Daily variability of dissolved inorganic radiocarbon at three sites in the surface ocean Marine Chemistry

We report radiocarbon measurements of dissolved inorganic carbon (DIC) in surface water samples collected daily during cruises to the central North Paci ﬁ c, the Sargasso Sea and the SouthernOcean.Therangesof Δ 14 C measurements foreachcruise(11 – 30 ‰ ) werelargerthanthe total uncertainty (7.8 ‰ , 2-sigma) of the measurements. The variability is attributed to changes in the upper water mass that took place at each site over a two to four week period. These results indicate that variability of surface Δ 14 C values is larger than the analytical precision, because of patchinessthatexistsintheDIC Δ 14 Csignatureofthesurfaceocean.Thisadditionalvariabilitycan affect estimates of geochemical parameters such as the air – sea CO 2 exchange rate using radiocarbon.


Introduction
Bomb radiocarbon ( 14 C) was produced in the late 1950s and early 1960s by thermonuclear weapons testing in the stratosphere and caused 14 C levels in tropospheric CO 2 to nearly double by 1964 (Nydal and Lovseth, 1983). After 1965, levels of 14 C in the atmosphere have decreased because of gas exchange with CO 2 in the surface ocean and incorporation into the terrestrial biosphere. Maximum Δ 14 C values measured in surface water dissolved inorganic carbon (DIC) were attained in the 1970s, indicating that the turnover time of CO 2 in the atmosphere with respect to transfer to the surface ocean is ∼10 years (Druffel and Suess, 1983). Measurements of Δ 14 C in water column profiles made since 1970 have been used to calculate the inventory of bomb 14 C in various oceanic regions (Broecker and Peng, 1994;Duffy and Caldeira,1995). The timescale of modification of 14 C is of the order of years, much longer than that for temperature, which is quasiconservative over a few weeks. This means that 14 C will "remember" a mixing event from a storm, entraining colder, usually lower 14 C water for a longer time than will SST.
Daily measurements of surface DIC Δ 14 C were reported previously for sites in the North central Pacific (NCP)  and the Sargasso Sea (SS) (McDuffee and Druffel, 2007). The Δ 14 C results from the NCP in November 1985 showed more variability after a 4-day storm, but accompanying chemical and physical data were not sufficient to determine the cause of the Δ 14 C variability. Daily measurements of chemical and physical parameters at the SS site indicated a change in water mass that was coincident with an increase in variability of Δ 14 C values (McDuffee and Druffel, 2007) half way through the cruise.
We report daily surface DIC Δ 14 C values obtained for cruises to the NCP and SS sites, and a site in the Southern Ocean. We wanted to determine if the variability of surface Δ 14 C values was greater than the total uncertainty of the measurements, because of changes in the water mass that occurred during the course of each cruise. Our results highlight the fact that the surface ocean Δ 14 C signature varies by a larger amount than previously indicated by uncertainties assigned to the individual values (3-4‰). This is relevant because surface radiocarbon values are used to calculate such quantities as air-sea CO 2 exchange rate and bomb 14 C inventory in the ocean, and additional error in the radiocarbon can impart larger error into these biogeochemical parameters. cruises: Alcyone-5 from October 8 to November 5, 1985, Eve-1 from June 6 to July 4, 1987, and Avon from May 28 to June 13, 1999. Samples were collected from a single site in the Sargasso Sea (31°50 VN, 63°30 VW, 100 km southeast of Bermuda, bottom depth 4380 m) during two cruises: Hydros-6 from May 29 to June 22, 1989 and SarC from June 14 to 29, 2000. Additionally, surface samples were collected from a site in the Southern Ocean (54°S, 176°W, bottom depth 5340 m) during the Boomerang cruise from December 14-31, 1995 (salinity was only available through day 12). The DIC Δ 14 C results of depth profiles taken during the Alcyone-5, Eve-1, Hydros-6  and Boomerang cruises (Druffel and Bauer, 2000) were reported earlier.
Seawater samples were collected from 0-0.5 m depth using a plastic bucket and rope for DIC Δ 14 C and δ 13 C, and concentration ([DIC]), alkalinity and salinity measurements. Results obtained using this collection method are equivalent to those obtained using Niskin bottle collection (Druffel, unpublished data). Sea surface temperature (SST) measurements were made using a mercury thermometer (±0.2°C). Samples were collected during daylight hours, usually between 1100 and 1400 h local time. Seawater samples for isotopic, [DIC] and alkalinity analyses were poisoned with saturated HgCl 2 solution to prevent biological remineralization of organic matter.
Water samples were processed for DIC Δ 14 C analysis using conventional counting (Alcyone, Eve and Hydros cruises) (Griffin and Druffel, 1985) and accelerator mass spectrometry (AMS) (SOce, Avon and SarC cruises) (McNichol et al., 1994;Southon et al., 2004). Radiocarbon measurements are reported as Δ 14 C in per mil (Stuiver and Polach, 1977). Statistical uncertainties for the individual conventional and AMS Δ 14 C measurements were ±2.5-3.0‰; the total uncertainty determined from replicate analyses of a standard seawater was ±3.9‰. Stable carbon isotope measurements (δ 13 C) were performed at WHOI or UCI on splits of CO 2 from the processed 14 C samples with a total uncertainty of ±0.06‰.
Alkalinity and [DIC] measurements were obtained by closed vessel titration of large volume (∼100 ml) samples using an automated titration system (Bradshaw et al., 1981;Brewer et al., 1986) in the laboratory of C. Goyet (WHOI) or D. McCorkle (WHOI). Measurements were determined using a nonlinear curve fitting approach (DOE, 1994) and standardized using certified reference materials obtained from Andrew Dickson (Scripps Institution of Oceanography). The standard deviation of pairs of replicate analyses of culture water was 4 μeq/kg for alkalinity and 6 μmol/kg for [DIC]. Alkalinity measurements from the Avon and SarC cruises were high by about 25 µeq/kg due to the long storage time of samples prior to analysis (N1 year) and are not reported.
In June 1999 (Avon, 17 days on station), daily Δ 14 C measurements at the NCP site averaged 89 ± 7‰ (n = 17); values were higher during the first 6 days (average 96 ± 3‰ n = 5) and lower and more  variable from days 7-17 (86 ± 5‰ n = 12) (Fig. 1a). The δ 13 C values from days 1-6 averaged 0.96±0.11‰, and were lower and more variable thereafter (0.69 ± 0.32‰) (Fig. 1b). Data from CTD casts made during this cruise showed a shift toward higher surface salinity values between days 8 and 11 (data not shown), which is consistent with a change in Δ 14 C and δ 13 C values during this time.
Values of [DIC] were constant throughout the cruise and averaged 2043 ± 6 µmol/kg. Alkalinity and salinity measurements are not available.
3.4. Implications for surface ocean variability in DIC Δ 14 C and δ 13 C Variability of the six Δ 14 C time series, as measured by the standard deviation of the averages, ranged from ±4.4‰ (SS 1989) to ± 8.4‰ (SS 2000) ( Table 1). The range of Δ 14 C values observed for the cruises was a minimum of 11.1‰ (SS 1989) and a maximum of 29.9‰ (NCP 1999) (Table 1). We note that the two cruises with the largest ranges of Δ 14 C values, NCP 1999 and S. Ocean 1995 (25.5‰), also had the largest ranges of δ 13 C values (1.2‰ and 0.75‰, respectively).
These results illustrate that, during all six cruises, repeated sampling at the same geographic location over the course of 2-4 weeks revealed surface Δ 14 C values that varied by more than the total uncertainty of the measurement (7.8‰ 2-sigma). Changes in the upper water mass were observed during most of these cruises, as determined by temperature-salinity relationships in the CTD data sets.
The source(s) of the variability in the isotopic measurements are likely changes in vertical mixing and/or spatial heterogeneity. Fig. 3 displays Δ 14 C measurements in samples collected from the upper ocean (0-300 m) during each of the six cruises (Druffel and Bauer, 2000;Druffel et al., 2008;Druffel et al., 1992) plotted versus density (sigma-t). The average and standard deviation of all surface values for each cruise are plotted as symbols with error bars, whereas Δ 14 C values for subsurface samples (10-250 m depth) are plotted as symbols with no error bars. Data from the earlier NCP cruises in 1985 (filled triangles) and 1987 (filled circles) show a larger gradient of Δ 14 C values with depth than that from the 1999 cruise (filled squares), in large part because atmospheric Δ 14 C values in the 1980s (160-270‰) were higher than in the 1990s (95-150‰). Surface ocean Δ 14 C values were lower in 1999 (NCP) and 2000 (SS) (Fig. 3a,  b), because more bomb 14 C had penetrated deeper into the main thermocline, and were replaced by 14 C-poor waters from below, causing a smaller gradient of Δ 14 C values with depth.
It seems likely that the variability of surface Δ 14 C is the result of sampling of different water masses that are floating by a single geographic location. Most of the subsurface Δ 14 C values are slightly lower than their average surface value. The least squares fit through each data set (lines in Fig. 3a,b) suggests an inverse relationship between Δ 14 C and sigma-t for most of the cruises. This inverse relationship suggests that Δ 14 C values are higher in surface waters that have limited contact with subsurface water, e.g. areas of little or no upwelling. The exception is the Southern Ocean where mixing with subsurface waters is prevalent. This is illustrative of the concept that mixing in the ocean occurs predominantly along surfaces of constant density. Discreet water sampling provides a snapshot of DIC Δ 14 C values at a single point in time. This is in contrast to geochemical proxies, such as shells, corals, forams and varved sediments that integrate Δ 14 C values over an extended period of time (weeks to years) depending on the sampling resolution. Most of the DIC Δ 14 C data available for the world ocean has been obtained from discreet water samples, e.g. Geosecs, WOCE, TTO. The reported uncertainty for DIC Δ 14 C values is based on repeated analyses of the same water sample and generally ranges from ± 3-4‰ (Key, 1996;Key, 1997;McNichol et al., 1994;Ostlund and Stuiver, 1980;Stuiver and Ostlund, 1980). Our study shows that for surface samples, the total uncertainty of a DIC Δ 14 C value at a given site over a several week period is approximately two times the reported uncertainty (∼ 7‰).
Therefore, depending on the application, users of post-bomb Δ 14 C data should consider this short-term variability of surface ocean Δ 14 C values and factor this into their analysis. For example, assessment of the bomb 14 C inventory in the water column requires numerous Δ 14 C measurements from a given depth profile (Broecker et al., 1995). Calculation of the bomb 14 C inventory at our NCP site in 1999 reveals a value of 1.8 × 10 14 atoms/m 2 with an error (based only on the 14 C measurement error of ± 3.5‰) of ± 2.2%. Using a larger error for Δ 14 C values of ± 7‰, our uncertainty for the bomb 14 C inventory increases to ± 5%, which still is not large. Another example is how variability of surface ocean Δ 14 C values affect estimates of air-sea CO 2 exchange rate. Using a multibox isopycnal mixing model to calculate the steady state, prebomb Δ 14 C value (−43.5‰) in the surface waters of the Sargasso Sea (Druffel, 1997), the average air-sea CO 2 exchange rate is 18.9 moles/m 2 /y. To obtain a pre-bomb value one-sigma lower than this (− 47‰), an air-sea CO 2 exchange rate of 9.9 moles/m 2 /y is needed, and to obtain a value one-sigma higher (− 40.0‰) requires an average air-sea CO 2 exchange rate of 28.6 moles/m 2 /y. Doubling the error for pre-bomb Δ 14 C values (±7‰) expands the range of air-sea CO 2 exchange rate values obtained to 2.3 to 40.4 moles/m 2 /y. We need to caveat that this example is for a prebomb ocean, based on uncertainties from post-bomb surface water masses, though pre-bomb variability is likely to be equally important at locations where different water masses mixe.g., tropical Pacific and subpolar/temperate boundaries.
Monthly surface Δ 14 C values from post-bomb corals displayed a seasonal amplitude that ranged from 30-80‰ in the eastern equatorial Pacific (Guilderson and Schrag, 1998) to 10-20‰ in the subtropical Atlantic and Pacific (Druffel, 1987;Guilderson et al., 2000). Thus, the ranges of daily Δ 14 C values (11-30‰) that we measured at our three sites in the NCP, SS and SOce are comparable to the range of Δ 14 C values observed seasonally at selected sites.
In summary, our results show that a single measurement of DIC Δ 14 C in surface seawater has a larger uncertainty than that accounted for by measurement error alone. The true range of Δ 14 C values that occur over a several day-to-several week period is approximately double the measurement precision. This is due to patchiness that exists in the DIC Δ 14 C signature of the surface ocean, and the movement of surface water masses relative to geographic location.