The Impact of Orbital Precession on Air‐Sea CO2 Exchange in the Southern Ocean

Orbital precession has been linked to glacial cycles and the atmospheric carbon dioxide (CO2) concentration, yet the direct impact of precession on the carbon cycle is not well understood. We analyze output from an Earth system model configured under different orbital parameters to isolate the impact of precession on air‐sea CO2 flux in the Southern Ocean—a component of the global carbon cycle that is thought to play a key role on past atmospheric CO2 variations. Here, we demonstrate that periods of high precession are coincident with anomalous CO2 outgassing from the Southern Ocean. Under high precession, we find a poleward shift in the southern westerly winds, enhanced Southern Ocean meridional overturning, and an increase in the surface ocean partial pressure of CO2 along the core of the Antarctic Circumpolar Current. These results suggest that orbital precession may have played an important role in driving changes in atmospheric CO2.

• Increased insolation during austral summer due to orbital precession shifts the southern westerlies poleward • Poleward shifted westerlies enhance CO 2 outgassing due to increased turbulent exchange and vertical transport of carbon-rich waters • Enhanced transport of carbon-rich waters is driven by a deepening of the overturning circulation in response to poleward shifted winds

Supporting Information:
Supporting Information may be found in the online version of this article. 10.1029/2023GL103820 2 of 9 Recent studies suggest that the climate in the high-latitude Southern Hemisphere responds to orbital precession, one of the Milankovitch cycles with a spectral peak at ∼21,000 years.Orbital precession increases the seasonal cycle of insolation over one hemisphere while decreasing the magnitude of the seasonal cycle in the opposite hemisphere; this is especially pronounced at high latitudes.Modeling and proxy studies have demonstrated that precession can significantly alter the position and strength of the SWW, which impacts circulation in the Southern Ocean (Lamy et al., 2019;Rutberg & Broccoli, 2019).Yet the models used in these studies are lacking a carbon cycle and thus are unable to predict if Southern Ocean air-sea CO 2 flux will be affected by precession.
Here, we use a state-of-the-art Earth system model which includes a representation of the carbon cycle to illustrate, for the first time, that orbital precession can have a marked impact on Southern Ocean air-sea CO 2 flux.We compare output from two simulations with different precessional states to illustrate the potential influence of precession on the Southern Ocean.As we will demonstrate, precession drives changes in the SWW and Southern Ocean circulation, alters the upwelling of deep, carbon-rich water, and produces anomalies in air-sea CO 2 flux.
Our results suggest that orbital precession plays an important role in regulating atmospheric CO 2 concentrations, and provide a possible mechanism to explain the precessional peak in the ice core atmospheric CO 2 spectra.

Methods
Our primary numerical modeling tool is the low-resolution configuration of the Community Earth System Model (CESM) version 2.1.1 (Danabasoglu et al., 2020), a fully coupled climate model designed for long climate integrations (Shields et al., 2012).The atmospheric component, CAM4, has a resolution of ∼3.75° × 3.75° and 26 vertical levels (Neale et al., 2013).The ocean component, POP2, has nominal 2° latitude × 4° longitude resolution (lowering to less than 2° latitude resolution in the Southern Ocean), and 60 vertical levels (Danabasoglu et al., 2012;Smith et al., 2010).POP2 represents subgrid-scale processes, such as mesoscale and submesoscale processes, via a collection of parameterizations (Danabasoglu et al., 2008).Importantly, the Gent and Mcwilliams (1990) mesoscale eddy parameterization includes a variable eddy-induced advection coefficient (Gent, 2016) which improves realism of eddy-driven mixing of carbon in the Southern Ocean at coarse resolution (Lovenduski et al., 2013).POP2 includes a biogeochemical model, MARBL (Long et al., 2021).MARBL contains multiple chemical tracers necessary for simulating ocean biogeochemistry such as carbon, nitrogen, phosphorus, iron, silicon, and oxygen.
Our experiment was designed to isolate the impact of precession on the Southern Ocean.We spun up CESM 2.1.1 for a 1000-year period with an eccentricity parameter of 0; since eccentricity modulates the strength of precession, this equilibration period had no precessional forcing.Carbon dioxide in the atmosphere is kept at a constant preindustrial value of 284.7 ppm.Over the last 500 years of the spinup, the globally integrated air-sea CO 2 flux drift is negligible (−1.9 ± 6.5 × 10 −6 Pg C yr −2 ; Figure S2 in Supporting Information S1).Following the spin-up period, two 100-year simulations were performed: the first, NoPrec, maintains an eccentricity parameter of 0, while the second, HighPrec, uses an eccentricity parameter of 0.058, which is the maximum value over the last one million years (Laskar et al., 2004).A high value of eccentricity was chosen to maximize the precessional signal in HighPrec.Because the eccentricty parameter was kept as zero in NoPrec, there is no precessional impact on insolation (Figure S1a in Supporting Information S1).In the HighPrec simulation, the Southern Hemisphere summer solstice was configured to occur at the perihelion of Earth's orbit, which maximizes seasonal variability of insolation in the Southern Hemisphere (Figure S1b in Supporting Information S1).A full depiction of the differences in insolation between HighPrec and NoPrec is shown in Figure S1c in Supporting Information S1.We are confident in using 100-year means to study changes in the Southern Ocean because our HighPrec simulation illustrates an immediate response of SWW, despite a minimal drift in global mean surface temperature.This result is consistent with the expectation of a negligible impact of precession on annual mean insolation.Both HighPrec and NoPrec show minimal drift in globally integrated air-sea CO 2 flux over the 100 years studied; NoPrec has a global drift of 0.4 ± 5.3 × 10 −4 Pg C yr −2 and HighPrec has a global drift of 3.2 ± 4.5 × 10 −4 Pg C yr −2 .
While CESM2 is a well-validated model (Danabasoglu et al., 2020;Long et al., 2021;Simpson et al., 2020), here we employ CESM2 components with lower resolution than the standard configuration, requiring an assessment of model validity at this resolution.Of particular interest in this study is the position and strength of the SWW.While the maximum zonal wind stress in the NoPrec SWW (0.14 N m −2 ) agrees with modern estimates (Large & Yeager, 2009, 0.14 N m −2 ), the modeled position of the SWW zonal wind stress (centered on 45°S) shows an equatorward bias relative to the estimated preindustrial value (Large & Yeager, 2009, centered on 53°S), which 10.1029/2023GL103820 3 of 9 is a consequence of the lower resolution of the atmospheric model (Shields et al., 2012).The modeled position and strength of the SWW in our NoPrec simulation is within the range reported by models that participated in the Palaeoclimate Model Intercomparison Projects PMIP2 and PMIP3 under preindustrial conditions (Rojas, 2013), whose model components are simulated at a similar resolution as in our experiment.We find that the version of CESM2 employed in this study captures the air-sea fluxes of pre-industrial/natural CO 2 as compared to an observation-based inversion, as biome-mean CO 2 fluxes over the last 500 years of the spin-up simulation are within the uncertainty of the observation-based fluxes (Mikaloff Fletcher et al., 2007) (Table S1 in Supporting Information S1).
A goal of this study is to isolate the different physical processes driving changes in air-sea CO 2 flux over the Southern Ocean.We approach this using the air-sea CO 2 flux equation as solved by the model: where k sol is the solubility of carbon in seawater, A noice is the surface area without ice, ΔpCO 2 is the difference in the partial pressure of CO 2 between the surface ocean and the atmosphere, and k gtv is the gas transfer velocity which is driven by surface winds (Wanninkhof, 2014;Wanninkhof et al., 2013).We isolate the contribution from each process as follows: where  [ ]  is used to quantify the impact of precession-driven changes in solubility (δk sol ) on air-sea CO 2 flux.The δ symbol corresponds to the algebraic difference between the HighPrec and NoPrec value of the designated variable, and the non-δ terms are taken directly from the NoPrec simulation.
We expanded this technique to include effects from changes in the covariance among the terms in Equation 1.Only one of these terms, the combination of ΔpCO 2 and k gtv , was significant relative to the changes computed using Equation 2. The influence of joint changes in these two processes was calculated as follows: (3) which isolates the impact of simultaneous, precession-driven changes in both ΔpCO 2 and k gtv .It is important to note that since CO 2 in the atmosphere is kept constant, the ΔpCO 2 term in the air-sea CO 2 flux decomposition corresponds to only changes in surface ocean pCO 2 .In this analysis, it also became important to distinguish between the different variables controlling surface ocean pCO 2 .Because pCO 2 is not calculated in CESM with a simple equation as in air-sea CO 2 flux (Equation 1), this decomposition was instead performed using a first order Taylor Expansion.Contributions from temperature, salinity, Dissolved Inorganic Carbon (DIC), alkalinity, and freshwater forcing are distinguished in their effects on surface ocean pCO 2 (see Appendix A of Lovenduski et al., 2007).
In this manuscript, we emphasize orbital precession-driven changes in the Southern Ocean, calculated as the difference between the century-mean values in the 100-year HighPrec simulation and the 100-year NoPrec simulation.This difference is only reported to be statistically significant if it surpasses the 99% confidence level (Text S1 in Supporting Information S1).

Results
Our simulations show that high precession produces a shift in the SWW that manifests most strongly in the austral summer relative to conditions with no precessional forcing.In the summer months (DJF), we find a ∼6 m s −1 increase in the zonal-mean wind speed extending from 300 to 100 hPa and centered at 50°S; we also find a ∼3 m s −1 decrease at the same heights centered on 30°S (Figure 1b).Whereas, in the winter months (JJA), high precession leads to a general weakening of the SWW in JJA (Figure 1c).The shift in the SWW during the DJF season exceeds the SWW weakening in the JJA season, resulting in an annual mean shift (Figure 1a).This poleward shift in the SWW appears throughout the entire vertical structure of the atmosphere, indicating a poleward intensification of the surface westerlies that drive Southern Ocean circulation.
The simulated precessional shift in the SWW corresponds to large deviations in the atmospheric temperature structure.We find that the strongest temperature anomalies occur during the DJF season, due to austral summer receiving significantly more insolation in periods of high precession (Figure S3 in Supporting Information S1).
We find a precession-driven increase in the pole-to-Equator temperature gradient around 200 hPa in both the annual-mean and DJF zonal-mean temperature profiles (Figures S3g and S3h in Supporting Information S1) that corresponds to the greatest wind anomalies (Figures 1a and 1b).These findings indicate that periods of high precession, or periods when the perihelion of Earth's orbit occur at the Southern Hemisphere summer solstice, are associated with an enhanced pole-to-Equator temperature gradient at the approximate position of the tropopause.
Carbon outgassing in the Southern Ocean increases by approximately 20% in HighPrec relative to NoPrec.The century-mean, integrated (<35°S) air-sea CO 2 flux increases from 0.264 ± 0.013 Pg C yr −1 in NoPrec to 0.322 ± 0.014 Pg C yr −1 in HighPrec.The null hypothesis that these values are the same can be rejected with a 99% confidence level, and the uncertainty values reported here represent the standard error of the mean (Text S1 in Supporting Information S1).This precession-driven anomalous air-sea CO 2 flux is most pronounced in the Indian and Pacific sectors of the Southern Ocean: regions typically characterized by outgassing or weak uptake of CO 2 (Figure 2c).North of the ACC streamlines, and in the Atlantic sector of the ACC, high precession is associated with anomalous uptake of CO 2 (Figure 2).Differences in air-sea CO 2 flux between HighPrec and NoPrec change depending on the season, and the greatest increases in carbon outgassing are found in austral summer and austral autumn (Figure S4 in Supporting Information S1).The precession-driven anomalous outgassing exceeds the anomalous uptake, such that the Southern Ocean becomes a larger net source of CO 2 to the atmosphere under high precession relative to no precession.The precession-driven increase in Southern Ocean sea-air CO 2 flux is a result of changes in both surface ocean pCO 2 and the gas transfer velocity caused by changes in precession.We isolated the influence of each physical process driving changes in air-sea flux using the technique outlined in Methods.We find that the spatial pattern of the changes in CO 2 flux is driven by the contribution from ΔpCO 2 (determined by surface ocean pCO 2 since carbon in the atmosphere is constant) (Figure 3), which itself is impacted by changing surface ocean DIC (Figure S5 in Supporting Information S1).This indicates that the surface ocean pCO 2 response to precession drives the anomalous outgassing in the Indian and Pacific sectors of the ACC and the anomalous uptake in the Atlantic sector of the ACC.Because of the strong correlation between surface ocean DIC and surface pCO 2 (Figure S5e in Supporting Information S1e), we can attribute the majority of these changes to an altered ocean circulation.When integrated over the Southern Ocean (<35°S), the large magnitude positive and negative ΔpCO 2 anomalies nearly balance, such that the net contribution to the integrated flux difference is small (0.019 Pg C yr −1 ; Figure 3d).The precession-driven CO 2 flux difference is also strongly affected by the simultaneous changes in the gas transfer velocity and ΔpCO 2 , which contribute to enhanced outgassing in the ACC and a large, positive Southern Ocean integrated flux contribution (0.091 Pg C yr −1 ; Figure 3f).The changes in gas transfer velocity contributes a moderate decrease in carbon outgassing of −0.038 Pg C yr −1 (Figure 3e).Whereas, the changes in air-sea CO 2 flux due to sea ice extent (Figure 3b) and solubility (Figure 3c) have minimal impacts on the flux difference, with the exception of sea ice extent near the West Antarctic Peninsula which drives localized anomalous CO 2 uptake (Figure 3b).
The core of the Southern Ocean meridional overturning circulation shifts poleward and deepens under high precession, tapping into a richer carbon source explaining the increase in surface pCO 2 .Relative to the NoPrec simulation, both the wind stress and overturning maxima shift southward by ∼1° in the HighPrec simulation (Figure 4).In its more poleward position, the meridional overturning circulation streamlines intersect waters with higher DIC concentrations (Figure 4b).For example, the 20 Sv streamline in the NoPrec simulation intersects waters only up to 1,250 m deep with maximum DIC concentrations of ∼2,330 mmol m −3 .In contrast, this streamline reaches a deeper depth of 1,500 m in the HighPrec simulation overlapping with higher DIC concentration of ∼2,340 mmol m −3 (Figure 4).This shifted and deepened meridional overturning increases the amount of carbon that is brought to the surface in HighPrec relative to NoPrec, which is a key component (Figure S5 in Supporting Information S1) of the simulated increase in CO 2 outgassing.The largest increases in air-sea CO 2 flux occur where precession drives both enhanced gas exchange velocities and anomalous meridional and vertical advection of carbon-rich water (Figures 2-4, and Figure S6 in Supporting Information S1).High precession is associated with increases in the modeled air-sea gas transfer velocity, k gtv , near the northern core of the ACC; these increases are especially pronounced in the Indian and western Pacific sectors (Figure S6b in Supporting Information S1).The SWW changes that induce increases in near surface turbulence and air-sea gas exchange also alter the ocean circulation (Figure 4), driving increases in surface ocean DIC and pCO 2 in the Indian and western Pacific sectors of the ACC (Figures S6b and S5a in Supporting Information S1).Where the gas transfer velocity and pCO 2 anomalies align, they combine to produce enhanced CO 2 outgassing (Figures 2b and 4).

Conclusions and Discussion
Our study demonstrates that high precessional states impact key Southern Ocean processes involved in the global carbon cycle, ultimately leading to a substantial increase in sea-air CO 2 flux.Under high precessional forcing of the Southern Hemisphere, our model predicts a ∼1° poleward shift of the SWW across the troposphere, likely caused by insolation-driven atmospheric temperature changes over Antarctica.The associated poleward shift in the SWW drives a stronger and deeper meridional overturning circulation, enhancing the vertical and lateral advection of carbon-rich water.The shifted SWW also increase turbulent air-sea exchange which combined with the changes ocean overturning combine to produce a 20% increase in CO 2 outgassing from the Southern Ocean.
The precession-driven poleward shift in the SWW predicted by our model strongly resembles a positive phase of the Southern Annular Mode (SAM; see, e.g., Figure 7 of Thompson et al., 2000), albeit with a different seasonality.While the SAM pattern has been linked to internal climate variability and anthropogenic forcing, here we demonstrate that the Southern Hemisphere seasonal insolation changes associated with precession produce a similar shift in the SWW.The simulated change in the equator-to-pole temperature gradient in the upper troposphere is similar to that of the positive SAM phase, when the polar atmosphere shows cooling aloft associated with Ozone forcing (see Figure 8 of Thompson et al., 2000).Periods of high precession shift and deepen the meridional overturning circulation in our model (Figure 4b), which has also been found to occur during positive phases of the SAM (Yang et al., 2007).Thus, results from our simulations suggest that the Southern Hemisphere response to precessional forcing exhibits similar features to the Southern Hemisphere response to variability associated with the SAM, suggesting that past changes could be used to understand ongoing changes in Southern Hemisphere climate.
The precession-driven changes in ocean meridional overturning and air-sea CO 2 flux that we report somewhat agrees with other modeling studies that have tested similar relationships.These studies impose direct changes to the SWWs by strength, position, or both and measure the resulting changes in atmospheric CO 2 concentrations (d'Orgeville et al., 2010;Menviel et al., 2008;Tschumi et al., 2008).In Menviel et al. (2008), a positive relationship is found between SWW strength and carbon outgassing in the Southern Ocean.In addition, the modeling studies performed in Tschumi et al. (2008) find a positive relationship between the strength of the SWWs and global atmospheric CO 2 concentrations.However, Tschumi et al. (2008) also finds that a poleward shift in the SWWs reduces the amount of atmospheric CO 2 .This is attributed to a reduction in the area of outcropping deep water in the Southern Ocean.This disagrees with the results presented here.While it is difficult to diagnose the reason for this difference, it is likely because the changes in the SWWs between HighPrec and NoPrec have variance in both seasonal (Figure 1) and spatial (Figure S6a in Supporting Information S1) structure.This is because the SWW change in this study is due to precessional changes, not imposed directly.This disagreement may also be due to a difference in ocean models; Tschumi et al. (2008) uses the Bern3D which is a frictional-geostrophic balance ocean model (Edwards & Marsh, 2005;Müller et al., 2006). Finally, d'Orgeville et al. (2010) suggests that a poleward shift in SWWs result in an increase of Southern-Ocean carbon outgassing, which agrees with the results presented here.
Our study uses an Earth system model that is configured with relatively coarse horizontal resolution in the atmosphere and ocean model components to support long integrations potentially affect the realism of our results.The average annual peak in zonal-mean wind stress occurs at 45°S in our model.While this shows good agreement with other models that have similar horizontal resolution (see Figure 3 of Shields et al., 2012), this position is equatorward relative to the modern-day position of 53° Large and Yeager (2009).Similar poleward shift in the position of the SWW in response to high precession is also found in other modeling studies with higher resolution, suggesting our results are not model dependent For instance, Rutberg and Broccoli (2019) used a model with a resolution of 2° latitude by 2.5° longitude in the atmosphere and found a poleward shift of 4° between extreme precessional states.The coarse resolution of our ocean model component requires that processes influenced by mesoscale eddies are parameterized.Numerous studies have emphasized the importance of mesoscale eddies in Southern Ocean meridional overturning, especially in its response to changes in surface wind stress (Abernathey et al., 2011;Doddridge et al., 2019;Hallberg & Gnanadesikan, 2006;Marshall & Radko, 2003;Marshall & Speer, 2012).Our model uses a variable eddy-induced advection coefficient (Gent, 2016), which has been shown to capture the sensitivity of these unresolved processes to changes in circulation (Lovenduski et al., 2013).Indeed, results from our model indicate that the eddy-induced meridional overturning circulation strengthens in response to SWW changes under high precession (counterclockwise anomalies in Figure S7b in Supporting Information S1), suggesting that our coarse resolution ocean model component is capable of capturing changes in unresolved eddy advection.However, there are studies that suggest that this response is model dependent and would likely alter the circulation response seen here (Bishop et al., 2016;Böning et al., 2008;Meredith et al., 2012;Tesdal et al., 2023).Future work should explore the responses identified here using a higher resolution configuration of capable of resolving these processes.
Taken together, our findings imply that orbital precession plays an important role in regulating atmospheric carbon dioxide concentration through its effect on the Southern Ocean.The present study is focused on the impact of precession on Southern Ocean CO 2 fluxes; however, it is likely that other regions could be similarly affected by changes in precession.While this analysis stems from the results of a single model, individual components of the dynamical mechanisms presented here have been described in prior studies using diverse sources of simulations, paleo-proxy evidence, and instrumental observations (Butler et al., 2007;Dufour et al., 2013;Lamy et al., 2019;Lovenduski et al., 2007;Nevison et al., 2020;Rutberg & Broccoli, 2019).As we have demonstrated, the changes in seasonal insolation associated with orbital precession could have driven to a shift in the position of the westerly winds over the Southern Ocean, increasing the upwelling of carbon-rich water to the surface exchanging more carbon with the atmosphere.This mechanism could explain variability in ice core records of atmospheric CO 2 variability on precessional timescales (Petit et al., 1999).

Figure 2 .
Figure 2. (a) Century-mean sea-air CO 2 flux from the NoPrec simulation.(b) Century-mean sea-air CO 2 flux from the HighPrec simulation.(c) Precession-driven change in sea-air CO 2 flux, calculated as the difference in century-mean CO 2 flux from the HighPrec and NoPrec simulations.Stippling indicates a statistically significant difference at the 99% confidence level (Text S1 in Supporting Information S1).Units are mol C m −2 yr −1 , and positive values correspond to CO 2 outgassing.Black lines show the Antarctic Circumpolar Current (ACC) in the NoPrec simulation, bound by the 7 and 100 Sv barotropic streamlines.Numbers under each map indicate the Southern Ocean (<35°S) integrated flux and anomalous flux, respectively (Pg C yr −1 ).

Figure 1 .
Figure 1.Precession-driven anomalies in Southern Hemisphere zonal-mean wind speed (m s −1 ), calculated as the century-mean difference from the HighPrec and NoPrec simulations.(a) Annual-mean anomalies, (b) Austral summer (DJF) anomalies, and (c) Austral winter (JJA) anomalies.Stippling indicates a statistically significant difference at the 99% confidence level (Text S1 in Supporting Information S1).Positive values/contours correspond to westerly wind anomalies.

Figure 3 .
Figure 3. Contribution of (b) sea ice extent, (c) solubility, (d) ΔpCO 2 , (e) gas transfer velocity, and (f) the combination of gas transfer velocity and ΔpCO 2 change to the total air-sea CO 2 flux difference (mol C m −2 yr −1 ) due to precession.Contributions calculated as in Equation 2 and Equation 3 using the century-mean differences in each variable from the HighPrec and NoPrec simulations.(a) Shows the sum of the five components (b-f), which is nearly identical to Figure 2. Black lines show the Antarctic Circumpolar Current (ACC) in the NoPrec simulation, bound by the 7 and 100 Sv barotropic streamlines.Numbers under each map indicate the Southern Ocean (<35°S) integrated contribution to the anomalous flux (Pg C yr −1 ).

Figure 4 .
Figure 4. Southern Ocean response to high precession.(a) Century-mean zonal-mean surface wind stress from (gray) the NoPrec simulation and (black) the HighPrec simulation.(b) Century-mean (colors) zonal-mean DIC concentration from the HighPrec simulation with meridional overturning streamlines from the (gray) NoPrec simulation and (black) HighPrec simulation.Overturning units are Sv with contour lines every 10 Sv; positive streamlines indicate clockwise flow.Squiggly arrows indicate the relative position and strength in the annual peak carbon outgassing in both simulations.