Small-scale heterogeneity of dissolved gas concentrations in marine continental shelf waters



[1] Marine continental shelf waters are known to contribute significantly to the global air-sea fluxes of many gases. Biogeochemical cycles in these regions are highly dynamic, and it is thus often difficult to fully resolve the spatial and temporal distribution of gases in the upper water column. High-frequency, real-time gas measurements with a membrane inlet mass spectrometer (MIMS) reveal significant small-scale heterogeneity in the distribution of CO2, O2/Ar ratios, and dimethylsulfide (DMS) in continental shelf waters of the Eastern Subarctic Pacific Ocean and Bering Sea. Decorrelation length scales for the gas distributions ranged from 15 to 25 km, with significant variability observed on subkilometer spatial scales. In the case of DMS, a number of rapid excursions were observed over distances that would be difficult to resolve with conventional methods. Across most of the sampling transects, CO2 and O2/Ar ratios were correlated, suggesting that biological processes dominated the cycling of these gases. In contrast, DMS concentrations were generally uncoupled from CO2 and O2/Ar, although concentrations often did change sharply across hydrographic and productivity fronts. The results presented here suggest that previous field studies may have underestimated the true spatial variability of dissolved gases (DMS in particular) in surface waters of highly dynamic marine systems. High-frequency gas measurements have significant promise for unraveling complex biogeochemical cycles in these regions.

1. Introduction

[2] Global oceanographic surveys have provided extensive information on the air-sea fluxes of a number of climatologically active gases. A strong oceanic sink for anthropogenic CO2 has been firmly established [Sabine et al., 2004], whereas net marine sources of many other trace gases including nitrous oxide (N2O), methane (CH4), and dimethylsulfide (DMS) have been identified in various regions [Owens et al., 1991; Bates et al., 1995, 1996; Kettle et al., 1999; Baker et al., 2000]. Over large spatial scales, oceanic sources and sinks have been relatively well constrained for some gases, CO2 in particular [Takahashi et al., 2002]. However, small-scale spatial and temporal variability in surface ocean gas concentrations leads to significant uncertainty in regional air-sea fluxes [Watson et al., 1991; Bange et al., 2001]. This uncertainty is particularly large for trace gases whose distribution depends upon complex hydrographic and biogeochemical interactions. For a number of trace gases such as DMS, local oceanic sources to the atmosphere are often associated with biogeochemical “hot spots” in highly productive continental shelf regions [Barnard et al., 1984; Turner et al., 1988; Matrai and Keller, 1993; Locarnini et al., 1998; Sharma et al., 1999]. Indeed, continental shelf waters are believed to play a disproportionately important role in oceanic gas cycles [Bange et al., 2001; Murata and Takaziwa, 2003; Chen et al., 2004; Thomas et al., 2004], and failure to resolve gas distributions in these regions may thus limit the accuracy of regional air-sea flux estimates.

[3] Adequate sampling of surface gases in highly dynamic coastal waters requires fine-scale spatial and temporal measurement resolution. Most current oceanic gas measurements are based upon the collection of discrete water samples, or the underway analysis of a seawater headspace gas. For gas phase measurements, underway sampling resolution is limited by the timescale of headspace equilibration [Johnson, 1999], and in some cases, by the need to cryogenically trap and concentrate analytes prior to detection [e.g., Baker et al., 2000]. Over the past decade, new aqueous-phase sensors have been developed to increase the potential measurement frequency for a number of gases including N2, O2, and CO2 [McNeil et al., 1995; Hales et al., 2004]. These instruments have provided a new level of spatial and temporal data resolution along underway sampling transects [McNeil et al., 1995] and on time series moorings [Emerson et al., 2002]. Very recently, membrane inlet mass spectrometry (MIMS) has emerged as a particularly powerful tool for high-resolution oceanic gas measurements [Short et al., 2001; Tortell, 2005]. Unlike most aqueous-phase methods, MIMS can be used to simultaneously measure, in real-time, the concentration of both major and trace gas species in seawater. Preliminary MIMS results have revealed fine-scale structure in gas distributions undetectable with other methods [Tortell, 2005].

[4] Following upon the preliminary tests of a MIMS system for open ocean gas sampling [Tortell, 2005], this report presents the first dedicated application of MIMS to the analysis of gas distributions in dynamic continental shelf waters. Measurements of CO2, O2, Ar, and DMS concentrations in the coastal Subarctic Pacific Ocean and Southeast (SE) Bering Sea reveal heterogeneity on spatial scales that have not been previously resolved. The results demonstrate that, for DMS in particular, current measurement programs do not adequately capture the true variability in surface ocean gas concentrations and the underlying biogeochemical dynamics.

2. Methods

[5] Dissolved gas measurements were made along a series of cruise tracks in the SE Bering Sea in August, 2003, and along one transect in coastal waters adjacent to the Queen Charlotte Islands, British Columbia (∼53°N, 133°W), in June of 2003. The Bering Sea transects covered several distinct hydrographic domains [see Kinder and Schumacher, 1981] in the vicinity of the continental shelf break at ∼55°N, ∼167°W (Figure 1). Sampling was conducted in the outer shelf region (∼100–250 m isobaths), the shelf break (250–2000 m), and the low-productivity waters of the deep Western gyre basin (>2000 m). These regions span a large range of phytoplankton biomass, productivity and species composition, and cross several prominent physical fronts [Loughlin and Ohtani, 1999].

Figure 1.

Location of sampling transects in the SE Bering Sea. Transect 1b is depicted with a blue line.

[6] Gas analysis was conducted using the membrane inlet mass spectrometer (MIMS) recently described in detail [Tortell, 2005]. Briefly, seawater from the ship's continuous supply was pumped through a sampling cuvette, allowing dissolved gases to diffuse across a silicon membrane into the vacuum inlet of a quadrupole mass spectrometer equipped with an electron impact ion source. Concentrations of O2, Ar, CO2, and DMS were measured sequentially by cycling the quadrupole mass filter to focus ions with m/z (mass to charge ratio) of 32, 40, 44, and 62. Each complete measurement cycle lasted approximately 25 seconds so that data were collected more than twice per minute, yielding an effective spatial resolution of ∼100–200 m along the cruise track (average cruising speeds were ∼8 knots). Instrument response time (including membrane permeation and ion detection) was typically less than 30 seconds for all gases. Measurement precision and accuracy have been assessed using a variety of standards and inter-calibrations. For DMS, CO2, and O2/Ar, measurement accuracy is better than 5%, 2.5%, and 0.5%, respectively, while the reproducibility of continuous measurements (standard deviation for a 5 sample running mean) is better than 1% for all gases.

[7] Instrument calibration was performed using DMS standards produced from the alkaline hydrolysis of DMSP [Dacey and Blough, 1987], and, for CO2, using commercial air/CO2 mixtures bubbled into temperature-controlled seawater (see Tortell [2005] for further details). Oxygen measurements were normalized to Argon in order to eliminate temperature and salinity-dependent solubility effects [Craig and Hayward, 1987]. Raw O2/Ar measurements obtained from the mass spectrometer (the ratio of ion currents at m/z 32 and 40) were expressed as a percent saturation anomaly (ΔO2/Ar [Emerson et al., 1991]) relative to the O2/Ar ratio measured in air-equilibrated seawater samples. Over the range of temperature and salinity encountered along the cruise track and used for equilibration standards, expected variation in the equilibrium O2/Ar ratio (0.05% [Garcia and Gordon, 1992; Hamme and Emerson, 2004]) was smaller than the analytical precision of measurements. Over a three week period, 15 air-equilibrated O2/Ar samples were measured with a relative standard deviation of 0.7%. Significant deviations in the O2/Ar ratios from equilibrium thus reflect biological O2 production and consumption as shown previously [Craig and Hayward, 1987]. Argon normalization is not appropriate for the other measured gases whose temperature and salinity-dependent solubility properties differ significantly from oxygen.

[8] To aid in the interpretation of underway gas data, ancillary measurements were collected along the sampling transects. Major nutrient concentrations were determined by ship-board autoanalyzer measurements on frequently collected discrete samples, while temperature, salinity, and chlorophyll a fluorescence data were obtained from the ship's thermosalinograph system. Note that chlorophyll a data were not available for all transects. Gas data and ancillary measurements were aligned by cross referencing sampling times.

3. Results

3.1. Gas Distributions in Coastal Subarctic Pacific Waters

[9] Coastal waters along the seaward side of the Queen Charlotte Islands, British Columbia are known to exhibit significant small-scale hydrographic and biological variability due (among other factors) to localized upwelling and the formation of mesoscale eddies [Crawford and Thomson, 1991; Crawford, 2002]. This spatial variability is reflected in the distribution of dissolved gases. Figure 2 shows the concentrations of CO2, O2/Ar ratios, and DMS measured during a two hour portion of a sampling transect in these waters, along with surface water salinity and chlorophyll a concentrations. In this productive coastal system, O2/Ar and CO2 were, respectively, above and below their atmospheric equilibrium values due to an apparent excess of photosynthesis over respiration. Concentrations of CO2 displayed a significant inverse correlation with O2/Ar ratios (r2 = 0.6, p < 0.0001), and these parameters exhibited a striking degree of similarity in their small-scale (subkilometer) spatial variability. Overall, CO2 and O2/Ar distributions generally followed surface water salinity gradients, and these parameters changed abruptly across a number of small salinity fronts. However, several large changes in CO2 and O2/Ar occurred in regions of nearly constant salinity (e.g., during the first 10 minutes of the sampling transect). Bulk phytoplankton biomass (Chlorophyll a) was relatively constant across the transect and did not show any clear relationship to CO2 concentrations or O2/Ar ratios.

Figure 2.

Underway measurements of (a) CO2, (b) O2/Ar, (c) DMS, and (d) salinity and calibrated chlorophyll a (Chl a) in coastal waters adjacent to the Queen Charlotte Islands, British Columbia (∼53°N, 133°W). The inset in Figure 2c shows a 10 minute portion of the transect data (from 50 to 60 minutes).

[10] Concentrations of DMS (Figure 2c) were unrelated to any of the other parameters measured along the transect. Across most of the sampling region, DMS levels exhibited relatively little variability. However, one extremely large and rapid excursion was observed during which DMS concentrations increased from 10 nM to 40 nM within the span of <5 minutes (<2 km), and returned to ∼15 nM within another 6–7 minutes (see Figure 2c, inset). This transient occurred several minutes (∼1 km) later than abrupt change in CO2, O2/Ar, and salinity.

3.2. Gas Distributions in the SE Bering Sea

[11] Following the initial application of the MIMS method in the coastal Subarctic Pacific, the technique was used to examine O2/Ar, CO2, and DMS distributions across the SE Bering Sea during late summer. This region is characterized by several hydrographic frontal systems [Kinder and Schumacher, 1981], and previous surveys have documented the existence of high spatial and temporal variability in surface water biogeochemistry [Iverson et al., 1979; Codispoti et al., 1982]. Across the outer shelf and shelf-break regions, high biological productivity is sustained by nutrient inputs derived from tidal mixing and local upwelling [Sambrotto et al., 1986]. In contrast, primary productivity in the Western Gyre appears to be limited by Fe availability, and this region exhibits the high-nutrient, low-chlorophyll (HNLC) conditions of other trace metal-limited oceanic waters [Leblanc et al., 2005; K. Bruland, personal communication, 2003].

[12] Gas measurements obtained with the underway MIMS system along transect 1 (Figure 3) clearly captured the abrupt transition from HNLC waters of the Western Gyre, into productive waters west of the shelf break. The gas and hydrographic data are plotted against longitude in Figure 3 and shown as spatial maps in Figures 7 and 8. The transition into high-productivity waters was marked by strong salinity and temperature gradients between 171.5° and 172°W (region 1 in Figures 3d and 7), and by a precipitous decrease in the concentration of CO2 and NO3 in surface waters (Figures 3a and 8a). Elevated net primary productivity was also evident by an abrupt increase in O2/Ar ratios above air equilibrium levels (Figures 3b and 8b). In the offshore waters east of 171°W, (water depth > 1000 m), CO2 concentrations exhibited relatively little variability and remained well below atmospheric equilibrium (195–225 μatm), while O2 remained significantly supersaturated (3–10%). A pronounced temperature and salinity front was encountered at the outer edge of the shelf break (∼168°W, region 2 in Figures 3d and 7). This front coincided with a small but significant decrease in CO2 concentrations and large increase in O2/Ar ratios (Figures 3b and 8b). To the east of this front, across the outer shelf domain, CO2 concentrations remained nearly constant, whereas O2/Ar ratios decreased to approach air saturation values. Across the entire sampling transect, CO2 concentrations covaried strongly with NO3 levels, and showed a general anticorrelation with O2/Ar ratios (see below for a more complete statistical analysis).

Figure 3.

Distribution of (a) NO3, (a–c) dissolved gases, and (d) salinity and temperature along a longitudinal transect (T1) in the SE Bering Sea. Numbers in Figure 3d refer to regions of rapid temperature and/or salinity changes depicted in Figure 7a.

[13] As in the case of the coastal Subarctic Pacific transect (Figure 2), the distribution of DMS across the SE Bering Sea was distinctly different from that of CO2 or O2/Ar (Figures 3 and 8). The highest DMS levels (∼15 nM) were measured in the open ocean waters (west of ∼172°), and only a small concentration change was observed across the offshore temperature and salinity gradients between 171.5° and 172°W (region 1). Significant variability in DMS levels was observed, however, over the eastern portion of the transect, with a approximately tenfold range in measured concentrations (1–10 nM). In the shelf-break frontal zone (region 2), DMS covaried with O2/Ar, CO2, temperature, and salinity, but overall, DMS levels were generally unrelated to any of these parameters across the majority of the sampling transect.

[14] Several additional transects were sampled within the open ocean domain, the outer shelf, and across the shelf-break front. The data obtained from these transects are shown in Figures 45678. Transect 1b was conducted adjacent to the shelf break frontal zone (region 2). As the ship's track was highly nonlinear for this transect, the data are plotted against sampling time rather than longitude. The salient feature of the shelf-break transect is the highly resolved, small-scale variability in dissolved gases (particularly CO2) associated with fine-scale structure in surface water temperature and salinity (Figure 4). By comparison, both transects 2 and 3 captured larger changes in surface water gas concentrations and hydrography. For transect 2, the concentrations of CO2, O2/Ar, and DMS changed rapidly in the region of sharp temperature and salinity gradients near 171.5°W (region 1; Figures 5d and 7). Moving east along the transect, gas concentrations also fluctuated significantly between ∼168° and 169°W in areas of rapid temperature and salinity changes (region 3). Transect 3, conducted within the outer shelf domain, showed the lowest CO2, and highest DMS and O2/Ar of all the sampling regions (Figure 6). The largest changes in CO2 and O2/Ar occurred in areas of abrupt temperature and/or salinity changes (regions 4 and 5). Across the western part of the transect (∼166–167°W) chlorophyll a fluorescence also covaried with CO2 and O2. In the case of DMS, the most abrupt concentration change (a threefold increase over the span of ∼10 minutes (165.55°W)) occurred in a region where CO2, O2, temperature, salinity, and chlorophyll a exhibited relatively little variability. Overall, the range of CO2 and DMS concentrations, and O2 supersaturation reported here are consistent with previous gas measurements in the SE Bering Sea [Barnard et al., 1984; Codispoti et al., 1986].

Figure 4.

Distribution of (a–c) dissolved gases and (d) salinity and temperature across the shelf-break front (T1b) in the SE Bering Sea. Note the small range in the salinity and temperature scales.

Figure 5.

Distribution of (a) NO3, (a–c) dissolved gases, and (d) salinity and temperature along a longitudinal transect (T2) in the SE Bering Sea. Numbers in Figure 5d refer to regions of rapid temperature and/or salinity changes depicted in Figure 7.

Figure 6.

Distribution of (a) NO3, (a–c) dissolved gases, (d) relative chlorophyll a fluorescence, and (e) salinity and temperature along a longitudinal transect (T3) in the SE Bering Sea. Numbers in Figure 6d refer to regions of rapid temperature and/or salinity changes depicted in Figure 7a.

Figure 7.

Spatial map of (a) temperature, (b) salinity, and (c) density distributions along the Bering Sea sampling transects.

Figure 8.

Spatial map of (a) Pequation image, (b) O2/Ar, and (c) DMS distributions along the Bering Sea sampling transects.

[15] Unfortunately, air-sea fluxes of gases could not be calculated along the measurement transects as wind speed data were not available. Gas fluxes depend upon the concentration difference between the atmosphere and surface ocean, and on a wind speed-dependent exchange coefficient [Wanninkhof, 1992]. Without calculating fluxes, it is still possible to conclude that the sampled waters of the Bering Sea and Subarctic Pacific were a clear CO2 sink given the large degree of Pequation image undersaturation (up to 200 μatm), and a significant source of DMS to the atmosphere. Although the magnitude of these gas fluxes cannot be determined, the data clearly show that potential surface water gas sources and sinks exhibited a high degree of spatial variability along the sampling transects.

3.3. Statistical and Spatial Analysis of Gas Distributions

[16] The covariance of gas concentrations and hydrographic measurements along the Bering Sea sampling transects was examined by computing a matrix of partial correlation coefficients for the pooled data set (Table 1). The partial correlation coefficients express the correlation between any two variables, after controlling for linear interactions with all other variables [Zar, 1999]. The correlation matrix reveals a number of clear statistical associations among CO2, O2/Ar and surface water hydrography. Concentrations of CO2 exhibited a strong positive correlation with NO3 levels and negative correlation to O2/Ar as is apparent in Figures 3456. Weaker, but statistically significant relationships (p < 0.05) were also evident between CO2 levels and surface water temperature and salinity (positive correlations in both cases). For O2/Ar, the strongest correlations were observed with CO2 and salinity, with a weaker correlation to DMS also apparent. Other than the correlation with O2/Ar, DMS levels were not significantly correlated with any other measured parameter.

Table 1. Partial Correlation Coefficients for the Bering Sea Gas and Hydrographic Dataa
  • a

    Degrees of freedom for the correlation coefficients were obtained by dividing the total number of sampling points by the mean decorrelation length scale (expressed as a sampling interval). This procedure yields an estimate of the number of statistically independent measurements.

  • b

    Statistical significance at the p < 0.05 level.

  • c

    Calculated for each column as the sum of all squared correlation coefficients.

CO2 −0.66b0.050.39b0.46b0.74b
O2/Ar−0.66b 0.39b0.010.75b0.25b
DMS0.050.39b −0.13−0.110.09
Temp0.39b0.01−0.13 0.13−0.70b
Sal0.46b0.75b−0.110.13 0.01
Sum of squaresc1.351.210.190.670.801.11

[17] The correlation analysis should be interpreted with some caveats. In particular, inspection of the coefficients in Table 1 reveals the presence of some residual interactions between variables which are not completely removed from the partial correlations. These residual nonlinear interactions cause the sum of squared correlation coefficients to exceed 1 for CO2, O2/Ar, and NO3 (Table 1). A full statistical description of the variable relationships would thus appear to require a more refined nonlinear model. Nonetheless, the linear correlation model can be extended through the use of a principal components analysis (PCA) in which patterns of variability can be extracted from multidimensional correlation matrices. Application of PCA to the Bering Sea gas and hydrographic data (Figure 9) reveals some patterns among the measured variables. For the pooled data set (Figure 9a) and for all of the individual transects (Figures 9b–9d), CO2 is always associated most closely with NO3 reflecting the high correlation between these variables. In addition, the strong anticorrelation between CO2 and O2/Ar was well resolved for transects 2 and 3, and to a lesser extent for transect 1. DMS appeared to fall in between CO2 and O2/Ar, and clustering tightly with salinity in transects 1 and 3, and strongly opposite to temperature in transects 1 and 2.

Figure 9.

Results of principal components analysis for the Bering Sea data set: (a) all data combined, (b) transect 1, (c) transect 2, and (d) transect 3. The axes in the factor loading plots represent the first (x axis) and second (y axis) principal components derived from the correlation matrix. The position of each variable on the plot reflects the strength of its association with each of these components.

[18] To quantify the spatial variability of surface water hydrography and gas concentrations along each transect, lagged autocorrelation functions were computed for all measured variables after applying a linear detrending procedure [Murphy et al., 2001]. The autocorrelation function expresses the correlation between any two points in a spatial series as a function of their separation (λ) along the measurement track. The first zero crossing of the function is taken as the decorrelation length scale (DLS), which represents the minimum distance at which two sampled points become statistically independent. Smaller values of the DLS represent smaller-scale spatial variability in a measured parameter.

[19] Computed autocorrelation functions (Figure 10) indicated that all measured parameters exhibited small-scale variability across the sampling transects. Decorrelation length scales ranged from 0.2° to 0.3° longitude, corresponding to a linear distance of ∼15 to 25 km. The DLS for CO2, O2, temperature, salinity, and NO3 were not significantly different from each other (Figure 9, inset; t-test, p > 0.05, df = 4). The derived decorrelation length scale for DMS was the shortest of all measured parameters, although the difference was only statistically significant relative to CO2 and NO3 (t-test; p < 0.05).

Figure 10.

Mean spatial autocorrelation functions (correlograms) for CO2, O2/Ar, DMS, temperature, salinity, and NO3. Plotted lines represent the means of 4 independently calculated functions. The inset shows the mean (± std. err.) decorrelation length scale for each parameter which is derived as the first zero crossing of the autocorrelation function.

4. Discussion

[20] The measurements presented here provide new insight into the fine-scale structure of gas distributions in high-latitude continental shelf waters. These regions are known to play a disproportionately large role in the air-sea exchange of many gases [Sharma et al., 1999; Thomas et al., 2004; Bange et al., 2001], and they are often characterized by high spatial and temporal variability [e.g., Murata and Takaziwa, 2003], which poses significant sampling challenges. The MIMS data show that CO2, O2 and DMS concentrations vary over smaller spatial scales than previously recognized, and highlight the biogeochemical difference between the surface ocean cycling of these gases.

[21] A number of previous studies have examined gas distributions in the high-latitude oceans, with a particular emphasis on CO2. Early work in the Bering Sea documented the “extreme” temporal and spatial variability of surface water CO2 [Codispoti et al., 1982] associated with small-scale variability in phytoplankton productivity [Iverson et al., 1979]. Subsequent field surveys have demonstrated significant small-scale (<100 km) heterogeneity in CO2 concentrations in productive regions of the N. Atlantic and Southern Ocean [Watson et al., 1991; Bates et al., 1998; Sweeney et al., 2000], yet few studies have quantified the spatial scales of the observed variability. Recently, Murphy et al. [2001] have computed decorrelation length scales for CO2 in open ocean waters of the Subarctic Pacific and Bering Sea (their analysis explicitly excluded measurements obtained in highly variable coastal regions). These authors report length scales ranging from 1.3° to 5.8° longitude, with the highest variability observed during periods of enhanced primary productivity in the late summer. By comparison, the observed spatial variability in the continental shelf transects sampled with MIMS (Figures 3456) was roughly an order of magnitude greater than this (Figure 10). The correlation length scales of the Bering Sea data ranged from 0.2° to 0.3° longitude (∼15–25 km). These length scales are similar to those reported previously for cross-shelf surveys of surface water temperature and plankton distributions in the Subarctic Pacific [Mackas, 1984; Denman and Freeland, 1985]. The analysis presented here thus suggests that very high frequency sampling is needed to fully resolve gas dynamics in nearshore marine waters.

[22] Small spatial-scale fluctuations in surface water gas concentrations may result from a combination of biological and physical processes. In the case of O2, biological processes can be specifically examined by following changes in the surface water O2/Ar ratio, since this ratio is largely insensitive to temperature and salinity-dependent solubility effects. The significant anticorrelation between O2/Ar and CO2 across the Bering Sea transects (Table 1, Figures 9c and 9d) can thus be interpreted as evidence of a strong biological imprint on gas distributions. Superimposed upon this biological signal, however, is the influence of water mass mixing, which is reflected in the salinity and temperature correlations seen for O2/Ar and CO2. The correlations between surface gas concentrations and hydrography likely reflect the importance of frontal zones in stimulating production over the SE Bering Sea [Iverson et al., 1979]. The position and strength of these frontal features are influenced by tidal cycles [Kinder and Schumacher, 1981], thus introducing temporal variability which is not explicitly resolved in spatial surveys [Kachel et al., 2002].

[23] A further complication is the effect of differential gas exchange which leads to an uncoupling of CO2 and O2 cycles on timescales of days to weeks. Surface water O2 concentrations equilibrate with the atmosphere ∼10 times faster than CO2 [Broecker and Peng, 1982], so that the photosynthetically derived O2 supersaturation is much more ephemeral than CO2 undersaturation. For the Bering Sea data set, rapid O2 exchange is likely responsible for the relatively weak correlation between NO3 and O2/Ar (Table 1), and the fact that the spatial autocorrelation functions for CO2 and NO3 (Figure 10) are nearly identical to each other, but qualitatively different from that of O2/Ar. Despite evidence for some uncoupling between O2/Ar and CO2 distributions, the strong overall correlation between these parameters is indicative of active net photosynthesis across the Bering Sea shelf. In particular, large simultaneous O2/Ar and CO2 excursions occurred along hydrographic fronts, where water mass mixing presumably introduces nutrients into surface waters [Sambrotto et al., 1986]. Along other portions of the sampling transects, significant CO2 drawdown occurred with only small O2/Ar supersaturation. These regions likely reflect later phases of phytoplankton blooms and/or a strong imprint of gas exchange. After accounting for gas exchange, the relationship between CO2 and O2/Ar thus provides insight into the temporal evolution of net primary production. This only applies in regions where CaCO3 producing organisms do not contribute significantly to primary production, as was the case in this study. In principle, calcification can introduce significant deviations in the CO2 – O2/Ar relationship through its effect on the surface water carbonate system [Robertson et al., 1994].

[24] Although CO2 and O2 are likely the most commonly measured gases in seawater, significant research interest has also focused on a variety of other trace gases. Dimethylsulfide (DMS) has attracted attention because of its potential role in climate regulation through the formation of cloud condensation nuclei [Charleson et al., 1987]. To date, nearly all oceanographic DMS measurements have been obtained using standard purge and trap gas chromatography (GC) methods, with a sampling frequency typically on the order of several measurements per hour [e.g., Baker et al., 2000]. A recently compiled global database [Kettle et al., 1999] has clearly demonstrated large-scale spatial and temporal variability in surface water DMS concentrations, yet significant uncertainty remains concerning the fine-scale dynamics of this gas [Belviso et al., 2004].

[25] High-frequency measurement of surface water DMS (Figures 23456) shows that this gas exhibits significant patchiness on subkilometer spatial scales, and suggests that the spatial scales of variability for DMS may be smaller than those of CO2, O2, temperature, and salinity (Figure 10). In the most extreme example, DMS concentrations along the Queen Charlotte Islands increased by 30 nM over the span of 3 minutes (∼750 m; Figure 2c, inset). This excursion was well sampled in the MIMS data set, with more than 20 individual measurements. In contrast, traditional field sampling and measurement protocols, with sampling and analysis times on the order of 10–20 minutes, would clearly fail to resolve this level of spatial heterogeneity. It therefore stands to reason that previous ship-board surveys in dynamic ocean regions may have often underestimated the true variability in surface water DMS concentrations. However, lower-frequency sampling does not necessarily lead to significant errors in the mean gas concentrations. When averaged over larger spatial scales (>100 km), the mean DMS concentration measured in the MIMS surveys (5.3 ± 4 nM) was very similar to previous estimates for the N. Pacific and Bering Sea (7.5 nM; data derived from the NOAA DMS Web server: This suggests that it is the magnitude of DMS variability rather than the mean concentration field which may be biased by low-resolution sampling.

[26] The apparent small-scale variability in DMS concentrations reflects the complex biogeochemical interactions which drive its oceanic cycling and that of its precursor dimethylsulfonioproprionate (DMSP). Previous field studies have identified many biotic variables which affect surface water DMS(P) concentrations, including the biomass, and species composition of phytoplankton and bacteria [DiTullio and Smith, 1995; Scarratt et al., 2002; Moran et al., 2003], and the rates of zooplankton grazing [Dacey and Wakeham, 1986; Wolfe et al., 1994] and viral lysis [Hill et al., 1998; Malin et al., 1998]. In addition, several abiotic factors including photolysis contribute significantly to DMS removal in the upper ocean [Toole et al., 2004], and strong correlations have been found between surface DMS concentrations and mixed layer depths [Simó and Dachs, 2002]. Recent field studies have demonstrated that the processes of DMS production and consumption are highly dynamic, resulting in surface water turnover times as short as several hours [Ledyard and Dacey, 1996; Simó and Pedrós-Alió, 1999; Merzouk et al., 2004]. It is thus not surprising that DMS levels often correlate poorly with the bulk water hydrographic variables [Kettle et al., 1999] as observed in the Bering Sea transects.

[27] While DMS concentrations are often poorly correlated to bulk phytoplankton biomass or productivity, a number of field studies have reported relationships between surface water DMS concentrations and the biomass of particular phytoplankton groups [e.g., DiTullio and Smith, 1995; Scarratt et al., 2002] that produce varying amounts of DMSP [Keller et al., 1989]. Previous field studies in the Bering Sea [Barnard et al., 1984] reported elevated concentrations of DMS in offshore waters dominated by the DMS-producing colonial haptophyte Phaeocystis pouchetii, relative to diatom-dominated waters of the continental shelf region. In agreement with this result, some of the highest DMS concentrations (∼15 nM) observed in the Bering Sea MIMS surveys were observed in the westernmost offshore waters despite lower overall phytoplankton biomass and productivity. The phytoplankton community in these offshore waters was dominated by a variety of nanoflagellates, although there was no evidence of a major bloom of either Phaeocystis or coccolithopores (calcification rates, measured by 14C incubations were always less than 5% of primary production). It thus seems that the high DMS concentrations in these waters may have been partially related to the taxonomic composition of the phytoplankton assemblages.

[28] High DMS concentrations were also associated in productive regions near several hydrographic fronts, where primary production was dominated by diatoms. Across these frontal regions, DMS concentrations covaried with CO2, temperature, and salinity, though not in a consistent manner. Moreover, principal components analysis revealed a tight association between DMS and salinity for transects 1 and 3, and a strong negative association between DMS and temperature for transect 2 (Figure 9). Previous field studies have also demonstrated large changes in DMS concentrations across frontal regions [Matrai et al., 1996; Jones et al., 1998].

[29] Despite the covariance of DMS with other measured parameters across frontal regions, the correlations were not statistically significant for the transects as a whole. For the pooled data set, the only significant correlation was observed between DMS and O2/Ar. This relationship may suggest that net DMS production is, at times, associated with early to middle phases of phytoplankton blooms, where the O2/Ar ratio exceeds atmospheric equilibrium values. It should be emphasized, however, that some of the largest changes in DMS concentrations occurred in regions where all other gases and hydrographic variables showed little variability. It is thus clear that the contribution of other biogeochemical parameters not measured in the present study must be considered before any mechanistic understanding of DMS distributions can be reached.

[30] While the underway data presented here do not allow a full interpretation of the observed DMS distributions across the sampling locations, the observation of fine-scale spatial structure provides new directions for future oceanographic studies. As it is impractical to sample all relevant biogeochemical parameters with high frequency along underway transects, the MIMS system can be used as a powerful tool for adaptive field sampling. The real-time data can be used to identify “hot spots” of elevated DMS concentrations (e.g., Figure 2) for focused process studies. By comparing a suite of hydrographic and biogeochemical variables across sharp DMS gradients, it may be possible to identify key parameters controlling the production and consumption of this gas in surface waters. Moreover, Lagrangian studies [e.g., Simó and Pedrós-Alió, 1999] could be used to sample, with high frequency, the temporal evolution of surface water gas concentrations and biogeochemistry in tracked water masses. We are currently attempting to optimize the MIMS system for the analysis of other gases such as CH4 whose fine-scale spatial and temporal distribution remains poorly resolved.


[31] The author wishes to thank Ken Bruland for generously providing space on his Bering Sea research cruise, and access to underway nutrient data, and to D. Hutchins, G. DiTullio, Peter Lee, and C. Martin for assistance at sea. This work was supported by grants from the Natural Sciences and Engineering Research Council of Canada (NSERC) to P. Tortell, and the U.S. National Science Foundation (grant OCE 032774) to D. Hutchins. Michael Bender and Jan Kaiser provided helpful criticism of the manuscript.