Long-term observations at several ocean time-series show upward trends of dissolved inorganic carbon (DIC) and seawater pCO₂ due to the uptake of anthropogenic CO₂ (Bates, 2001; Bates et al., 2002; Dore et al., 2003; Keeling et al., 2004; Brix et al., 2004). Any assessment of long-term trends in oceanic CO₂ is complicated by large seasonal variability of the inorganic carbon cycle due to processes such as seasonal temperature, salinity, and density changes, vertical and horizontal mixing, biological production, diurnal warming/cooling, and storm events. Interpretation of oceanic CO₂ time-series data is further complicated by variability imparted by spatial heterogeneity in the ocean as a result of mesoscale and sub-mesoscale phenomena, and meridional and zonal physical gradients.

Although quite a few open-ocean and coastal ocean CO₂ time-series have been initiated, only four time-series sites are of sufficient length to evaluate longer-term interannual trends. These are (1) ALOHA (A Long-term Oligotrophic Habitat Assessment), located near Hawaii (22°45'N, 158°W) in the North Pacific Ocean; (2) BATS (Bermuda Atlantic Time-series Study), located near Bermuda (32°10'N, 64°30'W) in the NW Atlantic Ocean; (3) Hydrostation S, (32°50'N, 64°10'W)) located near Bermuda in the NW Atlantic Ocean, and; (4) ESTOC (European Station for Time-series in the Ocean Canary Islands (ESTOC), located near Gran Canaria in the NE Atlantic Ocean. Several other ocean CO₂ time-series (e.g., site OSP/line P in the Gulf of Alaska, 50° N, 145° W; site KNOT in the North West Pacific Ocean 44° N, 155° E; site OWS Station M in the Norwegian Sea, 66° N, 2° E; transect BRAVO in the Labrador Sea, 57° N, 53° W; site DyFAMed in the Mediterranean Sea, 43° N, 7° E; site observations in the Irminger Sea, 60° N, 36° W) were not evaluated here due to patchy and intermittent data, relatively short duration of observations, or insufficient seasonal resolution (Table 1).

The four long-term ocean time-series (i.e. ALHOA; BATS, Hydrostation S, ESTOC) show upward trends of salinity normalized dissolved inorganic carbon (nDIC) and seawater pCO₂ over time. The anticipated rate of change surface ocean CO₂ due to the accumulation of anthropogenic CO₂ in the atmosphere and the surface ocean buffer factor (assuming that near-surface waters in the subtropical gyres have residence times long enough to equilibrate entirely with the anthropogenic perturbation in atmospheric CO₂) can be theoretically calculated. An equilibrium rate of DIC increase due to anthropogenic CO₂ of +0.9 µmoles kg^-1 yr^-1 was calculated for the subtropical gyres (Bates et al., 2002; Gruber and Sarmiento, 2002). At the four long-term ocean time-series, upward trends of CO₂ are variable. The causes of the trend variability is not certain, but is presently thought to relate to sub-decadal basinwide changes in biological (e.g., productivity) and physical properties (e.g., precipitation-evaporation balance; Dore et al., 2003; Keeling et al., 2004; Brix et al., 2004; and atmospheric annular mode influences such as Pacific decadal oscillation and North Atlantic Oscillation; Bates et al., 2002; Gruber et al., 2002) of the subtropical gyres

In the North Pacific Ocean, observations at the ALOHA site near Hawaii (22°45'N, 158°W), show upward trends of surface ocean salinity normalized dissolved inorganic carbon (nDIC) and seawater pCO₂. nDIC and seawater pCO₂ increased at a rate of +1.2 +0.1 µmoles kg^-1 year^-1, and +2.5 +0.3 µatm year^-1, respectively (Table 2; Dore et al., 2003; Keeling et al., 2004) for the 1988-2002 period. The observed rate of change of surface ocean DIC, for example, was slightly higher than the expected oceanic equilibration with anthropogenic CO₂ in the atmosphere.

In the Northwest Atlantic Ocean, observations at the BATS and Hydrostation S sites near Bermuda (32°N, 64°W), also show upward trends of surface ocean salinity normalized dissolved inorganic carbon (nDIC) and seawater pCO₂. During the first 10 years of observations at BATS (1988-1998), nDIC and seawater pCO₂ increased at a rate of +1.6 µmoles kg^-1 year^-1, and +1.4 µatm year^-1, respectively (Table 2; Figure 1; Bates, 2001). However, merging of BATS and Hydrostation S data show that over a longer twenty year period (1983-2003), nDIC increased at a rate of +0.83 + 0.13 µmoles kg^-1 year^-1 (Table 2; Figure 1; Bates and Keeling, unpublished data, 2004). Within the 95% confidence levels, this rate of oceanic CO₂ increase was similar to the expected oceanic equilibration (i.e., +0.9 µmoles kg^-1 yr^-1) with anthropogenic CO₂ in the atmosphere. Over the 1983-2003 period, seawater pCO₂ increased at a rate of +1.25 + 0.3 µatm year^-1, respectively (Table 2; Figure 1). Concurrently, seawater pH decreased by 0.0012 + 0.0004 pH units year^-1, representing a significant decline of 0.025 pH units (~8.125 to ~8.100) over the last 20 years. In addition, observed alkalinity increased slightly at a rate of +0.17 + 0.08 µmoles kg^-1 yr^-1 (Table 2; Figure 1), though this increase was statistically insignificant.

Long-term observations at the BATS site also indicate that the nDIC of surface and deeper water layers have increased at divergent rates over time since water-column sampling began in 1988. In deeper subtropical mode waters (STMW), the mean rate of change of nDIC over the 1988-2001 period was significantly higher than surface waters, increasing at a rate of 2.2 +0.26 µmoles kg^-1 year^-1 (Table 2). The STMW of the North Atlantic Ocean is formed each winter by cooling and convective mixing at the northern edges of the subtropical gyre south of the Gulf Stream (Klein and Hogg, 1996; Hazeleger and Drijfhout, 1998). The shallow depths of the subtropical gyre (~250-400m deep) are ventilated during STMW formation and the STMW layer is found throughout the subtropical gyre. This water mass is classically defined by temperatures ranging from 17.8° to 18.4°C, by a salinity of ~36.5 +0.05, and by a minimum in the vertical gradient of potential density (or isopycnic potential vorticity) (Klein and Hogg, 1996; Jenkins, 1998; Hanawa and Talley, 2001; Alfutis and Cornillon, 2001). The cause of the divergence between surface ocean and deeper STMW oceanic CO₂ trends is not certain, but is presently thought to relate to atmospheric/climatic variability of the North Atlantic subtropical gyre (Bates et al., 2002; Gruber et al., 2002).

In the Northwest Atlantic Ocean, observations at the ESTOC site near Gran Canaria (29°N, 15°W), also show upward trends of surface ocean salinity normalized dissolved inorganic carbon (nDIC) and seawater pCO₂. nDIC and seawater pCO₂ increased at a rate of +0.4 +1.6 µmoles kg^-1 year^-1, and +0.7 +5.1 µatm year^-1, respectively (Table 2; Gonzalez-Davila et al., 2003) for the 1995-2000 period. The observed rate of change of surface ocean DIC, for example, was slightly lower than the expected oceanic equilibration with anthropogenic CO₂ in the atmosphere. The causes for the lower than anticipated oceanic CO₂ increase are not certain, but probably relate to the relatively short period of observation (i.e., 1995-2000 data reported), and sub-decadal variability of the region close to upwelling off the African coast (Gonzalez-Davila et al., 2003).
 

We will continue to evaluate interannual to decadal variability in the carbon cycle of the subtropical gyres with our colleagues at Scripps (Charles Keeling) and UCLA (Niki Gruber). Interannual variability of the oceanic carbon cycle and the uptake of CO₂ from the atmosphere has been deduced from long-term upper ocean time-series records collected at stations ALOHA near Hawaii and BATS near Bermuda. In this context, we investigate the role of the accumulation (and release) of dissolved inorganic carbon in sub-surface ocean layers as reported for the North Atlantic by Bates et al. (2002), and their interplay with natural climate oscillations (like the NAO). We employ a simple diagnostic box model, a modification of (Gruber et al., 1998) to quantify the contribution of the processes controlling this variability (Brix et al., 2003; 2004). Near Bermuda, the variability in the carbon dynamics is largely driven by variations in winter mixed layer depths, which impact both the amount of DIC that gets entrained into the mixed layer and the magnitude of net community production. The variability of air-sea CO₂ fluxes tends to be controlled by sea-surface temperature (SST) anomalies with larger CO₂ uptake from the atmosphere during years of deeper than normal mixed layers. We find significant correlation of the magnitude of net community production and air-sea CO₂ fluxes with the North Atlantic Oscillation (NAO), attributed to a strong influence of the NAO on convection and SST during winter. Preliminary analyses of the HOT data indicate a much weaker role of mixed layer depth variability, and a much stronger role of SST and wind speed anomalies driving air-sea CO₂flux variations, and hence driving DIC variability. Interannual variations in net community production and air-sea CO₂ fluxes tend to be correlated, as is the case near Bermuda.