Assessment of Oceanic Productivity with the Triple-Isotope Composition of Dissolved Oxygen

See allHide authors and affiliations

Science  16 Jun 2000:
Vol. 288, Issue 5473, pp. 2028-2031
DOI: 10.1126/science.288.5473.2028


Plant production in the sea is a primary mechanism of global oxygen formation and carbon fixation. For this reason, and also because the ocean is a major sink for fossil fuel carbon dioxide, much attention has been given to estimating marine primary production. Here, we describe an approach for estimating production of photosynthetic oxygen, based on the isotopic composition of dissolved oxygen of seawater. This method allows the estimation of integrated oceanic productivity on a time scale of weeks.

Our knowledge of the rate of marine photosynthetic production is based primarily on bottle incubation experiments (1). These experiments provide local instantaneous primary production rates, which often miss the effects of significant blooms because of the heterogeneous distribution of plankton in time and space. A broader view of marine primary production can be obtained from satellite remote sensing (2). However, values derived by this method depend on the quality of calibration data obtained by actual productivity measurements in the ocean and cannot be better than the accuracy of this information. Here, we present a way to estimate marine production that alleviates the inherent problems of incubation methods. In this approach, gross production, integrated on spatial and temporal scales, is estimated from the difference between the triple isotope (16O,17O, and 18O) composition of atmospheric and dissolved O2 and the rate of air-sea O2exchange.

Most terrestrial processes fractionate O isotopes in a mass-dependent way, such that 17O enrichment is about half of 18O enrichment relative to 16O. As a result, δ17O and δ18O in terrestrial materials plot along a line with a mass-dependent slope of about 0.52. This δ17O/δ18O slope represents an average of slightly different slopes of various mass-dependent processes (3). For example, the δ17O and δ18O of O2 produced by photosynthesis and fractionated by respiration vary along a line with a slope of 0.521 (4), but fractionation of O isotopes in the formation of meteoric precipitation occurs with a slightly steeper slope, 0.534 (5). In contrast to these mass-dependent processes, ultraviolet (UV)-induced interactions among O2, O3, and CO2 in the stratosphere cause mass-independent fractionation (6) with equal lowering of δ17O and δ18O in atmospheric O2 (4). Therefore, for a given δ18O of O2 produced solely by biological production and consumption, there is an excess of 17O in comparison to air O2. This 17O excess (Δ17O) with respect to air O2 is defined asEmbedded Image(1)By definition, δ18O, δ17O, and Δ17O of air O2 equal zero (7).

The Δ17O value of dissolved O2diss) depends on the rate of air-water gas exchange, which tends to bring Δdiss to an equilibrium value with air, and the rate of in situ production of biological O2, which tends to increase Δdiss to a maximum value of pure biological O2max). In natural aquatic systems, Δdiss varies between these two extremes, and its value depends on the ratio of the rates of gross primary production and air-sea O2 exchange. Thus, gross production can be calculated from Δdiss if the rate of air-sea gas exchange is known.

Isotopic analysis of water seems to be a straightforward way to determine Δmax variations. However, the error of individual Δ17O measurements of H2O is relatively large [±75 per meg (1000 per meg = 1‰) (5)], and the value of Δ17O of H2O is very sensitive to the slope used in Eq. 1 for its calculation (8). For these reasons, we have determined Δmax in closed system experiments in which O2was produced and consumed in the absence of UV radiation.

In a previous study, Δmax was determined from the Δ17O of O2 produced byPhilodendron (a higher plant) in terrariums containing water from the Sea of Galilee and from the Dan River (4). This river is a major water source of the Sea of Galilee, but the latter is enriched in 18O by about 5‰ due to evaporative loss of water vapor. The Δmax values were 155 ± 15 and 156 ± 7 per meg for the Sea of Galilee and the Dan River, respectively. This shows that evaporation from the lake surface does not affect Δmax, because it causes δ17O and δ18O to increase along a slope of 0.521 (Fig. 1).

Figure 1

Schematic plot (not to scale) of δ17O, δ18O, and Δmax variations among seawater, meteoric water, air O2, and biological O2. Graphically, Δ17O is the vertical distance from the line with 0.521 slope going through the point representing the HLA air standard. In pure biological O2, Δ17O = Δmax. Point A represents ocean water and point D represents biological O2max = 249 per meg) produced from point A and fractionated by respiration along a line with δ17O/δ18O slope of 0.521. Point B represents meteoric water fractionated along the line with δ17O/δ18O slope of 0.534. Subsequent evaporation fractionated this water along a line with δ17O/δ18O slope of 0.521 to form the water of the Sea of Galilee (point C). Point E represents biological O2max = 159 per meg) produced from C and fractionated by respiration.

As discussed below, accurate Δmax values are critical for estimating true aquatic productivity. Therefore, we conducted additional experiments with marine and freshwater organisms (9). The Δmax of O2 produced from fresh water by Peridinum (a major producer in the Sea of Galilee) was 159 ± 10 per meg, very close to the value determined for the same water by a completely different plant,Philodendron (10). The Δmax values of O2 produced from seawater by marine organisms—planktonic algae (Nannochloropsis) and corals (Acropora) with their symbiotic algae—were 244 ± 20 and 252 ± 5 per meg, respectively. The different Δmax values of seawater and water from the Sea of Galilee are expected because meteoric water has different δ17O/δ18O slope than that of biological uptake (Fig. 1). It should be emphasized that our experiments clearly show that the values of Δmax are independent of the type of organisms producing and consuming O2. Finally, isotopic variations of seawater from different parts of the ocean are small (11), and thus an average Δmax value of 249 ± 15 per meg should be representative of the entire ocean.

In order to derive quantitative estimates of gross production, we consider a simple model in which the aquatic mixed layer, in contact with air, is in a steady state with respect to O2concentration and Δdiss and vertical mixing with deeper water is neglected. In the absence of biological activity, Δdiss is expected to be close to Δ17O of air O2, and thus approximately zero. However, in the case of air-water equilibration, we determined equilibrium Δdisseq) of 16 per meg (12). This small deviation of Δeq from zero suggests that the δ17O/δ18O slopes in invasion and evasion of O2 are slightly different than 0.521. We assume that the positive Δeq is the result of fractionation during O2 invasion alone (12). Thus, the Δ17O balance in the mixed layer is given byEmbedded Image(2)where I and E are the rates of atmospheric O2 invasion and evasion, respectively,GP is O2 gross production, and R is total oxygen consumption. The difference between evasion and invasion fluxes is the net biological O2 flux, and in our simplified case it equals PR. Noting thatI = KC o [where K is the coefficient of gas exchange (piston velocity) andC o is equilibrium O2concentration], Eq. 2 can be rewritten asEmbedded Image(3)We have applied the new method in the highly productive Sea of Galilee and in the BATS (Bermuda Atlantic Time-series Study) Station located in the low-productivity region of the Atlantic Ocean. The Sea of Galilee is a simple case where the entire photic zone, and thus all gross production, is confined to the mixed layer of about 10 m depth. The spatial distribution of O2 and Δdiss in the mixed layer is uniform and the residence time of O2 is about 5 days. The results for the Sea of Galilee (Table 1) show the expected seasonal cycle with higher productivity in spring and summer, and are in excellent agreement with the estimates obtained from bottle incubations with H2 18O spike (1). We notice that the uncertainty in GP17O) sharply increases when Δdiss values in the lake approach the biological limit of Δmax. With the exception of this extreme situation, errors in GP due to inaccuracies in Δmax, Δeq, and Δdiss are on the order of about 30%, and are smaller than errors associated with uncertainties in piston velocities (13). Therefore, improvement in estimation of GP is likely to come mainly from better knowledge of gas exchange rates, which can be gained by deliberate tracer experiments (14).

Table 1

Estimates of gross production (GP, mmol m−2 day−1) in the Sea of Galilee obtained from the Δ17O method [GP17)] and from bottle incubations with H2 18O spike [GP(H2 18O)]. TheGP(H2 18O) values represent depth integration over the entire euphotic zone. The Δdissvalues are in per meg units. For applying Eq. 3, we used Δmax = 159 per meg; values of O2solubility were taken from (19); piston velocities were calculated according to (13) from daily wind speeds and then averaged over 1 week. The error propagation forGP17) reflects uncertainties in Δmax, Δeq, and Δdiss.

View this table:

In the BATS Station, the situation is more complex because photosynthesis takes place in both the mixed layer as well as in the underlying thermocline. In Fig. 2, we show the depth distribution of Δdiss, temperature, and δO2/Ar (deviation of the O2/Ar ratio of dissolved gases from the O2/Ar ratio in air). Variations in δO2/Ar reflect production and consumption of O2 independent of physical processes affecting O2 concentration (15). In July, the δO2/Ar had a maximum at 40 m, corresponding to the typical subsurface oxygen maximum of the summer thermocline. As has been noted by Jenkins and co-workers (16), this originates from the seasonal net accumulation of photosynthetic O2. The Δdiss curve also has a maximum in the summer thermocline, indicating in situ gross O2 production. The maximum Δdiss is analogous to the subsurface δO2/Ar maximum—both maxima are formed from attenuation of vertical mixing caused by density stratification in the thermocline. In November, the mixed layer was deeper due to cooling and extended down to 60 m. As in July, a Δdiss maximum was present in the thermocline, but in contrast, there was no δO2/Ar maximum. The pronounced Δdissmaximum indicates continued in situ O2 production, but the relatively low values of δO2/Ar show that O2 uptake was greater than its generation. Finally, with further cooling in March the mixed layer extended to a depth below the photic zone, and in the absence of photosynthesis in the thermocline, the Δdiss maximum disappeared.

Figure 2

Depth profiles of δO2/Ar (‰ versus HLA), temperature (oC), and Δdiss (per meg versus HLA) in BATS Station near Bermuda. (A) Profiles for 7 July 1998. Note the maxima of δO2/Ar and Δdiss in the summer thermocline. (B) Profiles for 7 November 1998. Note the absence of the δO2/Ar maximum. (C) Profiles for 23 March 1999. Note the absence of both maxima.

Initial estimates of GP are given in Table 2. These estimates are time-integrated rates representing production in the mixed layer as well as gains and losses due to vertical mixing with the underlying thermocline. The residence time of O2 in the mixed layer is about 2 weeks, and thus effects of short events of high production are expected to average out by the lower background of the entire mixed layer. When the mixed layer is shallow and the thermocline is situated in the photic zone, the calculated GP rates should be considered as minimum values, because some of the production takes place below the mixed layer, as is evident in the Δdiss maximum in the thermocline (Fig. 2). This is probably an ocean-wide phenomenon, because in addition to BATS, we have observed Δdissmaxima in the thermocline in the Red Sea and in the equatorial Indian Ocean (17). Conversely, when deep mixing takes place in winter, the calculated GP should be considered as maximum because some of the dissolved O2 with high Δdiss in the thermocline is incorporated into the mixed layer. Over an annual cycle, the winter excess should compensate for the summer deficit, and an annual integration of the calculatedGP is expected to reliably reflect the true integrated production in a given area. Lateral transport has been neglected in this simple one-dimensional description, but it can be represented in future studies by applying general circulation models.

Table 2

Estimates of gross production (GP, mmol m−2 day−1) in the BATS Station obtained from the Δ17O method [GP17)]. The Δdiss values are in per meg units. For applying Eq. 3, we used Δmax = 249 per meg; values of O2solubility (C o, mmol m−3) were taken from (19); piston velocities (K, m day−1) were calculated from climatological wind speeds (20). Error propagation as in Table 1.

View this table:

The trend of seasonal variations of GP in BATS, with a late winter maximum and a summer minimum (Table 2), is similar to the general pattern of productivity observed in other studies (18). Annual integrated production was calculated from the data in Table 2 as 28 mol O2 m−2year−1. The integrated annual C fixation for the period 1988–94 estimated from 14C incubation experiments ranged from about 9 to 14 mol C m−2 year−1(18). Thus, annual O2 gross production is about two to three times the estimated annual C fixation. Similar ratios between O2 gross production and C fixation have been reported in studies based on direct comparisons of 14C and H2 18O incubation experiments (1), demonstrating the capability of the method we report to derive reliable estimates of gross production. Although more studies are needed in order to gain a comprehensive global picture, such a view is now more attainable.


View Abstract

Navigate This Article