A model study on the sensitivity of surface ocean CO 2 pressure with respect to the CO 2 gas exchange rate

Introduction Conclusions References


Introduction
Gas exchange is one of the important links between the ocean and the atmosphere regarding biochemical cycles.It plays an important role for global climate in particular with respect to atmospheric greenhouse gases.The ocean is an important carbon reservoir and a major sink for anthropogenic CO 2 which has been emitted to the atmosphere (Solomon et al., 2007).The exchange process is controlled by the gas concentration difference between the atmosphere and the ocean surface layer, while the gas flux further depends on the specific gas transfer rate, which is the product of the gas Figures transfer velocity and the gas solubility.In addition to the shifts in sea surface temperature and the surface wind stress, past and future greenhouse gas emissions provide changes in the ocean gas uptake.Trenberth et al. (2007) note that the strength of the mid-latitude winds have increased at least since the 1970s, especially in December, January and February.Model studies of Fyfe and Saenko (2006) confirm the strengthening and polewards shift of the Southern Hemisphere wind stress in the 20th century and predict the continuation of this trend during in the 21st century.These variations in the wind pattern provide changes in the gas uptake due to the wind dependency of the gas transfer velocity across the water-air interface.Studies of Zickfeld et al. (2007) and Tjiputra et al. (2010) predict an increasing carbon uptake in the high latitude Southern Hemisphere due to changes in the wind pattern and the resulting circulation changes.
In contrast, a study by Le Qu ér é et al. (2007) based on both forward and inverse analysis methods estimate using synoptic atmospheric forcing that the Southern Ocean carbon uptake has weakened by 0.08 PgC per year per decade relative to the atmospheric increase of CO 2 .This is attributed to the changing winds between 1981 and 2004 and is estimated to further weaken the Southern Ocean uptake strength in the near future.A study of Gruber et al. (2009) shows that there remain large uncertainties in the Southern Ocean carbon flux estimates.The study of Tjiputra et al. (2010) and numerous model studies presented in Meehl et al. (2007) predict a weakening of the Atlantic Meridional Overturning Circulation, which leads to a reduction of the CO 2 uptake in the North Atlantic, where according to Sabine et al. (2004) 23 % of the global anthropogenic carbon is stored.Observations suggest a decrease in the North Atlantic sink since 1990, linked to the phase changes in the North Atlantic Oscillation (Schuster and Watson, 2007;Watson et al., 2009).
Exploring the sensitivity of the carbon fluxes on the gas transfer rate can possibly improve our understanding of the uncertainties in future climate projections and the spread in results from different models on the ocean carbon cycle climate feedback (Friedlingstein et al., 2006).In the following study the effect of an increasing gas transfer rate is analyzed in 3 different ocean regions.The objective of the study is to assess Introduction

Conclusions References
Tables Figures

Back Close
Full the role of the gas transfer rate coefficient in influencing the global and regional uptake of CO 2 without changes in the ocean circulation.

The MICOM-HAMOCC isopycnic ocean carbon cycle model
We employ a state-of-the-art global biogeochemical ocean general circulation model (BOGCM) for our study.The physical component of the model is the Miami Isopycnic Coordinate Ocean Model based on the study by Bleck and Smith (1990) and Bleck et al. (1992).The current version of the model has been improved considerably as described in Bentsen et al. (2004) and Lohmann et al. (2009).The horizontal resolution of the model is approximately 2.4 × 2.4 • with 34 isopycnic layers.In addition, the model includes a single non-isopycnic surface mixed layer, which provides the linkage between the atmospheric forcing and the ocean interior.
The ocean biogeochemistry model is the Hamburg Ocean Carbon Cycle (HAMOCC5) model.It is based on the original work of Maier-Reimer et al. (1993), which has been extensively improved and used in many studies (e.g., Six and Maier-Reimer, 1996;Aumont et al., 2003;Maier-Reimer et al., 2005).The current version of the model contains an NPZD-type ecosystem model with multi-nutrient limitation on the phytoplankton growth formulation (Maier-Reimer et al., 2005).The surface pCO 2 in the model is computed from the prognostic variables temperature, salinity, pressure, dissolved inorganic carbon (DIC), and alkalinity.
The model was originally spun up for 950 yr (Assmann et al., 2010).While 600 yr of the spin up were performed with a monthly NCEP-based climatology (Kalnay et al., 1996), another 350 yr spin up was performed with repeated passes of the NCEP Reanalysis 1950-1999 to obtain realistic sea-ice distributions, surface mixing and mixed layers.The model simulations were all restarted in 1948 from an initialization run starting in 1860 using NCEP atmospheric forcing of the year 1959, since no gridded physical atmospheric data exists prior to that date (Assmann et al., 2010).Introduction

Conclusions References
Tables Figures

Back Close
Full A more detailed model description as well as the evaluation of the model performance on a global scale is documented in Assmann et al. (2010).The model simulations performed in this study are forced by the monthly atmospheric fields from the NCEP reanalysis.For the air-sea CO 2 fluxes computation, observed atmospheric CO 2 concentration from the Mauna Loa observatory in Hawaii is prescribed in the model.
Where available the monthly mean CO 2 and otherwise, the annual mean Mauna Loa CO 2 data are used to provide the atmospheric CO 2 concentration.

Gas Exchange
The gas flux between the ocean and the atmosphere is driven by the concentration difference between the atmosphere and the ocean and can be written, using the partial pressure difference of a gas as the thermodynamical driving force (Wanninkhof et al., 2009), as where pCO A 2 expresses the CO 2 partial pressure in the atmosphere and pCO O 2 the partial pressure in the ocean surface.S CO 2 describes the solubility of CO 2 in seawater and can be derived according to Weiss (1974) from temperature and salinity via (2) where T denotes the temperature of the sea surface, s describes the salinity of the seawater and A 1−3 and B 1−3 are constants for the calculation in molar units.During the last decades several research projects focused on the formulation of the gas transfer velocity k w .Deacon (1977) suggested in his formulation the transfer velocity k w to be a function of the dimensionless Schmidt number (Sc), the numerical constant (β) factor and the friction velocity (u * )

Conclusions References
Tables Figures

Back Close
Full Previous research focused on the quantitatively correct parameterizations of the gas transfer velocity (Liss and Merlivat, 1986;Wanninkhof, 1992;Wanninkhof and McGillis, 1999;Nightingale et al., 2000;Sweeney et al., 2007), resulting in different relationships between the gas transfer velocity and the wind speed.For the air-sea gas exchange, the model uses the formulation of Wanninkhof (1992) suggesting a quadratic relationship between the wind speed and k w .Furthermore, in ice covered areas, the gas exchange is scaled down to the ice free parts of the grid cell (Maier-Reimer et al., 2005).
3 Sensitivity experiment: regional changes of the gas fluxes The relations in Eqs. ( 2) and ( 3) suggest that the highest specific gas transfer rates arise in low temperature and strong wind regions.To investigate the effect of an increasing gas exchange rate on the total uptake of CO 2 , three different regions as listed in Table 1 were chosen.In the regions we substantially change the gas transfer velocity, however, do not take into account any change in the general pattern of gas exchange variability.Such a simplification is justified here as we first of all want to quantify the effect of a potential substantial change of the gas transfer on surface pCO 2 in general.The gas transfer rate in the various domains (Table 1) is increased in seperate experiments by 44 % which is equivalent to a 20 % increase in the wind speed as the model uses a quadratic relationship between the wind speed and the gas transfer velocity.The regions between 40-60 • N and polewards of 40 • S were selected because of their high specific gas transfer rates.The gas transfer was increased for the entire region polewards of 40 • S including the ice covered area polewards of 60 • S. The high specific gas transfer rates in these areas are caused by the influence of the strong westerly winds and the high gas solubilities due to the low sea surface temperatures.
Besides the high specific gas transfer rates in these regions the air-sea disequilibrium suggests these regions to be mainly CO 2 uptake regions.In contrast, the third region between 20 • S-20 Full warmer equatorial waters and the air-sea disequilibrium identifying large parts to be outgassing areas.
All the sensitivity runs were examined with the same model settings as for the standard run.To avoid changes in the ocean surface circulation, the wind field was left unchanged.The model integration time starts in 1948 and ends in 2009.To present the results the year 2000 was chosen because of the available literature values from Takahashi et al. (2009) in that year to compare the total integrated carbon uptake.

Results
Annual mean CO 2 flux differences for the reference year 2000 between the sensitivity experiments and the model standard run are shown in Fig. 1.The strongest flux differences are in the northern and southern westerlies region where flux differences exceeding ±2 molC m −2 yr −1 can be identified.In contrast the warm equatorial waters show flux differences less than ±0.5 molC m −2 yr −1 .Besides the differences in the experimental areas, changes in the carbon fluxes can likewise be discovered outside the experimental boarders.Water with higher carbon content gets transported by the surface currents resulting in an adaptation of the surface pCO 2 and the resulting gas fluxes outside the studied regions.
The increasing specific gas transfer rates influence the pCO 2 concentration as illustrated in Fig. 2. Changes in the surface pCO 2 result from modifications in the DIC concentration due to local variations of the carbon fluxes.Figure 2 indicates increasing partial pressures almost entirely in the uptake regions in the high latitudes and decreasing partial pressures in the equatorial region.The faster gas exchange results in pCO 2 differences up to ±10 ppm with the exception of the ice covered area of the highest latitudes of the Southern Ocean.Assmann et al. (2010) point out that the sea-ice cover in the model is overestimated.Furthermore, the artificial high pCO 2 underneath the ice is caused by the upwelling of remineralized organic matter resulting from the export production which is computed too far south in the model.Introduction

Conclusions References
Tables Figures

Back Close
Full The effect of an enhanced gas transfer rate on the global integrated CO 2 flux is visualized in Fig. 3.The net carbon uptake in the global ocean undergoes very little changes in all 3 scenarios compared to the standard run, varying mainly between ±0.1 PgC.Compared to the different scenarios experiment G03 shows stronger yearto-year variation caused by the ice covered area below 60 • S. A similar result can be identified observing the globally integrated mean pCO 2 difference between the runs, which vary less than ±2 ppm.Further analysis demonstrates that an increasing gas exchange rate does not simply show little effect on the global net uptake but has a notable effect on both the global upward and downward fluxes.Figure 4 illustrates the resulting fluxes plotted from 1990 onwards.Depending on the affected area the upward and downward fluxes increase simultaneously.The resulting fluxes for the year 2000 are displayed in Table 2.This result comes as a no surprise as the increase of the gas exchange rate shows its effect on both in-and outgassing.More surprising is that the slope shows no difference between the runs.By increasing the exchange rate of a net outgassing area one would expect a faster increase in the upward fluxes than in the downward fluxes.Instead, the lines only appear slightly shifted, leading to only small net differences between the runs.
To investigate the small annual differences an additional analysis for the year 2000 was performed on the intra-annual fluxes in the experimental regions.Figure 5 shows the monthly fluxes for the 3 experimental regions.The equatorial area shows a weak seasonal variation and describes a source for atmospheric CO 2 throughout the entire year, whereas both the high latitude regions are characterized by their intra-annual variability, but show a strong net uptake of CO 2 integrated over the entire year.The strong seasonality in the high latitudes of the Southern Hemisphere mainly results from the seasonal variation of the sea-ice.
Comparing the intra-annual fluxes of the standard run with the sensitivity runs, as illustrated in Fig. 5, indicates that the intra-annual signal gets strengthened in all cases.While the increasing gas transfer rate enhances the uptake from November until May in the Northern Hemisphere and from October until March in the Southern Ocean, Introduction

Conclusions References
Tables Figures

Back Close
Full the months of July until September and May until August indicate an equally strong increase of the outgassing fluxes of CO 2 .The equatorial area shows generally less variation.This outcome suggests that the intra-annual difference is related to the pCO 2 variation shown in Fig. 2. The increasing gas transfer rate leads to a perturbation of the surface pCO 2 in relation to the standard run, leading to higher partial pressures of CO 2 in uptake areas and lower partial pressures in supersaturated areas.In the high latitudes it results in less uptake or increased outgassing in months of low airsea disequilibrium, whereas the effect of the increased gas transfer rate dominates in months where the air sea disequilibrium is high.
Integrating the monthly fluxes between 40-60 • N for both the model standard and the sensitivity run G02 results in a net flux difference of −67 TgC for the year 2000, where −46 TgC is the contribution of the Atlantic and −21 TgC of the Pacific.This is more than twice as much as the −23 TgC calculated from Table 2 for the global ocean.The minus signs indicate more uptake by the ocean.These differences arise from the surface transport of water with high CO 2 concentration outside the experimental domain of artificially adjusted gas transfer velocities, where the gas exchange slows down due to a lower CO 2 disequilibrium (prescribed atmospheric pCO 2 minus increasing ocean pCO 2 outside the experimental domain) and unchanged gas transfer rate.A different result can be identified in the Southern Ocean.The integrated carbon flux difference for the experimental region G03 between the standard run (−1406 TgC) and the sensitivity run (−1344 TgC) is 62 TgC, hence, the increasing gas transfer rate leads to decreasing uptake in the year 2000.Figure 3 shows that this differs from year to year.Again, the identified flux deviates from the global net difference in Table 2, suggesting that transport of high CO 2 water equatorwards causes the global difference.Likewise, the equatorial region from experiment G04 shows a net decrease in the outgassing flux of −4 TgC in the year 2000 due to the increasing CO 2 partial pressure.
To investigate whether changing the gas transfer rate can improve regional model prediction, we compared the results from the model runs with the climatology of Takahashi et al. (2009) for the year 2000.The model output data were binned to the same Introduction

Conclusions References
Tables Figures

Back Close
Full 4 × 5 • grid boxes as presented for the climatology.Figure 6 shows the zonal averaged pCO 2 and CO 2 flux differences between the model runs compared to the Takahashi et al. (2009) data.The results from the sensitivity runs relative to the model standard run are plotted only in those latitudes where they differ from the standard run.While there is large disagreement between the model fluxes and the climatological data in the Southern Hemisphere, differences in the Northern Hemisphere mainly do not exceed ±50 TgC yr −1 .Besides the good correlation between the fluxes northwards of 30 • S (R = 0.88), the latitudes in the Southern Hemisphere southwards of 30 • S show a poor correlation (R = 0.23) which slightly improves by the increasing gas exchange rate in experiment G03 (R = 0.27).Figure 6 shows that both pCO 2 and CO 2 flux minima are shifted south relative to the observations and below 60 • S the ice cover causes the strongest disagreements.Therefore, in the following analysis, G03 is analyzed only between 40-60 • S. In addition to the small local improvements between the equator and 20 • S in experiment G04, Fig. 6 illustrates a remarkable deterioration resulting from experiment G02 at 40 • N due to the enhanced carbon fluxes.Beside the strong pCO 2 difference in the ice covered area of the Southern Ocean the strongest disagreement between the model outputs and the climatology data lie in the mid latitudes in both hemispheres and the high latitudes in the Northern Hemisphere.Stow et al. (2009) present a series of statistical tools to test the quality of the model output compared to observations.Some of these tools are used here to further investigate if the sensitivity runs improve the model prediction compared to the Takahashi et al. (2009) data for every zonal band.The results for the experimental areas are listed in Table 3.While the correlation coefficients between the CO 2 fluxes indicate that both the model and the sensitivity runs presents the climatology well between 40-60 • N and between 20 • S-20 • N, the correlation is negative between 40-60 • S due to the southward shift in the export production (Assmann et al., 2010).The root mean squared error (RMSE) and the average absolute error (AAE) results show that experiment G03 and G04 both decrease the discrepancies between model and observations, whereas experiment G02 shows and increasing bias.The model efficiency (MEF) indicates Introduction

Conclusions References
Tables Figures

Back Close
Full the accuracy of the model prediction relative to the average of the observations (Stow et al., 2009).A value close to 1 indicates a close match between model data and observations, whereas a value close to 0 states that the model predicts the observations not better than the mean of the observations.Negative values indicate that the observational mean would be a better predictor than the model output.Table 3 shows that model predictions are best between 20 • S-20 • N and get improved by the sensitivity experiment.A slight improvement can be discovered from experiment G03, whereas the results show that the observation mean is a much better predictor between 40-60 • S due to the shift in the biology.The area 40-60 • N is slightly better represented by the mean of the observations, whereas sensitivity run G02 shows a strong degradation of the model performance.The last result illustrates the potential to improve, or in this example rather deteriorate, the model prediction by adjusting the gas transfer rate in certain regions.
The pCO 2 RMSE and the AAE again show the largest model data discrepancies 40-60 • S, whilst the errors are in general smaller between 40-60 • N than between 20 • S-20 • N. The MEF indicates the same trend as before for the CO 2 flux.Experiment G03 and G04 represent the observation patterns better than the standard run, whereas G02 leads to a degradation of the model performance.Figure 6 reveals that both climatological CO 2 fluxes and pCO 2 data are lower between 40-60 • N compared to the standard run, whilst the increasing gas exchange rate in G02 leads to increasing fluxes and partial pressures relative to the model standard run.A decreasing gas transfer rate in contrast would potentionally lead to a better model performance in experiment region G02.Additionally, due to the stronger inter-annual variation of the CO 2 fluxes in G03 the model mean results from the years 1995 to 2004 were compared to the Takahashi climatology (not shown).The results indicate that averaged over the 10 yr period the G03 outputs show higher uptake fluxes compared to the model standard run and therefore do not improve the model performance for the CO 2 fluxes, whereas the pCO 2 values only slightly differ from the numbers obtained in Table 3.

BGD Introduction Conclusions References
Tables Figures

Back Close
Full

Conclusions
The model results indicate that regional changes in the gas transfer rate, as they are applied in this study, do not strongly effect the global air-sea net flux of CO 2 on the long run.The enhanced regional gas fluxes lead to a perturbation in the surface pCO 2 relative to the model standard run, resulting in a negative flux feedback as the concentration difference between the atmosphere and the ocean is the driving factor for the gas exchange.We therefore conclude that changes in the ocean carbon uptake are dominated by the changes in the ocean circulation rather than the gas transfer coefficient.
A comparison with climatology data from Takahashi et al. (2009) shows that an increasing gas transfer rate of 44 % in the experimental areas does not substantionally improve the model predictions in the higher latitudes of the Southern Hemisphere where the model data and the climatology data are poorly correlated.The results in the equatorial area and the Northern Hemisphere instead do support the idea that local enhancements of the gas transfer rate have the potential to improve local model predictions by adjusting the gas transfer rate.Furthermore, the study on the intra-annual variation between the sensitivity runs suggests that enhancements, some of them which are not considered in this study, could improve the local model performance relative to climatology data.Wanninkhof et al. (2009) suggest the gas exchange to be affected by naturally occurring surfactants, rain and bubble effects.An estimation of Woolf (1997) suggests a bubble contribution of 30 % on the global gas transfer velocity.Local model prediction results would undoubtedly benefit from including these effects.
This means, that after all, more detailed descriptions of variations in gas transfer velocities may have to be taken into account in advanced data assimilation schemes for parameter optimisations in biogeochemical ocean models.An omission of these descriptions may still result in acceptable global CO 2 uptake rates, but other processes than gas exchange may be corretced erroneously in order to correct locally for insufficient model representations of air-sea gas exchange processes.Introduction

Conclusions References
Tables Figures

Back Close
Full  Full  Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 1 .Fig. 2 .Fig. 4 .Fig. 5 .
Fig. 1.Annual surface flux difference between the model standard run and the G02 scenario (top), the G04 scenario (middle) and the G03 scenario (bottom) in molC m −2 yr −1 for the reference year 2000.Negative values describe increasing uptake fluxes and positive values describe decreasing uptake compared to the standard run.

Table 1 .
Description of the model simulations performed in this work.

Table 2 .
Takahashi et al. (2009)ntegrated carbon fluxes for the reference year 2000 and the best estimation fromTakahashi et al. (2009).