend recorded in the Gulf of Kutch may not be representative for the open northern Indian Ocean. If the Δ14C trend from the Gulf of Kutch coral is used then, the integrated mixed-layer Δ14C value would be an overestimate, due to the overall higher Δ14C values in the Gulf of Kutch than the contemporary open northern Indian Ocean waters. However, since for determining the 14C input, it is the atmosphere-ocean gradient of Δ14C that is integrated, the effect of using higher Δ14C values for both the atmosphere and the ocean mixed layer is likely to cancel each other. The 14C based air-sea CO2 exchange rates for the northern Indian Ocean stations are given in Table 1. As compared to the values reported in Bhushan et al. , the exchange rates recalculated in this study are higher by ∼9% on an average for all but one station in the Arabian Sea (the recalculated values are higher by −1 to 17% for different stations).
 The air-sea CO2exchange rates in the northern Indian Ocean calculated from the bomb-14C inventories range from ∼7 mol m−2 yr−1 (in the northern Bay of Bengal) to 20 mol m−2 yr−1 (in the equatorial Indian Ocean). High CO2 exchange rates of ∼17 mol m−2 yr−1 are measured in the western Arabian Sea, off Oman and near the Gulf of Aden (Table 1 and Figure 6). The high exchange rate obtained at the equator (and also possibly at SS132–3273) could be an artifact of lateral advection of 14C-enriched waters from the Pacific into this area from the Indonesian archipelago.
 The measured air-sea CO2exchange rates are plotted with the long-term wind speed inFigure 7. All points for the northern Indian Ocean station fall on a roughly linear trend, for wind speeds ranging from 5 to 8 m sec−1. From the GEOSECS 14C measurements, Broecker et al.  determined mean exchange rate of 20 mol m−2 yr−1 for the global ocean, using global average surface wind speed of 7.4 m sec−1. From 14C measurements in corals from the Red Sea, an invasion flux of 8 ± 2 mol m−2 yr−1 has been obtained, for a mean wind speed 4.7 m sec−1 [Cember, 1989]. As expected, the CO2 exchange rates calculated from the Red Sea coral, northern Indian Ocean measurements from this study and the global mean value obtained from GEOSECS 14C data [Broecker et al., 1985], follow a rough trend. Also shown in Figure 7are the empirical linear, quadratic and cubic relationships established between long-term average wind speed and CO2 exchange rates [Wanninkhof et al., 1985; Wanninkhof, 1992; Wanninkhof and McGillis, 1999; Sweeney et al., 2007]. The cubic relationship appears to fit better with the bomb-14C derived CO2 exchange rates in the northern Indian Ocean.
 In Figure 7, the exchange rate obtained for the equatorial Indian Ocean station SS152–3846 during 1997, is a clear outlier from the general polynomial trend. It is interesting to note here, that using the bomb-14C inventory measured at this station during 1978 (GEOSECS 448) and the local long-term wind speed of ∼5 m sec−1, one gets an exchange rate of 12.1 mol m−2 yr−1, which fits better with the general trend. As discussed, significant change in bomb-14C at this location is due to lateral advection of waters from the Pacific. Assuming lateral advection a steady state process, the fact that the inventory of GEOSECS 448 matches the regional wind speed indicates that the Δ14C of the water that is flowing in to this region has steadily increased in the last two decades since 1978. This increase is in response to the bomb transient, with a delay introduced by the transport processes. Thus, in the equatorial Indian Ocean, between 1978 and 1997, a major portion of the increment in the bomb-14C inventory is through lateral advection of 14C-enriched waters, rather than through air-sea CO2 exchange. The combined effects of uncertainties in the gas transfer velocity and wind fields lead to average difference of 27% between the lowest and highest estimates of the CO2 flux from the region [Feely et al., 2004].
4.2. Estimate of Net Sea-Air Flux of CO2 From the Northern Indian Ocean
 Both in the Arabian Sea and the Bay of Bengal, large seasonal variations of pCO2 in the surface waters are seen due to strong seasonality of wind induced upwelling and nutrient inputs through large amount of river discharge [Kumar et al., 1992]. Throughout the year surface pCO2 for the Arabian Sea is higher than in the Bay of Bengal [Sarma et al., 1998], which increases during the southwest monsoon season [Körtzinger et al., 1997].
 We have divided the northern Indian Ocean region into four regions: (i) Arabian Sea (6°N to 25°N and 40°E to 80°E, including Red Sea mouth, and excluding Persian Gulf); (ii) North Bay of Bengal (14°N to 22°N and 80°E to 97°E); (iii) South Bay of Bengal (6°N to 14°N and 80°E to 100°E); and (iv) North Tropical Indian Ocean (2°S to 6°N and 41°E to 100°E). The average ΔpCO2and air-sea CO2 exchange rates were interpolated for each of these regions in 4° × 5° grids, using the gridded ΔpCO2 values of Takahashi et al. . The net CO2fluxes from these northern Indian Ocean regions were calculated from the regional average exchange rates and the ocean-atmospherepCO2 gradients (ΔpCO2). We have used atmospheric pCO2 of 356 μatm, as mean difference of pCO2 of seawater and ΔpCO2 reported in Takahashi et al. . The results are summarized in Table 3 and Figure 8.
Table 3. Estimates of Sea to Air Flux of CO2 for the Northern Indian Ocean
Approx. Area (106 km2)
Mean ΔpCO2 (μatm)
Mean Air-Sea CO2 Exchange Rate (mol m−2 yr−1)
Mean CO2 Flux Rate (mol m−2 yr−1)
Net Regional Flux of CO2 (Tg C yr−1)
Arabian Sea (6°–25°N, 50°E–80°E)
14 ± 3
+69 ± 21
North Bay of Bengal (14°N–22°N, 80°E–97°E)
7 ± 1
−2.1 ± 0.6
South Bay of Bengal (6°N–14°N, 80°E–100°E)
9 ± 2
+1.5 ± 0.5
North Tropical Indian Ocean (2°S–6°N, 45°E–95°E)
14 ± 3
+49 ± 15
 We estimate net annual flux of CO2 from the northern Indian Ocean region (between 0° to 25°N) to ∼104 ± 30 Tg C yr−1. The source of this flux is entirely located in the Arabian Sea, with annual sea-air flux of ∼69 ± 21 Tg C yr−1. The Bay of Bengal as a whole is a sink of atmospheric CO2, where the sink is mainly located in its northern part. The estimated net uptake rate of CO2 by the Bay of Bengal is ∼1 ± 0.4 Tg C yr−1. The CO2 source in the tropical Indian Ocean within 0° and 6°N is estimated to ∼35 ± 11 Tg C yr−1. Our estimated sea-air fluxes agree well with the values reported inTakahashi et al. , where the flux from the temperate Indian Ocean (north of 14°N) has been estimated to 0.02 Pg C yr−1, and 0.1 Pg C yr−1 for the tropical Indian Ocean between 14°S and 14°N. Bates et al. based on wind speed based exchange rates determined net sea-to-air CO2 flux of 237 ± 132 Tg C yr−1 for the Indian Ocean and found that the Arabian Sea, Bay of Bengal, and 10°N–10°S zones were perennial sources of CO2 to the atmosphere, although the Bay of Bengal became a sink for CO2 in December. Louanchi et al.  estimated annual flux of 169 TgC yr−1for the northern Indian Ocean between 10°S and 18°N, using a one-dimensional model and incorporating the effects of physical and biogeochemical processes.Gruber et al. based on the model calculations of estimates of the contemporary net air-sea CO2 fluxes, showed a consistent description of the regional distribution of annual mean sources and sinks of atmospheric CO2 for the decade of the 1990s and the early 2000s. They found that this distribution is characterized by outgassing in the tropics, uptake in midlatitudes, and comparatively small fluxes in the high latitudes. However, to constrain the derived estimates in the Arabian Sea and the Bay of Bengal, more detailed measurements of radiocarbon in the water column and surface ocean pCO2 in the Arabian Sea and the Bay of Bengal with more spatial and temporal variability are required to constrain the CO2 invasion and evasion rates.
 This study reports the changes in the distribution of 14C in the water column of the Bay of Bengal, the equatorial Indian Ocean and near the Red Sea mouth, about two decades after the GEOSECS expedition. Our measurements show on an average there was only marginal change in the inventory of nuclear-bomb produced14C in the Arabian Sea and the Bay of Bengal, in two decades between 1978 and 1999. The bomb-14C inventory has nearly doubled to 9.5 × 109 atoms m−2 in the equatorial Indian Ocean, likely due to lateral advection of 14C rich waters from the southeast. The air-sea CO2exchange rates based on bomb-14C inventories range from 7 to 17 mol m−2 yr−1. Except for the equatorial Indian Ocean, the 14C-derived exchange rates are mostly a function of long-term wind speeds over this region. From the average14C derived CO2 exchange rates and reported annual surface seawater pCO2 values, the net flux of CO2 from the Indian Ocean in the late 1990s has been estimated to ∼104 ± 30 TgC yr−1.
Appendix A:: Reconstruction of Pre-nuclear Δ14C Profiles
 To assess the temporal changes of bomb-14C, data on the steady state or pre-nuclear Δ14C profiles are required. In the absence of the measurements of pre-nuclear Δ14C profiles, they can be reconstructed based on the pre-nuclear surface Δ14C (derived from marine mollusk shells of known age) and the profiles of bomb produced tritium (3H) or dissolved silicate [Broecker et al., 1985, 1995]. In this work, the pre-nuclear Δ14C profiles for different stations are reconstructed from the measured silicate data, based on the empirical linear relationship for the global ocean between the natural Δ14C and silicate (Δ14Cnatural = −70 − silicate), established by Broecker et al. , where Δ14C is expressed in permil (‰) and Silicate in micromole per kg (μmol kg−1). The above linear relationship has been derived from the Δ14C-silicate data of GEOSECS samples deeper than 1000 m, which are uncontaminated with bomb-14C and 3H. Deep waters originating from Antarctic circum-polar regions have relatively higher Δ14C and therefore deviate from the above linear relationship. From the GEOSECS deep-water samples of the Arabian Sea and the Bay of Bengal with Silicate >50μmol kg−1, Bhushan et al.  obtained a relation Δ14Cnatural= −80 − 0.83*silicate. The pre-nuclear Δ14C profiles based on this relation and those derived from Broecker et al. are consistent within ±10‰. The estimated uncertainty of reconstructing the pre-nuclear Δ14C using the linear relation of Broecker et al.  is also ±10‰ [Peng et al., 1998]. Therefore, reconstruction of pre-nuclear Δ14C for the northern Indian Ocean using the relation (Δ14Cnatural = −80 − 0.83*silicate) will not be significantly different from that derived using the relation of Broecker et al. . In this study, the pre-nuclear profiles have been calculated from the relation (Δ14Cnatural = −70 − silicate). Rubin and Key  developed a new separation method based on the strong linear correlation between Δ14C and potential alkalinity. This method provides an estimate of surface ocean pre-bomb Δ14C concentrations. However, this method is not applicable to all regions. The primary disadvantage relative to the silicate method is that high quality alkalinity measurements are less common than high quality silicate measurements. For low- and midlatitude waters, both methods give comparable results and for high latitude southern waters, the potential alkalinity method predicts lower natural Δ14C and consequently higher bomb 14C.
 The pre-nuclear surface Δ14C values used in this study for different regions are based on the Δ14C measurements of pre-nuclear shells [Dutta et al., 2001; Southon et al., 2002], rounded off to nearest 10‰. The surface Silicate values are adjusted to match the reconstructed pre-nuclear surface Δ14C with the values estimated from shells. The pre-nuclear surface Δ14C values adopted for different regions of the northern Indian Ocean are given in Table A1.
Table A1. Pre-nuclear Surface Δ14C Values Adopted for the Northern Indian Oceana
Central Bay of Bengal and Andaman Sea (Stewart Sd.)
Southern Bay of Bengal (Rameswaram)
Equatorial Indian Ocean (Rameswaram)
Southern Arabian Sea (Rameswaram)
Central Arabian Sea (Bombay, Goa and Malabar)
Northern Arabian Sea (Dwarka and Port Okha)
Appendix B:: Determination of Bomb-14C Inventories and Air-Sea CO2 Exchange Rates
 The inventories of bomb-14C are calculated following the procedures of Broecker et al. . The area (A) between the measured Δ14C depth profile and the reconstructed pre-nuclear Δ14C depth profile (Δ14C0) is calculated as
Here, z0 is the depth at which the difference between Δ14C and Δ14C0 becomes negligible (<1‰). The mean penetration depth (Z) of bomb-14C is obtained as
Here, ΔΔ14C = (Δ14C − Δ14C0)surface is the measured excess of surface water Δ14C from its pre-nuclear value. The total column-inventory of bomb-14C is given by
Here ΣCO2is the mean DIC of the water column up to which bomb-14C has penetrated. The proportionality constant ‘k’ takes into account density of seawater (average ∼1025 kg m−3), Avogadro's number (6.023 × 1023 atoms mol−1), and the 14C/12C atom ratio of the NBS Oxalic Acid-I standard (1.176 × 1012 [Karlen et al., 1968]). The inventory of bomb-14C is expressed in units of atoms of 14C per square unit area of the seawater column. Following this procedure, the estimate uncertainty of the calculated inventory is about ±10% [Peng et al., 1998]. To compute the bomb-14C inventories, the measured depth profiles of Δ14C, silicate and ΣCO2values were interpolated to every 1 m. Pre-nuclear Δ14C profiles (Δ14C0) were then reconstructed from the measured values of silicate, using the empirical relationship of Broecker et al. , Δ14C0 (‰) = −70 − silicate (μmol kg−1). The reconstructed pre-nuclear Δ14C of surface was forced to match with the values obtained from pre-bomb shells. The excess (measured Δ14C − Δ14C0) are then integrated until the difference becomes negligible and the inventory of bomb-14C is obtained using (B1) and (B3).
 The air-sea CO2 exchange rates were calculated following the procedure outlined by Stuiver The total amount of bomb-14C penetrated in ocean is proportional to the air-sea exchange rate of CO2 and the time integrated gradient of Δ14C between the atmosphere (Δ14Catm) and the ocean surface mixed layer (Δ14Cmix). Assuming the mean steady state difference in Δ14C values between atmosphere and oceanic mixed layer is −60‰ for the northern Indian Ocean, the total amount of bomb-14C (Q14), transferred per unit of ocean surface over time t(in years) is related to the air-sea exchange rate of CO2 (F12) as
The constant term on the right hand side of (B4), takes into account the isotopic fractionation factors for ocean-atmosphere transfer and normalization of14C activity to δ13C value of −25‰ [Stuiver, 1980]. Here Q14 is expressed in mol m−2 and F12 in mol m−2 yr−1. For the onset of input of bomb-14C to the oceans, 1954 is taken as the initial year (t = 0). The values of the integrals for Δ14Catm and Δ14Cmix are obtained from Δ14C measurements in the atmosphere and in corals. The tropospheric Δ14C from 1954 and 1968 has been adopted from the reported measurements from the tropical regions [Nydal and Lövseth, 1996; Hua and Barbetti, 2004; Chakraborty et al., 2008], with peak Δ14C of 700‰ in the year 1965. From 1968 onward, an exponentially decreasing atmospheric Δ14C trend with e-folding time of 17 years has been adopted, fixing the Δ14C for the years 1980 and 1999 at 265‰ and 88‰ respectively. There are too few coral Δ14C measurements in the open northern Indian Ocean, to reconstruct the mixed layer Δ14C history for different regions. Between the years 1954 and 1973, the value of the integral of Δ14Cmix is taken as 1400 [Stuiver, 1980]. From 1973 onward, a linear decrease of surface ocean Δ14C has been assumed, from a value of 140‰ in 1973 to 50‰ in 2000, to match with the observed values from the GEOSECS measurements during 1977–1978 [Stuiver and Östlund, 1983] and during late 1990s [Dutta et al., 2006]. Q14 are calculated from the bomb 14C inventories for different stations, to determine the air-sea CO2 exchange rates (F12) using (B4). Based on the uncertainty in bomb-14C inventory (15%) and the integral of the atmosphere-ocean Δ14C gradient (5%), the uncertainty of 14C-based exchange rates estimated to ∼20%.Druffel and Griffin  showed that for surface samples, the total uncertainty of a DIC Δ14C value at a given site over a several week period is approximately two times the reported uncertainty (∼7‰). Thus, depending on the application, post-bomb Δ14C data should consider this short-term variability of surface ocean Δ14C values and factor this into their analysis as it might affect the estimates of air-sea CO2 exchange rates.
 Once the rate of air-sea CO2 exchange (F12) is known, the net transfer rate (F) of atmospheric CO2 across the ocean surface can be determined from the difference of the partial pressure of CO2 (ΔpCO2) in the surface ocean and that in the atmosphere [Wanninkhof, 1992]. The net transfer rate of CO2 (F) is given by
In this case, the sign of F (or direction of flux) is positive upwards. Thus the net flux of CO2 will be from sea to air, in regions where ΔpCO2 is positive. This condition is common in most tropical oceanic areas due to equatorial upwelling, which brings deeper CO2rich waters to the surface. The exchange rates calculated from bomb-14C inventories are the long-term averaged values. Thus, if average ocean-atmospherepCO2 gradient (ΔpCO2) over a given oceanic region is known, it is possible to compute the net transfer rate of CO2 from the exchange rates. Based on the uncertainties in the 14C-based exchange rates (20%) and in the estimates of ΔpCO2 (10%) the maximum uncertainty on F is expected to be within 30%.
 We are very much indebted to B. L. K. Somayajulu for his support and guidance, who was inspiration in initiation of this program. Critical comments and suggestions from S. Krishnaswami and two anonymous reviewers greatly helped in improving this paper. We are thankful to the Captain and crew of FORV Sagar Sampada for providing all onboard logistics support and large volume sampling. We are thankful to the Ministry of Earth Sciences (previously Department of Ocean Development), Government of India, for providing ship time. We are grateful to the Director and staff of Centre of Marine Living Resources, Cochin for helping in managing the various cruises and the technical staff of NORINCO for providing help in onboard instrument operation. The onboard help provided by J. P. Bhavsar and R. Agnihotri is sincerely acknowledged. Authors are grateful to M. Dileep Kumar for his critical suggestions.