Atmospheric Molecular Oxygen 21 Overview

Oxygen is one of the most abundant elements on earth. It is contained in most rocks and it is a fundamental constituent of the water molecule. Oxygen exists in the atmosphere in the form of molecular 02; with a content of 20.95%, it is the second most abundant atmospheric gas after molecular nitrogen. Atmospheric molecular oxygen is produced during photosynthesis by terrestrial vegetation and marine phytoplankton, and it is consumed during autotrophic respiration by plants and respiration of organic carbon by heterotrophic organisms on land and in the sea. Both of these processes also involve the transformation of carbon from inorganic to organic forms and vice versa; hence they form the fundamental linkage points between the biogeochemical cycles of carbon and oxygen. Since both processes involve water, they also constitute important linkage points between molecular oxygen and the hydrological cycle, which are important for isotopic exchanges as described further below.

Figure 1 shows in the upper two panels a simplified scheme of the natural global cycles of carbon and atmospheric molecular oxygen. The scheme depicts only the fundamental flows between global bicjgeochemical cycles in THE climate system

Copyright >; , 2001 by Academic i'rcs. Aii rights of reproduction in any torm reserved.

Carbon Cycle

Molecular Oxygen Cycle

Atmospheric C02


Atmospheric 02

Vege-j tation

i Air-Sea f Gasexchange

Vege- ! tation j

Litter + j Soils j


' MB Res

Litter + Soils




Air-Sea Gasexchange







Ground water


Stratospheric C02 Tropospheric C02

Oxygen Isotope Cycles

„ Photochemical _ Isotopic Exchanges

Stratospheric 02 Tropospheric 02

Air-Sea Gasexchange

Ground water




- Res h


Sea water


Litter h Soils

Air-Sea Gasexchange



n MB Resu


Ocean Ocean

FIGURE 1 Upper panels: simplified scheme of the natural global cycles of carbon (left) and atmospheric molecular oxygen (right). Ph, photosynthesis; Res, respiration; MB, marine biota; DIC, dissolved inorganic carbon (H2CO„ HC03~, CO,2-). Pool sizes are not shown to scale. Lower panels: corresponding schemes of the cycles of the oxygen isotopes in C02 (lower left panel) and in 02 (lower right panel). Dark blue arrows indicate links to oxygen isotopes in the hydrological cycle. Red whiskers on arrows indicate exchanges during which fractionation processes occur.

the atmosphere, the terrestrial biosphere, and the oceans. Minor exchanges with the geosphere (e.g., volcanism, weathering) and by river flows are neglected. Also ignored are the pools and reactions with minor atmospheric-carbon containing constituents: CO, CH4, hydrocarbons, etc. At first sight, the carbon and oxygen cycles seen almost reciprocal, with 02 and C02 being produced and consumed during photosynthesis and respiration in clearly defined stoichiometric ratios. However, there is a fundamental difference between the two cycles in the oceans. Atmospheric C02 is buffered by the large oceanic carbonate system (dissolved inorganic carbon: DIC = H2C03 + HC03~ + CO,2-), which comprises more than 50 times the carbon contained in the atmosphere. Depending on the time scale, any perturbation to atmospheric C02 is diluted to a considerable extent by this large oceanic carbon reservoir. On the other hand, molecular oxygen is dissolved in the ocean only in minute amounts; hence perturbations to the oxygen cycle are not damped by exchanges with the ocean. Therefore, the dynamics of the two cycles are quite different, and this forms the basis of the diagnostic approaches described in the sections below.

2.2 Measurement Techniques

In photosynthesis and respiration processes, about 1.1 mol of 02 is exchanged for 1 mol of C02, which implies that the induced atmospheric variations in 02 are similar in magnitude to those in

CO,. Hence, to be useful, atmospheric O, concentrations have to be measured with an accuracy of at least 0.1 ppnrv. Considering the background atmospheric 02 concentration of about 20%, this implies a measurement sensitivity of better than 10-6. Keeling and Shertz (1992) reported the first accurate measurements of variations in atmospheric 02 using an interferometric technique. Since these pioneering measurements, O, variations have also been measured by means of mass spectroscopy (Bender ct it/., 1996). Both of these techniques determine the ratio of 0,/N, of an air sample relative to a laboratory gas standard. Based on a rough quantitative assessment of the atmospheric N, cycle, it is easy to see that atmospheric variations of N2 are expected to be on the order of 10-8 or smaller, implying that changes in the 0,/N, ratio primarily reflect changes in O,.

The long-term maintenance of the constancy of the gas standards constitutes one of the big challenges in oxygen measurement work, because a technique accurate enough to determine the absolute oxygen content of air samples has not existed as yet. Recently, the group of R. Keeling has developed two new approaches to measure continuously 02/N, ratios in atmospheric air, one by ultraviolet spectroscopy (Stephens, 1999) and the other by measuring the paramagnetic susceptibility of oxygen (Manning ct al., 1999), both of which also demonstrate the required sensitivity. These new continuous measurement techniques have opened up the possibility of much more extensive global monitoring of 0,/N, ratios than has been possible so far.

02/N, ratios are commonly expressed in "permeg," a unit that is defined as the relative deviation of the measured 0,/N2 ratio from the standard multiplied by 106. Because the atmosphere contains 20.95 vol % 02, a variation of 1 ppmv of 02 corresponds to a shift in the 02/N2 ratio of 4.773 permeg.

2.3 Global Atmospheric Trends in C02 and 02

The most significant information from oxygen measurements to date has been the separation of the net terrestrial uptake from the oceanic uptake of anthropogenic excess C02. The burning of fossil fuels and the emissions from cement production induce an in creasing trend in atmospheric CO, concentration, and, because the burning of fossil fuels requires oxygen, a decreasing trend in atmospheric O,. The oceans and the land take up a sizeable fraction of the excess CO,. But, as discussed above, while terrestrial uptake by photosynthesis involves the generation of 02, the ocean does not affect the 02 balance. Hence global budget equations for C02 and 02 may be formulated as d

Here NCO a and N0 a a are the global atmospheric contents of CO, and 02, while Qfoss denotes the carbon emissions from the burning of fossil fuels and from cement production. Socean and S[and denote the CO, sinks on land and in the ocean./foss is the stoichiometric factor for the industrial emissions (mol of 02 consumed per mol of CO, generated) and /¡am1 is the stoichiometric factor for terrestrial carbon uptake. The term Qocean denotes potential outgassing of dissolved oxygen from the ocean, for example, that one induced by global warming. The size of this term is believed to be small although not entirely negligible.

The two equations are readily solved for the two unknowns Sland and Socean. Since they constitute two linear equations in two unknowns, their solution can also be represented in graphical form (Keeling et al, 1996; see also Fig. 2 below). Direct atmospheric measurements of 02/N2 started in 1989 (Keeling and Shertz, 1992) in La Jolla and at several stations in the early 1990s (Bender et al, 1996; Keeling et al, 1996; Battle et al, 2000). Using analyses of 02 in archived air from Cape Grim, Langenfelds et al. (1999) were able to extend the record back to 1979. Oxygen measurements have also been reported from air extracted from Antarctic firn, dating back to the late 1970s (Battle et al, 1996).

An update of the global budget representating 1990-1997 is presented in Table 1. The atmospheric 02 trends averaged over


Global 02 and C02 Budgets Averaged over 1990-1997: Numerical Values of Terms in

Eqs. (1) and (2)*



Standard Error

Variance Fraction to Error of Land Uptake (%)

Variance Fraction to Error of Ocean Uptake


— 15.6 permeg year-1

± 0.87 permeg year-1




1.34 ppmv year !

± 0.02 ppmv year-1



0.3 permeg year-! 1.94 GtC year-1 1.47 GtC year-1

± 0.6 permeg year-1 ± 0.65 GtC year-1 ±0.80 GtC year-1


*For the explanation of the last two columns see text.

*For the explanation of the last two columns see text.

this period have been determined by merging the data from Alert, Canada and La Jolla, California (Keeling et al, 1996) with observations from Point Barrow, Alaska and Cape Grim, Tasmania (Battle et al, 2000). In being merged, the records, were first desea-sonalized by fitting functions consisting of a seasonal cycle represented by four harmonics and a linear trend to the individual records. Subsequently, the records were merged based on the offsets determined by fitting linear trends to overlapping parts of the records. Then, annual averages overlapping by 6 months were formed. The global inventory change over 1990-1997 was determined from the difference between the annual averages of 1997 and 1990. Since observations in the early record were somewhat sparse, the value for 1990 was computed as the average of the five annual means centered at 1989.5, 1990.0, 1990.5, 1991.0, and 1991.5. The corresponding global average C02 trend was determined with a similar procedure from monthly C02 observations of the Point Barrow and the Cape Grim monitoring stations reported by the Climate and Monitoring Diagnostics Laboratory of the NOAA (Conway et al, 1994).

Table 1 includes also the error analysis and the fraction of the error variance of the land and ocean uptake estimates generated by the uncertainties in the individual terms. Interestingly, the uncertainty in the land uptake is dominated about equally by errors in the fossil fuel emissions and the atmospheric 02/N2 trend, while the uncertainty of the ocean uptake is dominated by the error of the global 02/N2 trend only. This behavior results from canceling effects in the solutions of Eqs. (1) and (2) for the two sink terms. Thus, although the oxygen budget [Eq. (2)] in principle only determines the land uptake term, it is the fact that we know the atmospheric C02 trend very well, which tightly couples the two equations and leads to a smaller overall error estimate of the ocean uptake term. It is readily seen that a further reduction of the error in the global 02/N2 trend will primarily reduce the error of the ocean uptake estimate. Table 1 also shows that the errors contributed by uncertainties in the stoichiometric factors (/¡and and /¡oss) are at present relatively minor.

There is also a significant uncertainty induced by the largely unknown ocean outgassing term (Qocean). The value 0.3 permeg year-1 adopted for this term in Table 1 reflects an ocean warming rate of the order of 1 W m-2 as inferred from oceanographic data (Levitus et al, 2000) and from global warming simulations (Roeckner et al, 1999). The conversion of warming to outgassing of 02 is computed with a ratio of approximately 1.5 X 10-9 mol 02 per J of heat uptake (see Keeling et al, 1993). Thereby the effect on the atmospheric 02/N2 ratio of the corresponding outgassing of N2 has also to be taken into account. Although the corresponding conversion factor of 2.2 X 10-9 mol N2 per J of heat uptake is slightly larger than that for 02, molecular nitrogen is four times more abundant in the atmosphere, hence the effect of thermally driven N2 outgassing on the atmospheric 02/N 2 ratio is only half as large as for 02 and of opposite sign. There is a considerable uncertainty in the value for the ocean outgassing component. Furthermore, a warming ocean may also affect the natural cycling of 02 between the atmosphere and the sea: the increased stratification of the ocean may prevent deeper, oxygen-undersatu-rated waters to come into contact with the atmosphere. This may result in an enhanced 02 outgassing as compared to the purely thermal outgassing effect described above. The magnitude of this enhancement, however, is very difficult to assess. In the present analysis we only include the direct thermal effect based on an ocean warming rate of 1 W m-2 and include the potential enhancement due to increases in stratification in the uncertainty estimate of this term. A graphical representation of the global budget equations (1) and (2) in the form of an arrow diagram is shown in Figure 2.

2.4 Seasonal Cycles and Mean Annual Spatial Gradients

The seasonal signal in atmospheric C02 in the northern hemisphere is mostly generated by the terrestrial biosphere (Heimann et al, 1986; 1989; 1998; Fung et al, 1987; Knorr and Heimann, 1995), oceanic seasonal fluxes being largely buffered by the ocean chemistry and the slow sea-air gas exchange of C02. This is not true for 02 which in the Northern Hemisphere, at least in oceanic



1990 i 1991 yv


1992 TV

1993 \ \

Oceanic Uptake


1994 ï\ \


\ \ Effect of Observations \ \ Fossil Fuel input


1995 n \ 1996 L \


1997 «





C02 [ppmv]

FIGURE 2 Globally averaged 02/N2 ratio (vertical axis) versus globally averaged C02 mixing ratio (horizontal axis). The annually averaged observations have been determined from the station records as described in the text. The black arrow shows the observed trend for 1990-1997 determined by the fitting procedure as described in the text. The red arrows depicts the expected change due to the fossil emissions during 1990-1997. The effects of the ocean and the land biosphere is shown with the blue and the green arrow, whereby their slopes are determined by their respective 02 versus C02 contributions (see inset). The purple vertical arrow reflects an estimate of the oceanic 02 outgassing induced by ocean warming.

areas, exhibits a seasonal signal about twice as large as the corresponding seasonal cycle in atmospheric CO,. Hence the magnitude of the seasonal oceanic component of O, is similar to that of the terrestrial component. In the Southern Hemisphere, with small land areas the oceanic O, signal dominates largely the seasonal signal (Keeling and Shertz, 1992).

Since terrestrial exchanges in O, occur in relatively fixed stoichiometric ratios (Severinghaus, 1995), it is the oceanic O, signal that is of primary interest for monitoring in the atmosphere. The terrestrial component can be removed from the atmospheric measurements conveniently by introducing of the tracer atmospheric potential oxygen (APO), conveniently defined as the atmospheric oxygen signal that would indeed result if all atmospheric carbon were oxidized with the stochiometric constant of terrestrial bios-pheric carbon (Stephens eta/., 1998):

APO = S02 + /¡and (fCOzl + 2 [CH,] + 0.5 [COj). (3)

I Iere S02 denotes the observed deviation of the atmospheric O, concentration from a standard. The atmospheric tracer APO is dominated primarily by oceanic gas exchanges in addition to a relatively small contribution from fossil fuel not accounted for by the terrestrial stoichiometric factor (i.e., the fossil fuel component scaled by the factor /tosi — fiilnLi). Observations of the seasonal variation of APO in conduction with surface-water oxygen measurements have been used to constrain the large-scale magnitude of the air-sea gas exchange coefficient (Keeling et al, 1998) and of marine productivity (Six and Maier-Reimer, 1996; Balkanski el al., 1999). Mean annual gradients of APO have also been shown to provide powerful constraints on biogeochemical air-sea fluxes computed by ocean-circulation models with an embedded ocean carbon cycle (Stephens et al., 1998).

2.5 Continental Dilution of the Oceanic 02/N2 Signal

Observations of the atmospheric 02/N2 ratio will also be an important tool to constrain the zonal transport of air between the continents and the oceans within the Northern Hemisphere. Recently, atmospheric measurements of C02 from the global monitoring networks have been used to discriminate net carbon balances of different continental-scale regions by inversion studies (e.g., Rayner et al., 1999; Fan et al, 1998; Kaminski et al, 1999; Heimann and Kaminski, 1999; see also the contribution by Rayner in this volume). While these approaches yield relatively robust estimates for the carbon balances of the whole northern and southern extra-tropics and the tropics, credible estimates for smaller regions such as Europe, Asia, or North America are difficult to establish. The reasons for this difficulty can be traced to the presently inadequate monitoring network (~ 100 monitoring stations, mostly located in oceanic areas) and to difficulties in faithfully representing the flushing of air over the continents and the oceans in the atmospheric-transport models employed in the inversion studies. Of particular concern are "rectifying" effects

(Heimann et al, 1986; Keeling et al, 1989; Denning et al, 1995) generated by the strong seasonally varying surface sources, such as C02. Because of the temporal covariance between seasonal atmospheric transport (seasonal changes of the vertical stability over the continents, monsoon circulations, seasonal ITCZ movements, etc.) and a seasonal source at the surface, mean annual atmospheric concentration patterns are generated. These patterns are in principle indistinguishable from the patterns generated by net annual sources and sinks. Hence if one wants to invert atmospheric concentration patterns in terms of net surface sources and sinks, the "rectifying" patterns have to be correctly represented in the employed transport models. However, the magnitude of these effects is largely unknown; different atmospheric models yield dramatically different "rectifying" patterns when forced with the same purely seasonal surface source (Law et al, 1996). Unfortunately, there does not exist a direct atmospheric tracer that can be used to evaluate the rectifying effects simulated by the various models.

As described in the previous section, the tracer APO has no significant seasonal sources over the continents, but is mostly generated by the seasonal oceanic O, exchanges. Hence the dilution of this signal into the interiors of the continents in principle provides a test for the transport representation in the models. Of course, the usefulness of this test depends on continental O, monitoring stations, which currently are not existing, but are planned for the near future.

As an example, Figure 3 shows the modeled amplitude of the seasonal signal of terrestrial CO, and of the oceanic O, (oceanic component of APO) in the lower planetary boundary layer (at about 380 m above the surface) simulated with the global TM3 transport model (updated from Heimann, 1995) by using a horizontal resolution of approximately. 4° latitude by 5° longitude and 19 layers in the vertical dimension. The model predictions of the Hamburg model of the oceanic carbon cycle (HAMOCC3; Maier-Reimer, 1993) with the phytoplankton-zooplankton model of Six and Maier-Reimer (1996) have been used to prescribe the monthly oceanic 02 exchanges (atmospheric simulation in the upper panel). The simple diagnostic biosphere model (SDBM) of Knorr and Heimann (1995) has been used to prescribe the seasonal terrestrial sources in the C02 simulation (lower panel). The significant zonal structures of the oceanic 02 amplitude field and its dilution over the continents is a feature that remains to be verified by atmospheric measurements.

Detecting the dilution of the oceanic seasonal cycle signal in 02 over the Northern Hemisphere continents is relatively straightforward and does not involve a detailed analysis and determination of the seasonal signal. All that has to be monitored is the O, versus C02 relationship over the course of one year. Since the seasonal signals in O, from the ocean and the terrestrial biosphere are closely in phase, this 02-C0, relationship is expected to fall approximately on one line, with, however, a slope determined by the magnitude of the oceanic signal. The principle is shown in Figure 4. If there were no oceanic contribution, the slope would merely reflect the biosphere stoichiometric factor (—1.1). A larger slope indicates a significant oceanic contribution. Figure 5 shows the

-180 -150 -120 -90 -60 -30 0 30 60 90 120 150
-180 -150 -120 -90 -60 -30 0 30 60 90 120 150

0 5 10 15 20 25 30 35 40 C02 Amplitude [ppmv]

FIGURE 3 Amplitude of the seasonal signal in the lower planetary boundary layer (at approximately 380 m above the surface) generated by the terrestrial biosphere in the C02 mixing ratio (lower panel) and by oceanic exchanges in the atmospheric 02/N2 ratio (upper panel) as simulated with the TM3 atmospheric transport model. See text for the model setup description.

0 5 10 15 20 25 30 35 40 C02 Amplitude [ppmv]

FIGURE 3 Amplitude of the seasonal signal in the lower planetary boundary layer (at approximately 380 m above the surface) generated by the terrestrial biosphere in the C02 mixing ratio (lower panel) and by oceanic exchanges in the atmospheric 02/N2 ratio (upper panel) as simulated with the TM3 atmospheric transport model. See text for the model setup description.

spatial variation of the slope between O, and CO, in the lower troposphere resulting from the model simulations described above. The color code has been chosen, such that only the variations in the Northern Hemisphere are highlighted; i.e., where the aforementioned relationship between the terrestrial and oceanic seasonal cycle is expected to hold. This is no longer the case in the tropics and in the southern hemisphere, where a more complex relationship exists between 02 and C02. It is seen that the slope of the relationship over the Atlantic and Pacific oceans reaches values above 2. Over the interior of the continent this ratio is progressively reduced to values of 1.3-1.4. It is expected that this pattern will vary considerably between different models, and should therefore provide a critical check on the realism of the modeled transport.

0 0

Post a comment