Photosynthetic Control of Atmospheric Carbonyl Sulfide During the Growing Season

Climate models incorporate photosynthesis-climate feedbacks, yet we lack robust tools for large-scale assessments of these processes. Recent work suggests that carbonyl sulfide (COS), a trace gas consumed by plants, could provide a valuable constraint on photosynthesis. Here we analyze airborne observations of COS and carbon dioxide concentrations during the growing season over North America with a three-dimensional atmospheric transport model. We successfully modeled the persistent vertical drawdown of atmospheric COS using the quantitative relation between COS and photosynthesis that has been measured in plant chamber experiments. Furthermore, this drawdown is driven by plant uptake rather than other continental and oceanic fluxes in the model. These results provide quantitative evidence that COS gradients in the continental growing season may have broad use as a measurement-based photosynthesis tracer.

Climate models incorporate photosynthesis-climate feedbacks, yet we lack robust tools for large-scale assessments of these processes. Recent work suggests that carbonyl sulfide (COS), a trace gas consumed by plants, could provide a valuable constraint on photosynthesis. Here we analyze airborne observations of COS and carbon dioxide concentrations during the growing season over North America with a three-dimensional atmospheric transport model. We successfully modeled the persistent vertical drawdown of atmospheric COS using the quantitative relation between COS and photosynthesis that has been measured in plant chamber experiments. Furthermore, this drawdown is driven by plant uptake rather than other continental and oceanic fluxes in the model. These results provide quantitative evidence that COS gradients in the continental growing season may have broad use as a measurement-based photosynthesis tracer. P arameterizations of carbon-climate feedbacks in climate models are based on climate sensitivities for photosynthesis and respiration that are highly uncertain (1)(2)(3). Measurementbased estimates of photosynthesis or respiration fluxes at large scales (>10 4 km 2 ) are needed to investigate these feedback mechanisms. Whereas photosynthesis and respiration-flux estimates have been made using eddy flux (4,5) and isotope techniques (6), robust tools for investigating these processes at large scales are currently lacking.
Recent work suggests potential for the use of atmospheric carbonyl sulfide (COS) as a photosynthesis tracer, on the basis of similarities observed between COS and CO 2 in a global airmonitoring network (7). The similarities are attributable to the simultaneous uptake of COS and CO 2 in photosynthetic gas exchange by terrestrial plants (8). COS has also been studied as a source of stratospheric aerosol (9,10), with recent reports suggesting that COS is a major source and that its contribution may be closely linked to continental surface fluxes (11,12).
Past models of COS plant uptake assume a 1:1 relation between relative uptake of COS and net primary productivity (NPP) (13)(14)(15)(16). This relation was challenged by plant chamber (8) and atmospheric measurement studies (7), which suggest a new model of uptake that is related to photosynthesis [gross primary productivity (GPP)] and yields four to six times the uptake of the NPP-based models. COS uptake is related to GPP because atmospheric COS and CO 2 diffuse at similar rates into stomata, dissolve at similar rates into intercellular plant water, and are consumed by photosynthesis enzymes (17,18). COS is taken up preferentially to CO 2 because photosynthesis enzymes transform one-third of the dissolved CO 2 in leaf water but irreversibly transform most of the dissolved COS (19,20). The remainder of the dissolved CO 2 diffuses back to the atmosphere. The chamber studies suggest a GPP-based uptake model where F is COS plant uptake, GPP is the photosynthetic uptake of CO 2 by terrestrial plants, [COS]/[CO 2 ] is the ratio of ambient concentrations, and V COS=CO2 is the leaf-scale relative uptake measured during plant chamber experiments. This GPP-based model was found in recent work to be qualitatively consistent with variations of global atmospheric measurements, suggesting that the total atmospheric lifetime of COS is only 1.5 to 3 years (7,21). However, it remains to be determined whether atmospheric COS measurements are quantitatively consistent with GPPbased plant uptake.
To provide a quantitative test of the relation between GPP and atmospheric COS, we compared atmospheric COS measurements from an airborne experiment with two simulations from a three-dimensional atmospheric transport model; one simulation was driven by the GPP-based uptake, and the second was driven by the NPP-based uptake. The airborne experiment, the Intercontinental Chemical Transport Experiment-North America (INTEX-NA), included 1741 daytime measurements of COS over continental North America between the surface and 12 km in altitude during July and August 2004. The experiment also included measurements of CO 2 (22) and many other species (23 (7). To interpret the INTEX-NA observations, we simulated COS and CO 2 concentrations over North America for the INTEX-NA period (24,25). The COS simulations were driven by plant uptake, soil sinks, and ocean and anthropogenic sources (direct and indirect). We calculated GPP-based plant uptake (Eq. 1) by scaling regional GPP fluxes (26) by leaf-scale relative uptake estimates. The NPPbased uptake and other surface fluxes were taken from a recent inventory of gridded surface fluxes (14). We simulated CO 2 concentrations using ecosystem, ocean, and anthropogenic surface fluxes. See the supporting online material (SOM) for details on observations and model simulations.
The persistent vertical drawdown (Fig. 1) and variability (Fig. 2) in the boundary layer are well represented by the GPP model, whereas the NPP model performs poorly. Plant uptake was found to be dominant over other sources and sinks in the continental COS budget during the growing season (Fig. 3), a necessary condition for the use of COS as a photosynthesis tracer. These results are discussed below.
The mean modeled and measured CO 2 concentrations along the INTEX-NA flight paths (Fig.  1A) show the expected net uptake of CO 2 and boundary-layer mixing during the growing season (27). The agreement between the observed and modeled drawdown indicates that atmospheric mixing is well represented in the model. Whereas model underestimation in the 2-to 5-km altitude range may suggest some deficiencies in the simulated mixing, there is only a 10% difference between the observed and modeled estimates of the column-integrated drawdown (21).  The mean COS vertical profile (from the 1741 INTEX-NA samples) also shows considerable drawdown in the boundary layer (Fig. 1B). The INTEX-NA concentration drawdown (difference between 6 to 8 km and 0 to 2 km in altitude) of 59.9 T 8.9 parts per thousand (ppt) [mean T 95% confidence interval (CI), n = 50 vertical profiles] for sampling in July and August of 2004 is consistent with the NOAA/ESRL drawdown of 55.9 T 19.0 ppt (mean T 95% CI, n = 12 airborne sites) for sampling in July and August of 2005 through 2007. The NPP model largely underestimates the observed drawdown as expected (13,16), whereas the GPP model has good agreement with the observed drawdown. As with CO 2 , some deficiencies in the GPP model are apparent in the 2-to 5-km altitude range. However, there is only a 15% difference between the observed and GPP model estimates of the column-integrated drawdown (21). The columnintegrated drawdown for the observed data was 4.2 times the NPP model estimate (21). Sensitivity analysis showed that the drawdown estimates are robust with respect to boundary condition uncertainty (21).
Maps of INTEX-NA boundary-layer COS concentrations show considerable variability in the boundary-layer observations (SD = 39 ppt), and this variability is similar to that in the GPP model (SD = 34 ppt) but much larger than that in the NPP model (SD = 24 ppt) (Fig. 2). The GPP model captures much of the observed variability in the boundary-layer COS concentrations with a correlation coefficient of 0.71 (n = 440). This GPP model performance is similar to the a priori performance in CO 2 studies that use a model-observations analysis to infer surface-flux estimates (28). The variability in the GPP model is largely driven by the magnitude of the plant uptake and caused by the mixing of background air with boundary-layer air that is depleted of COS (21).
The combined evidence from the concentration drawdown, column-integrated drawdown, boundarylayer variation, CO 2 profiles, and NOAA/ESRL data are consistent with the GPP-based model rather than the NPP-based model. Next, we consider the relative influence of the different surface fluxes on the COS airborne samples using transport simulations driven by only one surface flux at a time (Fig. 3A). Anthropogenic COS emissions (direct and indirect) are concentrated in the eastern United States but result in a boundarylayer enhancement that is less than one-third of the vegetative drawdown (21). The COS soil up-   take is <10% of the plant uptake, which is consistent with available field observations (29). Although ocean fluxes are a large source globally (14), they have only a small influence on the vertical profile for the continental growing season. There may be large missing sources in the global COS budget that could be important in relation to plant uptake for some regions (8). However, the good agreement of modeled and observed COS during INTEX-NA suggests that these missing sources are not located in North America during the growing season. Whereas the COS drawdown is dominated by plant uptake, the CO 2 drawdown has offsetting influences of photosynthesis and respiration components that are many times the net CO 2 drawdown (Fig. 3B). These offsetting components make it challenging to apply CO 2 measurements to separately investigate photosynthesis and respiration, reinforcing the need for a tracer such as COS. This evidence that continental-scale variations of atmospheric COS over North America are driven by the COS/photosynthesis relation suggests the potential use of COS as a carboncycle tracer. One tracer application is to estimate the ratio of GPP CO 2 fluxes relative to net ecosystem exchange (NEE) CO 2 fluxes using relations between simultaneous observations of COS and CO 2 . For vertical profile observations, a useful relation is the ecosystem-scale relative uptake (ERU), which is the ratio of the relative drawdown of COS to CO 2 (7,21). When plant uptake is the dominant flux, the ERU is proportional to the ratio of GPP/NEE with a proportionality constant that is the leaf-scale relative uptake. The INTEX-NA and NOAA/ESRL observations have a mean ERU of 5.7 T 0.6 (mean T SD, n = 31 vertical profiles, July and August 2004) and 5.7 T 2.1 (mean T SD, n = 10 airborne sites, July and August of 2005 through 2007), respectively. For a mean continental leaf-scale relative uptake of 2.2 (21), these ERU values imply a GPP/NEE ratio of 2.6, which is similar to estimates of GPP/NEE of crops during mid-growing season from eddy-flux studies (5,7,30). Both INTEX-NA and NOAA/ ESRL observations show lower ERU values over the mid-continent where C4 corn is extensive (Fig. 4). This regional depression in ERU could reflect a decrease in the ratio of GPP/NEE for C4 corn plants that has also been observed in eddy covariance studies of the North American growing season (5). Alternatively, lower leaf-scale relative uptake values have been hypothesized for C4 plants (7,8) and we suggest that the leaf-scale relative uptake and the ratio of GPP/NEE for C4 plants be explored more widely. Although the ERU could also be influenced by a regional anthropogenic COS source, the anthropogenic source relative to the plant uptake is rather weak in this region (21).
Another tracer application is to estimate the photosynthesis CO 2 flux with the use of an inverse analysis (31) of the COS model concentration error. Inverse analyses must consider multiple sources of model error, including transport parameterizations, different surface fluxes, boundary conditions, and representation error. For the COS inversion, the two flux parameters, other than GPP (Eq. 1), must be well constrained, and the COS fluxes other than plant uptake must be well known or relatively small, as may be the case for the continental growing season. One flux parameter, the concentration ratio parameter (Eq. 1), is well constrained by observations and has an observed variability of <10% in the INTEX-NA boundarylayer samples (n = 440). The other flux parameter, leaf-scale relative uptake, may be more uncertain because of the dependence on plant type and growth conditions, but the success of the simple leaf-scale uptake mapping used in this work is promising (21). Knowledge of leaf-scale uptake could be improved with additional plant chamber and ambient studies in support of efforts to recover GPP values.
The results presented here suggest global COS plant uptake and vertical gradients that are more than four times those predicted in previous global models ( [14][15][16]. This finding implies a large missing source in the global COS budget (7,8) and large uncertainty in how previous global models have predicted the transfer of COS to the stratosphere. Applications of the GPP model at a global scale should help to resolve uncertainties in COS budgets and could improve our understanding of the relation between COS surface fluxes and stratospheric aerosol. However, the most intriguing application may be to recover GPP and ecosystem respiration information by inverse analysis of atmospheric COS measurements.