High-resolution biogeochemical investigation of the Ross Sea, Antarctica, during the AESOPS (U. S. JGOFS) Program

Authors


Abstract

[1] The results of high-resolution biogeochemical measurements in the upper 200 m of the Ross Sea, Antarctica, obtained during the AESOPS (U. S. JGOFS) program using the Lamont Pumping SeaSoar (LPS) are presented. They consist of three west-east transects from 170°E to 180° longitude along the AESOPS study line at 76.5°S and three short north-south transects in the Ross Sea polynya during the initial and maturing stages of phytoplankton blooms in the austral spring and early summer of 1997. The LPS carried an in situ instrument array for measurement of temperature, salinity, fluorescence, photosynthetically active radiation (PAR), and dissolved oxygen. In addition, a high-pressure pump mounted aboard the LPS fish delivered a continuous seawater sample stream to the shipboard laboratory for high-speed analysis of its nutrient (nitrate plus nitrate, phosphate, and silicate) and total CO2 concentrations and CO2 partial pressure (PCO2). Vertical resolution of this sampling equaled or exceeded that of hydrostation-style conductivity-temperature-depth (CTD) casts; horizontal resolution (nominally equal to a vertical cast every 3–5 km) exceeded station resolution by a factor of 10. While not perfectly synoptic, the 20-hour duration of these transects is far shorter than the time typically taken to complete the line with conventional sampling methods. These surveys clearly identified three distinct deep water masses below about 100 m: High-Salinity Shelf Water (HSSW) in the western end of the transects, Modified Circumpolar Deep Water (MCDW) in the middle of the transects, and Low-Salinity Shelf Water (LSSW) to the east. The regions to the west were characterized by high biological productivity with high N:P and C:P uptake ratios, but little Si uptake, indicating that the production was dominated by Phaeocystis. To the east, biological productivity was lower than in the west, and low N:P and C:P uptake ratios and high Si uptake indicated the dominance of diatoms. The difference in uptake ratios appears to be entirely due to anomalously high P uptake by diatoms; N:C uptake ratios are similar in the two regions and very near canonical Redfield stoichiometry. The area in the center of the transects was characterized by high stratification and low, diatom-dominated productivity; the reason for this low productivity is unclear but is speculated to be due to the short time which this water has been exposed to an ice-free surface. The presence of strong variability at horizontal length scales of order 10 km is evident in nearly all fields especially in upper 100 m, although many of the features are resolved only by two or three LPS tracks (each separated by only a few kilometers). Physical, bio-optical, and chemical variability is observed to extend vertically from sea surface to depths as deep as 140 m far below the 1% light level and the mixed layer depth. This may be attributed to downwelling of waters associated with eddies and/or meandering. A simple statistical measure presented to quantify the errors associated with the undersampling of such a highly variable field shows that biogeochemically important parameters like PCO2 and oxygen and chlorophyll concentrations are poorly resolved by a commonly used 50-km hydro-station spacing. Longitudinal averages of these parameters, however, are predicted fairly well at coarser resolutions.

1. Introduction

[2] The Ross Sea is a partially enclosed embayment along the Pacific sector of the Antarctic coast, located due south of New Zealand (Figure 1). It is bound to the west by Victoria Land, to the south and east by Edward VII Land, and is bordered to the north by the continental slope. It covers an area of a little less than 106 km2; about half of this is permanently covered by the Ross Ice Shelf (RIS) to the south. The southwestern portion of the Ross Sea north of the RIS consistently has lower sea ice coverage than areas to the north and east; this polynya is the largest in coastal Antarctica and is one of the most productive waters of the global oceans. It exists because strong katabatic winds blowing from the Antarctic continent move sea ice away from the region and concentrate it offshore to the northeast. While ice coverage is always lowest to the southwest, the polynya begins to expand as insolation increases in late October each year, and by late January, most of the Ross Sea north of the RIS is ice-free. (The preceding discussion is derived from Smith et al. [2000a], Dunbar et al. [1998], Emery et al. [1997], Comiso et al. [1993], Sullivan et al. [1993], Jacobs and Comiso [1989], and Zwally et al. [1985].) Our study area is located within the polynya along 76.5°S in a nutrient-rich continental shelf environment with water depths of 300–700 m. The Ross Sea shelf area is also an important region of deepwater formation as Circumpolar Deep Waters (CDW) move onto the shelf and are modified by cooling and freshening during subsurface interaction with the RIS, and by gaining salt through brine rejection by sea ice formation over the continental shelf [Orsi et al., 1999; Jacobs and Giulivi, 1999; Jacobs et al., 1985, 1970].

Figure 1.

Map of the Ross Sea study area. Tracks of the four Lamont Pumping SeaSoar (LPS) surveys superimposed as thick black lines centered on the 76.5°S latitude line between 170°E and 180° longitude.

[3] Because of the persistence of the Ross Sea polynya and associated high productivity in the austral spring and summer, because of the recognition of the high nutrient regions of the Southern Ocean as an important region in determining the oceans' response to increased atmospheric CO2 [Sarmiento et al., 1998], and because of its close proximity to the McMurdo Station support facility, the Ross Sea has been a region of frequent study, including Research on Ocean/Atmosphere Variability and Ecosystem Response in the Ross Sea (ROAVERRS [Arrigo et al., 1999]); the Climatic Long-term Interaction for the Mass balance in Antarctica (CLIMA [Spezie and Mansella, 1999]), the Ross Sea Flux Experiment [Marra, 1996], and the Antarctic Environment and Southern Ocean Process Study (AESOPS) of the Joint Global Ocean Flux Study (JGOFS [Smith et al., 2000a; Anderson and Smith, 2001]).

[4] The primary purpose of our study is to document the distribution and spatial/temporal variability of biogeochemical properties of the upper 200-m layers of the Ross Sea polynya during the early stages of phytoplankton blooms in 1997 with a horizontal resolution of 2–5 km. Our study covered the early bloom period, spanning the transition from high-productivity low-biomass to full high-biomass expression of the bloom. This paper shows the results of the first deployments of the Lamont Pumping SeaSoar (LPS) [Hales and Takahashi, 2002], a system designed to couple the high-resolution surveying capabilities of the SeaSoar towed undulating vehicle (SeaSoar, Chelsea Instruments, UK; www.chelsea.co.uk) with a pumped sample stream. High-frequency chemical analyses of the pumped sample stream were performed in the shipboard laboratory to determine nutrient concentrations, following Hales et al. [2004], and total CO2 (TCO2) concentrations and CO2 partial pressure (B. Hales et al., High-frequency measurement of partial pressure and total concentration of carbon dioxide in seawater using microporous hydrophobic membrane contactors, submitted to Limnology and Oceanography, 2004) (hereinafter referred to as Hales et al., submitted manuscript, 2004). We participated in the Ross Sea Process IV study of the JGOFS AESOPS program, sailing on the RVIB N. B. Palmer cruise 97-08 (from Lyttleton, New Zealand, 5 November 1997, to the McMurdo Station, 13 December 1997). The corrected and gridded data from this investigation are archived at the JGOFS Data Center at the Woods Hole Oceanographic Institution, and may be accessed through the Center (http://usjgofs.whoi.edu/jg/dir/jgofs/southern/nbp97_8/; then click on “SeaSoar”; the data is also available at http://cdiac.ornl.gov/oceans/datmet.html).

2. Biological and Chemical Environments

[5] High biomass and high productivity associated with the opening of the polynya is a regular feature of austral spring and summer, and makes the Ross Sea one of the most productive regions of the Southern Ocean. There is a well-defined temporal progression from pre-bloom, low-biomass, and low-productivity winter conditions to a period of high-productivity when the phytoplankton seed stocks take maximal advantage of the juxtaposition of high nutrient concentrations, greater insolation, and meltwater-induced stabilization of the upper water column. A period of high biomass and low productivity will follow as production consumes available macro- and micro-nutrients and insolation decreases. Finally, conditions return to winter environments as biomass is degraded and deep mixing resumes. (The preceding discussion is summarized from Smith et al. [2000a, 2000b], Arrigo et al. [1998a, 1998b], Smith and Gordon [1997], Smith et al. [1996], Arrigo and McClain [1994], Comiso et al. [1993], Sullivan et al. [1993], Demaster et al. [1992], Wilson et al. [1986], and El-Sayed et al. [1983]). The bloom is dominated by two different phytoplankton assemblages: One consists primarily of Phaeocystis antarctica and the other consists of a variety of Fragillariopsis, Thalassiosira, and Nitschia diatom species. It has been proposed that the former assemblage favors poorly stratified open-water environments, whereas the latter dominates in ice-edge and more stratified environments [Arrigo et al., 2003, 2000, 1999, 1998a, 1998b; DiTullio and Smith, 1996; Leventer and Dunbar, 1996; El-Sayed et al., 1983]. The bloom termination in late summer seems due to depletion of micro- or macro-nutrients, or onset of light limitation or water column destabilization, rather than grazing, as evidenced by the low levels of bacterivory [Ducklow et al., 2001, 2000] and microzooplankton herbivory activities [Tagliabue and Arrigo, 2003; Dennett et al., 2001; Caron et al., 2000; Lonsdale et al., 2000]. The results of many studies indicated that phytoplankton growth in the Southern Ocean is limited by the availability of micro-nutrients, especially of iron [e.g., Arrigo et al., 2003; Fitzwater et al., 2000; Martin et al., 1990; Olson et al., 2000; Sedwick et al., 2000; Sedwick and DiTullio, 1997]. Coale et al. [2003], however, pointed out that in contrast to the Antarctic Circumpolar Current area, the southern Ross Sea (where our study was conducted) was not severely iron stressed, particularly early in the growing season. The photosynthetic products are mostly in the form of particulates, and about 10 mole% of reduced carbon present as dissolved organic matter [Carlson et al., 2000, 1998]. Biogenic particles sink out of the mixed layer either individually or as aggregates [DiTullio et al., 2000; Accornero et al., 1999; Asper and Smith, 1999; Dunbar et al., 1998].

3. Chemical Properties of the Ross Seawater

[6] The chemistry of the Ross Sea surface water is characterized by abundant winter-time macronutrients, with silicate, nitrate, and phosphate concentrations being about 75, 30, and 2 μmol kg−1, respectively, distributed homogeneously with depth and longitude. These are substantially reduced in surface waters through the austral spring and summer seasons by the high rates of biological productivity to much lower levels, <60, <10, and <0.7 μmol kg−1, respectively, in the surface, and this reduction shows variability as a function of longitude [Gordon et al., 2000]. Ammonium is initially near zero throughout the Ross Sea, but by middle to late summer, areas with concentrations exceeding 2 μmol kg−1 can be found. Carbonate chemistry is defined by high wintertime total CO2 (TCO2) concentrations and CO2 partial pressures (as high as 2260 μmol kg−1 and 440 μatm, respectively), and, as a result of phytoplankton blooms, both properties are reduced in surface waters to <2100 μmol kg−1 and <200 μatm, respectively [Gordon et al., 2000; Sweeney et al., 2000a]. By early fall, much of the high original wintertime nutrient and TCO2 concentrations are restored by remineralization of biogenic particulates within the mixed layer and/or the upward mixing of subsurface waters which contain excess CO2 and nutrients from deeper remineralization of organic particles. Some portion of the biogenic particles are presumably exported to depth and respired later in the year [Sweeney et al., 2000a]. Most of this biogenic material exported to depth is remineralized in the deep water column and within the sediments [Nelson et al., 1996], and the decomposition products are returned to the overlying water column. Dissolved organic carbon (DOC) and ammonium increases were not likely significant factors in the carbon and nitrogen budgets [Carlson et al., 2000; Gordon et al., 2000]. Although there is some continual resupply of the surface waters' carbon and nutrients by mixing from below [e.g., Sweeney et al., 2000b], our early season measurements of these concentrations, and their depletions from their wintertime values, should be qualitatively comparable to biological standing stocks and uptake from the beginning of the growing season. Trace metal concentrations in general parallel the seasonal and depth variation of the macronutrients, but may be additionally contributed from melting of sea ice and RIS [Coale et al., 2003; Edwards and Sedwick, 2001; Fitzwater et al., 2000; Sedwick et al., 2000; Sedwick and DiTullio, 1997].

[7] The two ecological assemblages dominated by Phaeocystis or diatom impact the chemistry of the water in which they grow with distinctly different signatures: Waters in the diatom-dominated regions experience significant silicate depletion, while those in the Phaeocystis-dominated regions do not, and the net utilization ratio of nitrate to phosphate is significantly lower than Redfield stoichiometry [Redfield et al., 1963] in diatom-dominated areas and significantly higher than Redfield in Phaeocystis-dominated areas (Arrigo et al. [2002], Sweeney et al. [2000b], Arrigo et al. [1999], and Bates et al. [1998] first noted the different N:P uptake ratios, but did not attribute this to plankton assemblage differences), although the difference decreases as the growing season progresses [Rubin, 2003; Sweeney et al., 2000b]. This may be due to water mixing as well as changes in the ratio of gross and net productivity [Brezezinski et al., 2003].

[8] Studies of mesoscale (order 10–100 km in the horizontal) and sub-mesoscale variability in the Southern Ocean in general, and the Ross Sea in particular, have been limited to observations of short length scale variability of physical and bio-optical parameters at fronts [Abbott et al., 2001, 2000; Barth et al., 2001; Muench, 1990] and at the retreating edges of melting sea ice [Asper and Smith, 1999; Sedwick and DiTullio, 1997; Sullivan et al., 1988; Nelson and Smith, 1986; Smith and Nelson, 1985; Smith and Gordon, 1997; Smith et al., 1996]. However, none of the studies have addressed short length scale chemical variability, and none have addressed short length scale variability within the Ross Sea polynya far from its marginal meltwater inputs.

4. Methods

4.1. Lamont Pumping SeaSoar (LPS) System

[9] We performed four 24-hour surveys of the upper water column (shallower than 200 m) between 25 November and 7 December using the Lamont Pumping SeaSoar (LPS) along the 76.5°S latitude AESOPS study line between 170°E and 180° longitude (Figure 1). The first of these transects followed by only a few days a strong katabatic wind event where winds were sustained at nearly 60 knots for about 24 hours, but conditions were fairly placid during the remainder of the time. The LPS is described in detail elsewhere [Hales and Takahashi, 2002]; briefly, a SeaSoar towed undulating vehicle carrying a CTD and sensor array was outfitted with a positive displacement high-pressure pump which delivered seawater samples continuously through a Nylon tube imbedded in the tow cable to the shipboard laboratory for rapid chemical analyses. Shipboard-analyzed samples were corrected for flow lag through the tube by comparing salinity traces at the in situ intake and the shipboard outlet and phase-shifted to coincide with in situ measurements. The LPS was towed at about 6–7 knots, and oscillated between about 15 m and 200 m depths at dive and climb rates of about 0.3 m s−1, and the resulting horizontal spacing between successive ascents and descents was approximately 2.5 km. While there is potential for undersampling the surface mixed layer by only sampling to minimum depths of 10–20 m, we verified that we had in fact reached the mixed layer by the good agreement between temperature, salinity, and PCO2 as seen by the LPS at its shallowest depths, and the same parameters measured in the ship's surface uncontaminated seawater intake, located at the bow and at a depth of ∼3 m.

4.2. CTD and Other In Situ Measurements

[10] In situ data were collected using a SeaBird 9+ CTD system aboard the LPS. The system consisted of two temperature-conductivity (TC) sensor pairs, with seawater drawn through each by one of two SeaBird pumps. The sensor array also included a SeaBird Beckman DO sensor for measuring dissolved oxygen, a WetLabs (www.wetlabs.com) WetStar fluorometer, and a Biospherical (www.biospherical.com) QSP200L photosynthetically active radiation (PAR) sensor. Data were collected at a rate of 12 Hz, and reduced to 1-Hz frequency by taking medians of bins of 12 data. Salinity was calculated from measured temperature and conductivity, taking into account the time-lag between temperature and conductivity and the thermal mass of the conductivity cell. Density anomaly, or σt, was calculated from temperature and salinity using the International Equation of State for seawater [Millero et al., 1980] at one atmosphere to determine density, and subtracting 1000 kg m−3. The deepest measurements were made at pressures of about 200 dbar; the difference between σt and σΘ over this pressure range is negligible. Irradiance was calculated from sensor output using factory-supplied calibration coefficients. Fluorescence was logged simply as fluorometer voltage (0–5 V scale) rather than chlorophyll concentration because fluorescence yield as a function of species distribution and light conditions was uncertain and much higher than factory calibrations. Oxygen concentrations were initially estimated using the factory-supplied calibration algorithm. Salinity, fluorescence, and oxygen concentrations were ground-truthed and corrected by collecting check samples at the shipboard end of the sample water stream; this procedure is discussed below.

4.3. Ground-Truthing

[11] This combination of a pumped sampling system with an in situ sensor array allows unique opportunities to verify the accuracy of many of the in situ measurements through the collection of check samples and their independent analyses. We drew water samples for verification of the salinity, oxygen, and fluorescence-based chlorophyll measurements. Salinity samples were always within a few parts per million (order 0.001 on the practical salinity scale) of the in situ measurements (Figure 2), and further correction was deemed unnecessary. Oxygen measured in check samples, however, was as much as 10% higher than calculated from in situ sensor measurements, and we felt it was necessary to correct it. Linear regression of measured concentrations against in situ measurements (Figure 3a) showed a very tight correlation (r2 = 0.99), and we corrected the oxygen data using this regression equation, finding very good agreement between surface and in situ measurements even for fine structure in the in situ data (Figure 3b). Fluorometer voltage was calibrated to give in situ chlorophyll concentration through regression against check samples (Figure 4a) collected from the LPS sample line throughout the sampled depth range, and was shown to be consistently and simply related to chlorophyll concentrations, even through the subsurface fluorescence maxima discussed in later sections (Figure 4b). The accuracy of nutrient samples was also verified in this way; see section 4.4. Additional verification of the accuracy and consistency of our measurements and sampling procedure is provided by the good agreement of shipboard surface water measurements discussed in section 4.1 and section 4.5.

Figure 2.

Verification of in situ salinity measurements. Solid line is salinity (versus time) calculated from in situ measurements of temperature and conductivity made aboard the LPS; large solid squares are measurements on check samples drawn from the ship board end of the pumped sample stream.

Figure 3.

Correction of in situ dissolved oxygen measurements, using discrete check samples drawn at the ship-board outlet of the LPS sample line synchronized with the in situ measurements. (a) Regression of all check samples versus corresponding in situ measurements. (b) Illustration of one sampling sequence covering a single up-down cycle of the LPS (elapsed time about 30 min) and several check samples. The dashed black line is oxygen calculated from measured electrode temperature and current using the factory-supplied calibration algorithm; light blue symbols are oxygen concentrations measured by Winkler titration on check samples drawn from shipboard end of the sample stream. The dashed blue line is calculated from a linear regression of in situ oxygen and check samples (Figure 3a).

Figure 4.

Calibration of in situ fluorometer. (a) Regression of all check samples versus corresponding in situ measurement of fluorometer voltage. (b) Illustration of one sampling sequence covering a single up-down cycle of the LPS (elapsed time about 30 min) and several check samples. The dashed blue line is calculated from a linear regression of flourometer voltage with the chlorophyll-a concentrations in check samples (Figure 4a). The solid green symbols in Figure 4b are chlorophyll-a concentrations measured in check samples drawn from the shipboard end of the sample stream, and the dashed green line is chlorophyll-a calculated from the regression shown in Figure 4a.

4.4. Calculation of N2, Apparent Mixed Layer Depths, and 99% Light Attenuation Levels

[12] Buoyancy frequency (N2) was calculated from the ungridded sigma-t data. Derivatives of density versus depth were determined by calculating best fit (least squares) linear slopes of moving 3-m windows of sigma-t versus depth. Three-meter windows were chosen because they reduced differentiation noise to acceptable levels while maintaining as much vertical resolution as possible. N2 was calculated by normalizing these derivatives to the actual density and multiplying by gravitational acceleration (g) and dividing by the Coriolis parameter (f). The apparent mixed layer depth (ZAML) was defined as the shallowest “significant” maximum in N2. This significant maximum was operationally defined as having at least five consecutive values of N2 greater than 10−5 s−2. This operational definition minimized the contribution of noise in the differentiated data to the definition of the ZAML. This definition was used rather than an absolute change in sigma-t because the LPS fish did not always reach shallow enough depths to unequivocally capture the surface density value, and the data from the temperature and conductivity sensors in the ship's surface water intake did not always agree with those aboard the LPS fish to allow reliably calculate mixed-layer depths based on a small difference between surface and in situ density. Because the LPS did not routinely reach depths shallower than 10–20 m, this calculation of ZAML will not capture very shallow mixed layer depths. The effective euphotic depth (ZEEZ) was determined by calculating an attenuation coefficient from the irradiance data in the upper 5 dbar of data sampled (centered normally at depths from 10 to 20 m), and using this coefficient to estimate at what depth 99% of incident light would be absorbed (equivalent to the 1% light level).

4.5. Nutrients

[13] Concentrations for macronutrients were determined using a Lachat QuikChem 8000 Autoanalyzer modified to run nitrate once every 12 s, and phosphate and silicate every 24 s; the full method is described elsewhere [Hales et al., 2004]. Calibrations were run every 2 hours, and calibration functions were interpolated between bi-hourly runs and applied to the raw data. Accuracy was verified by periodically drawing discrete samples for analysis by routine JGOFS analytical procedures [Gordon et al., 2000]. Nitrate and silicate concentrations determined by the high-speed system were in good agreement with the discrete analyses; phosphate, however, was offset by about 7%. Roughly half of this offset was due to a discrepancy between the concentrations of our standard solutions and those measured by the discrete methods. Even when correcting for this difference, the phosphate concentrations measured in the discrete check samples are still 3–5% higher than those measured by the high-speed continuous system. We cannot explain the discrepancy. Although the fault probably lies with us, we felt it was appropriate to present our data as we determined them, rather than adjusting the entire data set to match the few discrete analyses. Whereas the precision of measurements was satisfactory (∼1%) and the relative variability shown in the results section is robust, the absolute value of the phosphate data presented here must be considered to be uncertain by 7% relative to the results obtained by the standard JGOFS protocols [Gordon et al., 2000].

[14] Despite the good agreement between the high-speed continuous measurements of silicate and the few discrete check samples, the high-resolution silicate measurements were affected for some instances due to the extremely poor temperature control in the laboratory assigned to us aboard the N. B. Palmer. These are not unexpected consequences of our high-speed analysis method, in which the readings were taken before the chemical reactions in coils reached completion. This procedure causes greater sensitivity to reaction rate and therefore room temperature fluctuations. This effect was not present in the phosphate channel because its primary reaction took place in a heated, thermostated reactor, and was small for the nitrate system because of the faster kinetics of those reactions. The thermal artifact in the silicate data was corrected by examining the ratios of silicate to phosphate at depth. For all transects, the ratio of silicate concentration to phosphate concentration in deep water with T < −1.5°C was found to be about 41.5, with scatter related to thermal variability. We therefore used the phosphate concentration to determine correction factors to apply to the silicate data, based on deviations of the silicate measurements from the product of this average ratio and the phosphate concentration. Correction factors found in this way were in the range 0.95–1.05. Because the absolute value of the silicate concentrations determined in this way are tainted by this correction procedure, we present only the depletion of silicate relative to local deep waters rather than the absolute concentrations. This was determined by finding the average silicate concentration for the portion of each dive/climb cycle below a depth of 140 dbar, and subtracting the individual measurements from these deep concentrations.

4.6. Carbonate Chemistry

[15] The partial pressure of dissolved CO2 (PCO2) in the pumped sample stream was determined by a technique described more thoroughly elsewhere (Hales et al., submitted manuscript, 2004). Briefly, the seawater sample stream flowed through a membrane contactor unit based on a micro-porous, hydrophobic membrane material, and a carrier gas was flowed counter to the sample stream on the opposite side of the membrane. CO2 concentration in the equilibrated gas stream was measured by non-dispersive infrared (NDIR) absorbance, and calibrated using five reference gases consisting of a mixture of air and CO2 ranging from 100 ppm to 800 ppm. The CO2 concentration for each reference gas mixture is tied to the manometric measurements of C. D. Keeling. PCO2 data was collected once every 2 s, and the response time of the system was determined in laboratory experiments to be about 3 s. PCO2 accuracy was verified by consistency of LPS-determined near-surface water values with those obtained using the large-volume showerhead equilibrator plumbed to the ship's surface underway uncontaminated seawater intake.

[16] The total dissolved CO2 (TCO2) concentration of seawater was determined by injecting small-volume (1 mL) loops of seawater into an acidified carrier stream, and then measuring the partial pressure of CO2 generated in the acidified stream with a NDIR absorbance method similar to that used for PCO2. Sample loops were injected once every 36 s. Precision of this system was adequate (0.2%), but direct comparisons of two supposedly identical systems (using Dickson certified reference material) revealed an unexplained 1% difference in accuracy between the two systems. As a result, we use the high-speed TCO2 measurements only to quantify the relationship between relative changes in PCO2 and TCO2 (the Revelle Factor).

4.7. Gridding

[17] Grids with 1-m vertical and 500-m horizontal resolution were created, spanning a depth range of 5–205 dbar and a longitude range of 170°E–180°. Grid points bracketed by data within 12 m and 10 km in the vertical and horizontal, respectively, in both depth and longitude were filled with a normalized-distance-squared weighted average of the bracketing data. The normalized distance of each bracketing datum from the bracketed grid point was calculated by dividing the horizontal distance by 500 m and the vertical distance by 2 m and 5 m for the in situ and shipboard data, respectively. The more coarsely vertically resolved surface nutrient data were interpolated to 1-m resolution along the sampling track before completing the grid filling exercise. Data were gridded for the sake of presentation purposes only; no calculations (e.g., of N2, ZAML or ZEEZ or undersampling errors) were performed on the gridded data.

4.8. Estimation of Errors Due to Undersampling

[18] We calculated the errors due to undersampling using a simple representation of the distributions of our highly spatially resolved measurements that would result from coarser-resolution measurements. To do this, we interpolated the LPS data at uniform spacing equivalent to the minimum spacing of the LPS (∼0.5 km at 20 m), and then resampled that data at ever coarser resolutions ranging from that corresponding to the nominal LPS spacing (∼2 km) up to spacing about 2 times coarser than the AESOPS station spacing (∼100 km). The coarse-resolution resampled data were then linearly interpolated to the spacing of the LPS data, and average absolute differences between selected measured parameters at the two resolutions were calculated for each resampling resolution. These errors should be thought of as those expected in estimating unsampled data by interpolation between sampling locations.

5. Results

5.1. Physical Observations

[19] Figure 5 shows three sections of temperature, as recorded by in situ sensors aboard the LPS, along the 76.5°S study line, each completed in approximately 20 hours on 25 November, 2 December, and 7 December. Perhaps the most striking features are the strong subsurface temperature maxima between about 174°E and 177°E observed in all three transects. In all cases the warmest water is at depths and is removed from the surface. The western boundary of this region is a strong thermal front separating colder waters to the west, and coincides with a local steepening of the bottom-depth slope; to the east the boundary is more diffuse, and the waters east of there are intermediate in temperature between the western and central waters. Temperature variability is overall quite low, ranging from a minimum of about −1.9°C to a maximum of only about −0.6°C. Despite the small range in temperatures over the course of this study, the variability over short length scales is quite high relative to the total variance of the data. There are several well-defined features in all three transects that are sampled by only two or three up/down tracks of the LPS; given this spacing, some of these features have horizontal scales that appear to be as short as a few kilometers. The expected temporal pattern of warming surface waters as the spring season progresses into summer is evident. The extent to which this warming has penetrated to depth is complicated by the temporally varying contribution of the aforementioned warm subsurface water, which appears to strengthen and become more clearly defined between the first (25 November) and second (2 December) transect, but weaken and become more diffuse between the second and third (7 December).

Figure 5.

Depth and longitudinal distributions of temperature along the 76.5°S study line during three surveys on (top) 26 November, (middle) 2 December, and (bottom) 7 December 1997. The dashed lines overlaid on the sections show the path of the LPS and demonstrate data density. Bathymetry (black solid line, bottom panel) was taken directly from the N. B. Palmer's shipboard echosounder.

[20] Figure 6 shows the salinity and density structure along the same three sections. Total variability is small, with a range of about 34.2 to 34.7. Demonstrating the dominant effect of salinity on density at these low temperatures are the overlain contours of sigma-t, which closely track the salinity variations. In all three transects, the saltiest, densest water is to the west (the left-hand side of the figure), and the freshest, least-dense water is in the center, with water of intermediate character in the east (the right-hand side). The warm subsurface waters identified in the previous figure are also lower in salinity than the waters to the east or west. The surface waters with lowest salinities (hence the lowest densities) are situated directly atop the warm, fresh subsurface water in the center of the region. This fresh surface water is hereafter referred to as the surface salinity minimum (SSM). This separates two hydrographically distinct regions to the east and west. Throughout the section, both salinity and density decrease over this time, probably due to ice-melt. On both sides of the warm, fresh water in the center of the study region are strongly sloping isopycnals, implying that some dynamic mechanisms must be maintaining this density structure.

Figure 6.

Salinity distributions for the same three surveys as shown in Figure 5. The labeled, white curves are isopycnal (presented as sigma-t). The density contours overlain on the sections show the close relationship between density and salinity in these cold waters, indicating that the density is governed primarily by the salinity. The dashed black lines show the depth of the bottom of the mixed layer, defined as the depth of the shallowest observed pycnocline (ZAML; see section 4). Bathymetry (black solid line, bottom panel) is as in Figure 5.

[21] Plots of temperature versus salinity for the three transects (Figure 7) show clearly the contribution of three previously defined subsurface water masses. Both high-salinity shelf water (HSSW) and low-salinity shelf water (LSSW), as summarized by Carmack [1990], are evident. As the warm subsurface water in the center of the transects is distinct from the insolation and meltwater influenced surface waters, it can only be attributed to a mixture of the warm Modified Circumpolar Deep Water (MCDW) with the HSSW and LSSW. MCDW is often seen on the continental shelf of the Ross Sea [Jacobs et al., 1985; Carmack, 1990], and plays an important role in formation of waters that ultimately sink from the Ross Sea and become Antarctic Bottom Water (AABW [Jacobs et al., 1985]).

Figure 7.

T-S diagrams for the three sections, denoting the positions of high-salinity shelf water (HSSW), low-salinity shelf water (LSSW), and modified circumpolar deep water (MCDW), schematically with the correspondingly labeled ovals. Symbols are colored as pressure (see color bars) to distinguish the warming and freshening of surface waters by increased insolation and ice-melt from the different deep-water properties.

[22] Two pieces of relevant information can be extracted from the density distributions. The first is stratification, or analogously buoyancy frequency (N2) (see section 4). Stratification is generally weak, with maximum values of N2 only rarely reaching 10−4 s−2. In the upper 50 m, as expected from the temperature and salinity distributions, stratification is weakest at the western end of the line and highest in the middle, with intermediate values to the east. Strongest stratification is consistently situated atop the warm subsurface water. Stratification strengthens over time as the influence of meltwater and increased insolation warms and freshens the surface waters. Overlain on the salinity and density sections are the apparent mixed layer depths (ZAML; thick dashed line in Figure 6), defined as the depth of the shallowest significant maximum in N2. As expected, mixed layers are shallowest in the SSM zone, deepest and stratification weakest west of the SSM, and intermediate to the east (Table 1). Overall stratification increases with time, and concurrently the ZAML shoals and horizontal variability in this depth decreases. The strength of stratification at the base of the apparent mixed layer mirrors its depth, with weaker stratification coincident with deeper mixed layers.

Table 1. Mixed Layer Depths (ZAML, m), Euphotic Zone Depths (ZEEZ, m), and Buoyancy Frequencies (N2, s−2) at ZAML for East, West, and SSM Surface Waters on the Three Transects
DateWestSSMEast
ZEEZZAMLN2ZEEZZAMLN2ZEEZZAMLN2
27 Nov.29 ± 549 ± 283 ± 245 ± 540 ± 133 ± 237 ± 650 ± 303 ± 2
2 Dec.28 ± 434 ± 173 ± 242 ± 725 ± 69 ± 537 ± 626 ± 84 ± 3
7 Dec.35 ± 926 ± 74 ± 340 ± 1117 ± 412 ± 844 ± 920 ± 56 ± 3

[23] The second information that can be calculated from the density distributions is the geostrophic velocity, which can be determined by density gradients along isobars. While this calculation probably does not yield accurate absolute velocities without some direct current velocity measurement (e.g., by ADCP), it does give insight to horizontal shear rates. This exercise revealed near-zero average velocities (relative to an assumed reference depth of 180 m, chosen to facilitate possible future synthesis of ADCP-based velocities within the LPS-sampled field), but higher velocities on the east and west edges of the MCDW corresponding to the sloping isopycnals and strong lateral density gradients there. Maximum relative velocities on the western side of this feature were the most intense, reaching 20 cm s−1 with a southward direction, while those on the eastern edge are less intense, only about 10 cm s−1 to the north. Apparently, anticyclonic (or counter clockwise) circulation around the SSM-MCDW zone, possibly an eddy or a meandering filament, supports the observed lateral density gradients. However, we cannot distinguish between these two circulation patterns with only two-dimensional coverage.

5.2. Bio-Optical Observations

[24] Chlorophyll concentrations, determined from calibrated in situ fluorescence measurements, are shown in Figure 8. Concentrations are generally high, at least as high as 6 μg L−1 in the later two surveys, but as high as 3 μg L−1 even in some portions of the first. Highest overall concentrations and deepest penetrations are in the western end of the study area; lowest surface concentrations are in the central SSM part of the study area over the MCDW. Lowest overall concentrations, less than 0.1 μg L−1, are found within this subsurface MCDW mass. In the west side of the region, chlorophyll concentrations in the upper waters are high to the shallowest depths sampled; in the eastern portion of the region, and even some parts of the western portion in the final transect, chlorophyll concentrations are highest below the surface. On the basis of the chlorophyll check samples as discussed in section 4, we believe that these subsurface maxima are truly maxima in chlorophyll concentrations, as opposed to some decoupling of the relationship between chlorophyll concentration and fluorescence. Further distinguishing the eastern and western sections is the presence of vertical finger-like high-chlorophyll features penetrating to depths of 100 m or greater west of the SSM, but their absence from waters to the east. In addition to the large observed range in chlorophyll concentrations, there is strong variability over short horizontal length scales. Like temperature, there are features in the chlorophyll data that have horizontal scales in a range from only a few kilometers to 25 km. In the Ross Sea area, the water depth is about 500 m and the water column is weakly stratified. The Rossby radius of deformation “a” may be estimated to be 10–20 km using a = (g′ H) equation image/f, where g′ is the reduced-gravitational acceleration, g (ρo − ρ1)/ρo, H is the water depth, and F is Coriolis parameter [Cushman-Roisin, 1994]. The observed variability of temperature and biologically mediated properties is consistent with the Rossby radius. This suggests that the biological variability is caused by meso- and submeso-scale ocean dynamics.

Figure 8.

Chlorophyll concentration on the same sections, determined from calibrated in situ fluorescence measurements. The equivalent euphotic zone depth (ZEEZ, the 1% light level or 99% attenuation depth, determined as described in section 4) is denoted by the dashed yellow line; density contours (white solid lines) and ZAML (black dashed line) are as in Figure 7. Bathymetry (black solid line, bottom panel) is as in Figure 5. Density contour labels are omitted from this figure for clarity; refer to Figure 5.

[25] Quite noteworthy in these chlorophyll distributions is their apparent disregard for physical environmental conditions such as those represented by the apparent mixed layer depth (ZAML, overlain black dashed line) and euphotic zone depth (ZEEZ, dashed yellow line), as summarized in Table 1. The ZEEZ appears to have shoaled slightly through time; however, ZAML has shoaled faster, with the result that ZAML > ZEEZ across the entire region during the first transect, but ZEEZ almost always exceeds ZAML by the time of the second survey. Euphotic depths were deepest, and exceeded ZAML by the most, in the low-productivity, low-salinity SSM water in the middle of each transect. Despite these apparently favorable conditions of a stable, well-illuminated mixed layer, the SSM water is always the lowest chlorophyll region of the surface water. In contrast, ZEEZ was shallowest, and most likely to be equal or less than ZAML in the high-chlorophyll regions west of the SSM. High chlorophyll concentrations do not seem limited to the depths shallower than ZAML or ZEEZ; in fact, high chlorophyll concentrations appear even at depths far below either of these. While the likelihood that advected water carrying non-locally produced chlorophyll is accepted, it is clear that chlorophyll concentrations at depths far greater than either ZEEZ or ZAML have increased over the time of these surveys.

5.3. Chemical Observations

[26] Dissolved oxygen concentrations, determined from calibrated in situ oxygen measurements, are presented in Figure 9. Concentrations range from as low as 240 μmol kg−1 in the core of the warm MCDW in the center of the study region to over 400 μmol kg−1 in high-chlorophyll, near-surface areas. Saturation (with respect to the atmosphere) in surface waters along the transect ranges from less than 90% (100% saturation indicated by the dashed red contour) in waters overlying the warm, fresh water in the center of the study region to over 110% in high chlorophyll areas. The oxygen concentrations in the upper layers observed on 7 December are consistently greater than those observed on 2 December as a result of the progressing phytoplankton blooms. Following chlorophyll to some extent, the high-oxygen waters extend well below ZAML and ZEEZ. Furthermore, the chlorophyll and oxygen relationships are distinctly different from the east to west. Unlike chlorophyll, oxygen shows less evidence of subsurface maxima at the eastern end of the line, whereas it does track the deep chlorophyll “fingers” seen to the west. Above about 80 m depth, oxygen shows large horizontal variability and the scale of this variability is short, like temperature and chlorophyll, with several features having horizontal scale as short as a few kilometers.

Figure 9.

Oxygen concentrations on the same sections. The dashed red contours show the position of 100% saturation of dissolved O2 with respect to the atmosphere; ZEEZ (yellow dashed lines), density contours (white solid lines), and ZAML (black dashed lines) are as in Figure 8. Bathymetry (black solid line, bottom panel) is as in Figure 5. Density contour labels are omitted from this figure for clarity; refer to Figure 5.

[27] Distributions of CO2 partial pressure (PCO2, normalized to a constant temperature of −1.8°C to remove the effect of temperature using the temperature effect on PCO2 of (∂ln PCO2/∂T) = 0.0423°C−1 [Takahashi et al., 1993]) are shown in Figure 10. A progression of PCO2 drawdown over the 2-week period is clearly depicted in the three panels. Highest PCO2 levels are found in the core of the warm, fresh, MCDW-influenced water in the center of the study region; lowest levels are found in regions of high chlorophyll and high oxygen near the surface at the western end of the study area. Remarkable is the strong inverse similarity to the dissolved oxygen distributions: Where oxygen is high, PCO2 is low, even in the finest structure. This perceived correlation is borne out in plots of PCO2 versus oxygen, and will be discussed later (see Figure 15 in section 6.1.4). While we have the highest confidence in the integrity of our data obtained at the shipboard end of the sampling stream, this correlation between oxygen, measured with an in situ sensor, and PCO2, analyzed at the shipboard outlet of the sample stream, should convince even the most skeptical of the validity of our pumped sampling approach.

Figure 10.

CO2 partial pressure (PCO2) for the same sections, measured in the shipboard end of the sample stream and normalized to a constant temperature of −1.8°C. Red dashed lines show the position of 100% saturation of dissolved CO2 with respect to the atmosphere; euphotic zone depth (yellow dashed lines), density contours (white solid lines), and the mixed layer depth (black dashed line) are as in Figure 8. Bathymetry (black solid line, bottom panel) is as in Figure 5. Density contour labels are omitted from this figure for clarity; refer to Figure 5.

[28] Distributions of nitrate are shown in Figure 11. Concentrations are generally high, never falling below 18 μmol kg−1 in the course of the study and reaching values over 30 μmol kg−1. The concentrations in the upper layers did decrease significantly in the week between the first and second transects, indicating high rates of net primary productivity. Following PCO2 and mirroring oxygen, highest nitrate concentrations are found in the core of the warm, fresh (SSM) water in the center of the study region, and lowest nitrate concentrations are found in the regions of high chlorophyll toward the west. Not surprisingly, the apparent horizontal scale of several of the features in the nitrate distributions is on the order of a few kilometers, like chlorophyll, PCO2, and oxygen. PCO2 and nitrate are well correlated as will be discussed later (see Figure 16 in section 6.1.4).

Figure 11.

Nitrate concentrations for the same sections, measured in the shipboard end of the sample stream. Labeled gray contours illustrate the magnitude of silicate depletion relative to local deep water; ZEEZ (yellow dashed lines) and ZAML (black dashed lines) are as in Figure 10. Bathymetry (black solid line, bottom panel) is as in Figure 5.

[29] Phosphate distributions are shown in Figure 12. Like nitrate, concentrations are generally high, from no less than 1.2 μmol kg−1 to over 2 μmol kg−1 over the course of the experiment. Tracking nitrate and PCO2 and mirroring oxygen, phosphate concentrations are highest in the core of the warm, fresh water in the center of the study region and lowest in surface waters where chlorophyll concentrations are high. However, significant differences are observed. Unlike the distributions of these other chemical species, surface phosphate concentrations are more depleted in the eastern study area, where chlorophyll distributions show strong subsurface maxima than in the western area where chlorophyll concentrations are higher but show no subsurface maxima. While the pattern of phosphate variability is somewhat different than that of the chemical species discussed previously, the scale of the horizontal variability still appears quite small.

Figure 12.

Phosphate concentrations for the same sections, measured in the shipboard end of the sample stream. Silicate depletions (gray contours), ZEEZ (yellow dashed lines), and ZAML (black dashed lines) are as in Figure 11. Bathymetry (black solid line, bottom panel) is as in Figure 5.

[30] Some light can be shed on the apparent differences between the nitrate and phosphate distributions by the distributions of silicate depletion, shown by the thick gray contours in Figures 11 and 12. Where there is little, if any, silicate depletion, nitrate depletion is high and phosphate depletion is only moderate. Conversely, where silicate depletion is high, phosphate depletion is highest. Previous studies of the Ross Sea [El-Sayed et al., 1983; Wilson et al., 1986] show two distinct phytoplankton groups accounting for most of the productivity in the central polynya: diatoms, which form their silicate frustules, and Phaeocystis, which use only vanishingly small amounts of silicon, if any. Quantitatively, diatoms may be dominant east of the SSM, given the near equivalence of nitrate and silicate depletion typical of diatom growth in less stressful, iron-replete environments [Takeda, 1998; Fitzwater et al., 2000; Brezezinski et al., 2003]. This is consistent with the likely adequate supply of dissolved iron in these waters at this time of the year as shown by Coale et al. [2003]. On the other hand, diatoms must be nearly absent from the high-chlorophyll regions west of the SSM where silicate uptake is near zero. Plots of nitrate depletion (with respect to wintertime or deep-water concentrations) versus phosphate depletion (Figure 13), colored as a function of silicate depletion, show two distinctly different nitrate versus phosphate uptake trends: Where diatoms dominate, as evidenced by high silicate depletion, the N:P uptake ratio is significantly lower, about 9; where Phaeocystis dominate, the uptake ratio is about 19. These different uptake ratios have been observed by many researchers previously, even without the advantage of the high-resolution sampling employed here [Bates et al., 1998; Arrigo et al., 1999; Sweeney et al., 2000a, 2000b; Arrigo et al., 2002, 2000; Smith and Asper, 2001]. These observations, therefore, indicate that the classical Redfield ratio of 16 [Redfield et al., 1963] represents the nutrient utilization ratio for a mixture of these organisms.

Figure 13.

Nitrate depletion as a function of phosphate depletion; symbols colored as function of silicate depletion. Areas of high diatom productivity, indicated by high silicate depletion, have a distinctly different depletion ratio than areas of high Phaeocystis productivity. For reference, the slopes resulting from several different reported uptake ratios are shown: nine for diatom productivity and 19 for Phaeocystis productivity as given by Arrigo et al. [1999], Bates et al. [1998], and Sweeney et al. [2000b], and 16 for canonical Redfield stoichiometry.

6. Discussion

[31] We will break our discussion into two broad categories: The first will address some of the questions that the observations prompt regarding the observed pattern and character of productivity, and the second will briefly address the importance of mesoscale-resolution measurement capability in the Ross Sea.

6.1. Productivity Questions

6.1.1. Low Productivity in SSM

[32] One of the most puzzling features seen in these distributions is that of low chlorophyll and oxygen and high PCO2, nitrate, and phosphate in the surface waters of the SSM. The physical condition in this region ought to be the most favorable for productivity along the 76.5°S study line, because of the fact that here apparent mixed layers are the shallowest and most stable, and light penetration is good, with ZEEZ deeper than ZAML on all three transects. The concentrations of macronutrients are high in these surface waters: In the SSM, nitrate exceeds 25 μmol kg−1, phosphate exceeds 1.4 μmol kg−1, and silicate is greater than 70 μmol kg−1. However, indicators of productivity within the SSM are consistently the lowest seen in surface waters. Chlorophyll concentrations are several-fold lower than they are to either the east or the west. Oxygen concentrations are low, below saturation with respect to the atmosphere, and PCO2 levels are the highest seen in surface waters anywhere else along the transect. Trace nutrients (e.g., iron or zinc) can become limiting late in austral summer [Coale et al., 2003; Fitzwater et al., 2000; Sedwick and DiTullio, 1997]. No measurements of these were made during this cruise; however, the low-salinity water at the surface can only come from ice-melt, and this is generally thought to be a source of trace metals in the Southern Ocean [Edwards and Sedwick, 2001; Sedwick and DiTullio, 1997; Smith and Nelson, 1985; de Baar et al., 1995]. Grazing is unlikely to be a major impediment to productivity here at this time, since zooplankton abundances in the Ross Sea early in the productivity season are usually low [Caron et al., 2000; Lonsdale et al., 2000], and measurements of reduced nitrogen species (i.e., nitrite and ammonia[Gordon et al., 2000]) do not show the high concentrations indicative of grazing activity during this time frame.

[33] We believe this low productivity is the result of a short exposure of this water to an ice-free, sunlit surface, and that the plankton there simply have not had enough time to take advantage of the growth-favorable conditions. This is implied by the low oxygen concentrations in this surface layer. Mixed layers are shallow in the SSM, only about 20 m, and even moderate gas exchange should bring a layer like this into equilibrium with the atmosphere in a matter of several days. We can roughly quantify the time over which this water has been exposed to the atmosphere: Assuming that it started with an O2 concentration of 240 μmol kg−1 (a typical value for the MCDW), and that productivity has added approximately 50 μmol kg−1 (given the observed nitrate depletion) of oxygen, then gas exchange has added approximately 30 μmol kg−1 to this water since it reached the surface. As shown below, even for a conservative estimated gas exchange velocity of 1 m d−1, this implies that the water could have only been at the surface for about 7 days.

[34] This leads us to ask whether the SSM water and MCDW below arrived to our study section separately or coupled together. Freshwater input to the Ross Sea can come only from ice-melt. Although the stratification associated with this fresh surface water is the highest observed in the entire study, it is not high enough to allow significant vertical shear. The stratification maximum lying underneath the low-salinity surface water has N2 ≈ 10−4 over a depth range of less than 10 m. In order to maintain a Richardson number of greater than or equal to 1, the threshold at which mixing would break down stratification, the vertical velocity change over this depth interval must be less than 2 cm s−1. Such vertical velocity gradients are not supported, at least in the north-south direction, by the geostrophic calculations, which show very little vertical variability in the velocity fields. This implies that the SSM water above the MCDW is intimately coupled to the MCDW intrusion itself and that the meltwater-driven freshening of these surface waters must have occurred prior to the MCDW's appearance at this location. Had the SSM and MCDW arrived at our section from different directions separately, then much greater velocity shear would be required.

[35] We can think of two possible scenarios that bring the SSM/MCDW intrusion to the 76.5°S study line: (1) It could be part of a current meander, flowing from the southeast in a northwesterly direction, crossing 76.5°S at about 175°E and then turning back to the southwest and crossing 76.5°S again at about 173°E; or (2) it could be an anticyclonic eddy crossing the line from either northeast to southwest or from southwest to northeast directions. We believe that the second scenario is most likely, specifically with the eddy moving from southwest to northeast, given the location of meltwater sources and distributions of chlorophyll as recorded by satellite images. There are essentially two sources of meltwater in the Ross Sea: sea ice, which is predominant seaward to the north and east, and the Ross Ice Shelf to the south. In the first scenario, this current meander would have to glean its freshwater signal from sea ice to the east-southeast of the study line. This is unlikely, however, as satellite images of chlorophyll distributions show high chlorophyll in surface waters in this direction, and the SSM is a low-productivity feature (B. G. Mitchell, Scripps Institution of Oceanography, personal communication, 2000). In the second scenario, an eddy could pick up its fresh water from sea ice to the northeast or southeast; however, satellite-derived chlorophyll in these directions is high, inconsistent with the low-productivity SSM. The eddy could, alternatively, pick up its freshwater from melting RIS ice, and move northward across the study line where observed. This scenario is consistent with that of Jacobs et al. [1985], who suggested a southward flow of MCDW to the RIS in the east, followed by westward flow along and under the ice shelf [Pillsbury and Jacobs, 1985], and finally a northward return through the Ross Sea. It is consistent with satellite-observed distributions of surface chlorophyll, which are lowest immediately north of the RIS, and south of the SSM/MCDW's appearance on the 76.5°S line. Finally, it is consistent with a divergence driven by the strong wind event preceding our surveys which could cause upwelling of any MCDW beneath the ice shelf and move it offshore toward the shelf-break. At the longitudes of the expression of the SSM/MCDW feature, the RIS is about 50 km to the south of the 76.5°S study line. Given this distance and our earlier discussion of the time that the SSM could have been at the surface (less than 7 days), the eddy must be moving at an average speed of about 8 cm s−1 or 7 km d−1. This velocity is of the same magnitude as the calculated geostrophic velocities at the edges of the feature (up to 20 cm s−1). An independent measure of the eddy's velocity can be obtained by assuming that it is a symmetrical feature and that its expression on 2 December represents its maximum cross-sectional length, about 70 km. On the preceding and following surveys, the width of the coherent feature is less, about 50 km. If the eddy is in fact moving from southwest to northeast, then it must have moved about 25 km in those time intervals, implying a velocity of 4–5 km d−1. Movement in this direction is consistent with the bathymetry of the Ross Sea, which shows a trough aligned to the north-northwest between the RIS and the shelf-break. Perhaps this feature helps guide flow through the Ross Sea.

6.1.2. Why Do Diatoms Dominate East of the SSM?

[36] The second interesting feature of these distributions is the clear delineation between the diatom-dominated region east of the SSM and the Phaeocystis-dominated region to the west. It is clear from the absence of silicate uptake in the west on 25 November and 2 December (Figures 11 and 12) that diatoms were not growing there. To the east, the near equivalence of silicate uptake and nitrate depletion, in the 1:1 ratio typical of diatoms growing in nutrient-replete waters [Brezezinski, 1985; Dugdale and Wilkerson, 1998; Takeda, 1998], seems to leave little room for nitrate-consuming Phaeocystis. In support of this segregation of the two communities are the distributions of phaeopigments, which indicate distinct Phaeocystis populations in the west and diatom colonies in the east [Sweeney et al., 2000a, 2000b; Bidigare et al., 1996].

[37] Macronutrients are abundant throughout the study region, and while micronutrients have been shown to reach low levels later in the season here [Coale et al., 2003; Fitzwater et al., 2000], they are not likely be a critical factor in determining community structure this early in the season. The only distinction we can see in our data between the two sites is in the physical characteristics of the surface waters. Apparent mixed layers are slightly shallower and more stratified in the east than they are in the west (Table 1), and euphotic depths more consistently exceed the mixed layer depths in the east. It has been reported that diatoms prefer shallower, more stable mixed layers than Phaeocystis [e.g., Arrigo et al., 2003, 1999]. We are surprised, however, at how slight are the differences in ZAML between two regions so distinctly different in phytoplankton community structure. The mixed-layer depths in the west are 49 ± 28, 34 ± 17, and 26 ± 7 m in the 25 November, 2 December, and 7 December transects, respectively, compared with 50 ± 20, 26 ± 8, and 20 ± 5 m in the east during the same time periods (see Table 1). The differences between the east and west are well within 1 standard deviation for each period. The same could be said of the differences in the strength of the stratification at the base of the mixed layers in the two regions: Stratification is stronger in the eastern region, but one would be hard-pressed to claim a statistically significant difference between the two regions (Table 1). Do the plankton communities respond as if this is a significant difference, as they are clearly segregated between the two settings? Perhaps the greater spatial variability in ZAML to the west (as indicated by the rapid up-and-down variability of ZAMl in the west as shown in Figures 6 and 8, and the larger standard deviation of ZAMl there as shown in Table 1) implies a more dynamic environment in the west and a more stable, quiescent environment in the east. Diatoms are perhaps unable to optimize their growth rates in the more variable west, while Phaeocystis are less sensitive to these dynamic conditions.

[38] While we can suggest a reason for diatoms to fare better in the slightly shallower and more stratified surface mixed layers to the east, and worse in the slightly deeper, more variable, and weakly stratified, mixed layers to the west, we are without an adequate explanation for the absence of Phaeocystis in the diatom-dominated region. Since macro-nutrients are replete in the surface waters, there should be no meaningful competition for these resources. Arrigo et al. [2003] contend that Phaeocystis have comparable iron demand. Micro-nutrients, however, are probably reasonably abundant in these early-season conditions as well [Coale et al., 2003], and if there is cross-polynya variability, it is expected that the ice-melt-driven lower-salinity surface waters of the east and SSM should have the highest iron content. Since the eastern water (LSSW) may have come from different location and environments from the western water (HSSW), it is possible that each of them might have carried different amounts of plankton seed stocks. However, qualitative microscopic observations show that while diatoms do indeed dominate Phaeocystis in the east, there are Phaeocystis cells there in sufficient quantity to constitute a seed population (S. Mathot, personal communication, 1998).

[39] Arrigo et al. [2003] suggest that Phaeocystis are more sensitive to photoinhibition than are diatoms, and are thus discriminated against in well-lit, clear, shallow mixed layers and are prone to develop in deeper subsurface layers when such conditions prevail. There are a number of problems using these concepts to describe our observations. First, it requires that phaeocystis are sensitive enough to photoinhibition to be overwhelmingly segregated between the two environments by the statistically insignificant differences between those environments' mixed-layer and euphotic zone depths. Second, the chlorophyll distributions in the Phaeocystis-dominated west show no subsurface maxima, as seen in the diatom-dominated east and SSM; they are highest at the shallowest depths sampled by the LPS. Third, there is no difference seen in the absolute surface irradiance in the Phaeocystis and diatom-dominated areas. Finally, while we do see lower light levels and shallower ZEEZ in the high-chlorophyll regions in the west, Phaeocystis cannot “self-shade” at the beginning of a bloom. The absence of significant silicate depletion in these regions rules out the existence of shallow-diatom strata above the Phaeocystis. A scenario where a surface diatom layer formed initially, providing the Phaeocystis with shade, and then was wholly recycled without leaving a chemical uptake signature following the development of a self-shading Phaeocystis bloom seems unlikely. Thus this explanation is not supported by the observation. We can only pose the question of why Phaeocystis do not prosper in the seemingly favorable conditions in the east.

6.1.3. Chlorophyll Maxima

[40] A third observation that is worthy of discussion is that of the subsurface maxima in chlorophyll in the diatom-dominated east observed in the latter two transects. These subsurface maxima are located in the vicinity of the ZEEZ but below the base of the mixed layer. Despite the clear maxima seen in chlorophyll, we see no corresponding maxima in oxygen concentrations, nor minima in PCO2 and the concentrations of nitrate, phosphate, and silicate. Chlorophyll concentrations in the layer appear to be unchanged from 2 to 7 December. This indicates that chlorophyll is not actively utilizing CO2 and nutrients in the chlorophyll maximum layer, and its concentration is approximately in a steady state. Assuming that the chlorophyll is entirely associated with diatoms, we interpret that diatoms photosynthesize and grow mostly in depths above the chlorophyll maximum layer, and that the grown diatoms sink to the chlorophyll maximum depth and accumulate at this depth. They become inactive with no net utilization of CO2 and nutrients due to low light levels. At this depth, the rate at which diatoms are accumulating is nearly balanced with the rate of loss out of the layer by the gravitational settling of individual and/or aggregated particles.

[41] Parslow et al. [2001] explained the formation of a persistent subsurface chlorophyll maximum, composed primarily of large diatoms, that was observed at water depths 60 m to 100 m near the 1% light level in the high-nutrient- low-chlorophyll Polar Front Zone (53°S–58°S) along 145°E, south of Tasmania. They considered that a chlorophyll maximum can occur when the supply rate from the waters above exceeds the rate of loss by sinking of diatoms to waters below. Of a number of physical, biological, and chemical factors, which they evaluated, the supply rates mediated by iron and/or silicic acid availability were considered to play an important role in the formation of the chlorophyll maximum. Thus, in the east, the settling flux of diatoms from the photic layer above may be faster than the loss rate, whereas in the west, Phaeocystis sink slowly first and then sink rapidly as they form large aggregates. This may be a factor in the absence of a deep chlorophyll maximum in the west.

[42] A second possible interpretation is that the chlorophyll maxima are maxima in the chlorophyll:carbon ratio in the plankton cells. We are unaware of any publications of these ratios; however, POC distributions generally show maximum values at the surface in early-bloom Ross Sea conditions [Gardner et al., 2000; Carlson et al., 2000]. Phytoplankton may adjust their chlorophyll content to reflect varying light intensities as a function of depth, and therefore, have more chlorophyll at depth where light intensity is lowest. This interpretation has the advantage of decoupling the productivity from carbon and nutrient uptake rates without requiring sinking of senescent diatom cells to a depth coincident with ZEEZ but not ZAML. This scenario requires that diatoms be better able to regulate their biochemical composition than are Phaeocystis; we are unaware of any evidence in support of this.

6.1.4. Net Utilization Ratios of Carbon, Oxygen, Nitrogen, and Phosphorus

[43] Our final productivity-related discussion point is that of the differences in the ratios of nutrient, carbon, and oxygen uptake between the diatom- and Phaeocystis-dominated regions. Figure 13 shows the uptake of nitrate versus phosphate uptake during all three 76.5°S surveys, colored to indicate the silicate uptake. There is a clear bifurcation of this relationship, with diatoms consuming about twice as much phosphate per mol of nitrate as Phaeocystis. This is similar to results seen by other researchers [Takeda, 1998; Arrigo et al., 1999; Gordon et al., 2000; Sweeney et al., 2000b]; however, we can definitively say that the bifurcation is only present due to an uptake of excess phosphate by diatoms. All other biologically-active chemical species, i.e., nitrate, carbon, and oxygen, correlate tightly to each other, and show no such split between diatom- and Phaeocystis -dominated regions. As stated in section 4, our measurements of TCO2 were lacking in terms of absolute accuracy, but Figure 14 demonstrates the tight correlation of TCO2 to PCO2 and constrains the value of the Revelle factor (γ = ∂ ln PCO2/∂ ln TCO2) to about 15. This relationship can be used to constrain changes in TCO2 from our higher quality PCO2 data. Figure 15 shows the relationship of PCO2 to oxygen; as expected from the mirror-image relationships between the PCO2 and O2 distributions in Figures 9 and 10, they are tightly negatively correlated with a single relationship. The slope of these correlations and the Revelle factor determined above indicates a C/-O2 relationship of 1.3–1.4, close to the canonical Redfield ratio of 1.3. The relationships between nitrate and PCO2 (Figure 16), like the relationships between PCO2 and O2, are tight linear correlations all grouped around a single trend on each transect. The slopes of these trends, coupled with the Revelle factor determined above, yield a single C/N net utilization ratio of 7 ± 1 for all three transects. This is consistent with the canonical Redfield ratio of 6.6, and falls between the C/N ratios of 5.2 ± 0.4 for gross diatom production and 9.1 ± 1.2 for net diatom production in the Southern Ocean reported by Brezezinski et al. [2003]. Thus carbon, nitrate, and oxygen are consumed/produced in near-constant ratios consistent with Redfield stoichiometry, while phosphate, on the other hand, is always taken up about twice as much by diatoms per mol of carbon or nitrate consumed, or oxygen produced, than it is by Phaeocystis. Sweeney et al. [2000b] show that the apparent excess phosphate uptake has nearly disappeared later in the season. However, it is not known whether this is due to changes in specific uptake ratios (for example, diatoms may have higher phosphorus demands early in the season, or may be able to take it up and store it for later use in more Redfield-like stoichiometry) or, alternatively, due to a breakdown of segregation between the two communities, or possibly, to greater relative contributions of vertical mixing of deeper waters with background N:P ratios.

Figure 14.

PCO2 as a function of TCO2 for the 7 December survey. While our high-speed TCO2 data were not of desired accuracy, the obvious correlation between PCO2 and TCO2 and the implied Revelle factor of 15 (red line) strongly support the use of PCO2 as a proxy for TCO2 in this environment.

Figure 15.

PCO2 as a function of O2 concentration. The red line is a regression of the data (over a somewhat restricted range to eliminate the complication of the slightly different PCO2:O2 relationship seen in the deep MCDW); the ΔO2:ΔC ratios shown result from combination of these slopes and the Revelle factor determined in Figure 14.

Figure 16.

Nitrate concentrations as a function of PCO2. The red line is a regression of the data (over a somewhat restricted range to eliminate the complication of the slightly different properties of deep MCDW); the ΔC:ΔN ratios shown result from combination of these slopes and the Revelle factor determined in Figure 14.

6.2. Mesoscale Variability of the Upper Layers of the Ross Sea

6.2.1. Variability of Physical and Biogeochemical Parameters

[44] The LPS transect data (Figures 5, 6, and 812) demonstrate clearly that all physical and biogeochemical parameters vary with horizontal length scales greater than several kilometers. Also, the data show that the variability in all the biogeochemical parameters extend vertically as deep as 120 m, which is considerably below the 1% light level located at depths of 30 m to 50 m (indicated by the dashed yellow curves in these figures). A unique feature is observed in the 7 December transect (bottom panels) in the area between 173°E and 174°E (west of the SSM), where the isopycnals dome upward. Here the concentrations of chlorophyll and oxygen in upper layers are the highest and the PCO2, nitrate, and phosphate are drawn down significantly. The chlorophyll and oxygen data show not only a maximum centered around 80 m deep off the western edge of the isopycnal dome, but also a minimum in the same depth range off the eastern edge of the isopycnal dome. The PCO2 and nitrate data show a distribution pattern reverse to chlorophyll and oxygen.

[45] In an effort to explain the patchy distribution of phytoplankton blooms in the northern edge of the North Atlantic subtropical gyre, Pollard and Regier [1990, 1992] and Woods and his associates [Woods, 1988; Onken, 1992; Fiekas et al., 1994; Strass, 1994] have developed various dynamic models for the formation of mesoscale meandering and eddies and their effects on phytoplankton blooms. More recently, the mathematical framework formulating the effects of ocean's mesoscale and submesoscale hydrodynamic behaviors on biological activities have been discussed extensively by Flierl and McGillicuddy [2002]. Their results show that vertical circulation cells with upwelling and downwelling occur along a meandering front. The vertical motion is driven by geostrophic advection of relative vorticity, and is as fast as 10 m d−1 reaching depths exceeding 100 m. It appears that our observations may be taken as a demonstration of the circulations which have been theorized and modeled by the researchers listed above.

6.2.2. Mesoscale Variability and its Significance to Sampling Intervals

[46] It is clear from simple comparison of our mesoscale-resolved sections (Figures 5, 6, and 812) with low-resolution sections, that would result from 50-km hydrostation-spacing of those same fields (Figure 17), that there is strong variability in most fields that cannot be resolved by widely spaced hydrostations. We have also seen that the LPS surveys found essentially the same results as many earlier coarse-resolution hydrostation-based studies when considering only property-property relationships such as T-S or nutrient:carbon uptake ratios. Hence the importance of short length scale variability must be assessed in an objective manner. Quantifying the importance of high-resolution sampling in characterizing the distributions of the parameters measured is a complicated task, and we are uncomfortable with many of the approaches typically employed to do this. Our data are not well suited for analyses with Fourier transform techniques (e.g., spectral power distributions [Sweeney, 2000] or autocorrelations) that assume periodic data records that are much longer with respect to their length scales of variability than ours. Even direct calculations of decorrelation length scales are lacking in quantitative representation of the significance of the variability; for example, an autocorrelation of the horizontal distribution of temperature at 20 m on 2 December shows a very short (<20 km) decorrelation distance (Figure 18). We know, however, that total variability in the temperature field is less than 1°C, and feel that discussing only length scales of variability without quantifying the significance of that variability is misleading. We chose, rather, to quantify the absolute errors in prediction of unsampled data as a function of sampling resolution.

Figure 17.

Examples of temperature and PCO2 distributions for the 2 December survey as they would appear if sampled at hydro-station resolution and gridded as the high-resolution results shown in Figures 5 and 10. Solid vertical lines in top panel indicate locations of “stations” extracted from the gridded data in Figure 5; vertical resolution of temperature data at these stations is 1 m. Squares in bottom panel indicate the hypothetical location of “samples” at the stations; vertical resolution of these samples is 10 m.

Figure 18.

Example of an autocorrelation analysis performed for temperature as a function of longitude at 20 m depth during the 2 December survey. Decorrelation at a phase shift of 19 km confirms the obviously short length scales of variability evident in Figure 5, but does not quantify the significance of this variability.

[47] Figure 19 summarizes the errors due to undersampling as a function of sampling resolution for each of the three LPS transects along 76.5°S during the AESOPS Process IV cruise. We have included analyses of two in-situ measured physical variables (T and S), one in situ measured bio-optical variable (chlorophyll, via fluorescence), one in situ measured chemical variable (O2), and one shipboard-measured chemical variable (PCO2). All of these have been analyzed at three depths: 20 m (mostly within the mixed layer), 50 m (both in and out of the mixed layer), and 100 m (mostly out of the mixed layer). From each of these traces, two pieces of key information can be extracted. The first is the asymptotic error, approached as sample spacing approaches the total length of the transect; this is taken directly from the graphs and tabulated in Table 2. The second is the characteristic error length scale, a parameter that quantifies the length scale of the variability. This is more subjective; however, we simply took advantage of the shape of the error function curves, which are reminiscent of the familiar first-order exponential rise function, and found the sampling resolution corresponding to the point on the error curves which was 63% of the asymptotic error shown in Table 2. These values, which correspond very well in magnitude to more traditional decorrelation length scales, are tabulated in Table 3, and should be thought of as the sample spacing below which most of the variability occurs.

Figure 19.

Errors in predicting unsampled data as a function of horizontal sampling resolution for all three transects. Examples are shown for two in situ measured physical parameters (temperature and salinity), one in situ measured bio-optical parameter (fluorescence), one in situ measured chemical parameter (oxygen), and one shipboard-measured chemical parameter (PCO2). Three depths were chosen for this illustration: 20 m (red lines), which is mostly in the mixed layer; 50 m (green lines), which is intermittently in and out of the mixed layer, and 100 m (purple lines), which is below the mixed layer.

Table 2. Absolute Horizontal Variability at Specified Depth Horizons
ParameterDepth, m25 November2 December7 December
Temperature,°C200.050.080.20
500.050.080.13
1000.100.180.22
Salinity200.050.070.06
500.040.070.04
1000.030.040.06
Fluorescence, mg m−3200.440.840.64
500.340.690.74
1000.130.120.12
Oxygen, μmol kg−120 1616
50 2120
100 119
PCO2, μatm20183130
50163535
100121411
Table 3. Characteristic Variability Length Scalesa
ParameterDepth, m25 November2 December7 December
  • a

    In kilometers. Characteristic variability length scale is determined by assuming that the error functions shown in Figure 19 are functionally described by an equation of the form equation image, where E is the error due to undersampling, E is the error asymptotically approached as the sample spacing, x, approaches infinity (E values are given in Table 3), and xchar is the characteristic length scale tabulated above.

Temperature20231915
5013910
100112124
Salinity20524724
50364457
100214147
Fluorescence20271711
5013810
100875
Oxygen20 2315
50 1413
100 1516
PCO220391920
50122113
10081310

[48] There are several conclusions that can be drawn from this analysis that illuminate the importance of high spatial resolution sampling. First, all data except salinity show most of their variability at spatial resolution of <25 km, as evidenced by the steep rise of the error functions toward their asymptotic values in the 0- to 25-km longitude resolution range and the characteristic variability length scales (Table 3) of 10–20 km. Second, at the 2° longitude AESOPS station spacing (∼50 km), undersampling errors for all data except salinity are at or near their asymptotic value; in other words, the unsampled data are represented about as well by stations at 200 km spacing as they are by the stations at 50 km spacing, and capturing the spatial variability present in the data would require at least 2–3 times more discrete hydrostations than were occupied during AESOPS. Third, chemical and bio-optical data show undersampling errors that increase with coarser resolution as fast as or faster than temperature, but with different patterns at different depths and times. Also, finally, shipboard-sampled chemical data (PCO2) shows similar spatial variability as seen in in situ measured O2 data. This is probably not surprising given the tight correlation between PCO2 and O2 shown earlier (Figure 15), but it does demonstrate that the pumped sampling and shipboard analysis approach employed with the LPS captures even the shortest lateral variability seen with in situ measured parameters.

[49] The importance of this short length scale variability, however, is dependent on the parameter in question, and the process to which that parameter's variability is important. For example, the absolute magnitude of the total variability in the temperature and salinity data is very small. Particularly for temperature, the asymptotic undersampling uncertainty (∼0.1°C) is so small that it makes almost no difference to any other property (e.g., seawater density or gas solubility) or process (e.g., microbial respiration or gas exchange rates) of interest. In contrast, the absolute magnitudes of the chlorophyll, PCO2, and O2 variability are quite important. At depths of 20 and 50 m, asymptotic chlorophyll undersampling uncertainty is ∼0.6 μg L−1; this would constitute a bloom in most areas of the ocean. Asymptotic PCO2 uncertainty is about 30 μatm; this would represent a significant error in estimation of gas exchange flux. Asymptotic O2 uncertainty is 15–20 μmol kg−1; this would represent a significant error in gas exchange flux or net community O2 production/consumption (as estimated, for example, by Bender et al. [2000]). Sampling chemical and biological parameters at high spatial resolution in this setting thus appears to be even more important than sampling physical parameters at similarly high resolution.

[50] Finally, if longitudinal averages and standard deviations are the desired end result, much coarser sampling can be accepted. As an example, the uncertainty of the mean PCO2 as a function of undersampling is shown in Figure 20. In this case, the error in predicting the mean PCO2 from an undersampled data set is only about 5 μatm at all depths when sampled at hydrostation resolution. Standard deviations are similarly well quantified at hydrostation resolution. Therefore the question of whether higher resolution sampling is necessary can be assessed with an initial coarse-resolution survey that quantifies averages and standard deviations of parameters of interest. If the standard deviation of the coarse resolution data is large compared to required data precision and accuracy, then higher spatial resolutions will be required.

Figure 20.

Example of errors in prediction of longitudinally-averaged PCO2 as a function of sampling resolution, showing the much lower uncertainty in prediction of averages from undersampled, highly spatially variable distributions. Resampling is as for PCO2 in Figure 19.

7. Conclusions

[51] In the preceding sections, we showed results from the first deployments of the Lamont Pumping SeaSoar, performed over a 2-week interval along the U.S. JGOFS AESOPS study line at 76.5°S in the Ross Sea polynya, Antarctica. High-spatial resolution measurements of physical, bio-optical, and chemical parameters were obtained during the on-set and early development stages of phytoplankton blooms. Our data show the presence of water masses known to make up the Ross Sea's volume, the high productivity expected for this season, and the distinct diatom-dominated phytoplankton area separated from the Phaeocystis -dominated area by an intrusion composed of a low-salinity surface water and MCDW. Horizontal variability of meso- and submeso-scales ranging from several to 30 km are observed, especially in biologically mediated properties (i.e., PCO2, and the concentrations of oxygen, TCO2, and macro-nutrients) within the upper 140 m. The observed scale of variability is consistent with the Rossby radius of deformation, which is typical of weakly stratified layers in the high-latitude Southern Ocean. The signals for photosynthesis are detected at depths considerably below the euphotic depths (<1% light levels). This may be caused by the up- and down-welling circulation associated with an eddy or a meandering filament. While property-property relationships and polynya-wide averages of properties are similar whether observed using high-resolution sampling schemes as deployed here or more traditional coarse-resolution hydrostation sampling approaches, actual distributions of biogeochemically important parameters are significantly in error when sampled at coarse resolution. The decision to sample at high spatial resolution is dependent on the process of interest, and its dependence on the magnitude of variability in the parameters upon which that process depends.

Acknowledgments

[52] We thank David Chipman for the construction and shipboard operations during this investigation of high-speed PCO2 and TCO2 measurement systems. Engineering assistance by Erich Scholz and technical assistance by John Goddard and Stewart Sutherland made the successful shipboard operations possible. We acknowledge helpful collaborations with Walker Smith, Jr., Chief Scientist, during the AESOPS expedition. Discussions with Arnold Gordon, Stan Jacobs, and Colm Sweeney on the Ross Sea oceanography are appreciated. We gratefully acknowledge the following financial support: the support for the field operations by the National Science Foundation; Global Change Post-doctoral fellowship to B. H. from the Department of Energy; and a grant to T. T. from National Oceanographic and Atmospheric Administration for the construction of the Lamont Pumping SeaSoar system. This is LDEO Contribution 6580.

Ancillary