A detectable change in the air-sea CO2 flux estimate from sailboat measurements

doi:10.21203/rs.3.rs-3468622/v1

Download PDF

Article

A detectable change in the air-sea CO2 flux estimate from sailboat measurements

https://doi.org/10.21203/rs.3.rs-3468622/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 09 Feb, 2024

Read the published version in Scientific Reports →

You are reading this latest preprint version

The sailboat Seaexplorer collected underway sea surface partial pressure of CO₂ (pCO₂) data for 129 days (2018–2021), including an Antarctic circumnavigation. By comparing ensembles of data-driven air-sea CO₂ fluxes computed with and without sailboat data and by applying a detection algorithm, we show that these sailboat observations significantly change the regional air-sea CO₂ flux density in the North Atlantic and Southern Ocean. While the globally integrated effect is small due to the compensating changes in both basins, the Southern Ocean, particularly frontal regions (40–60°S) during summertime, exhibited the largest air-sea CO₂ flux changes averaging 20% of the regional mean. Assessing the sensitivity of the air-sea CO₂ flux, the results stay robust within the expected random measurement uncertainty (± 5 µatm) but remain undetectable with a measurement offset of 5 µatm. We thus conclude that sailboats fill essential measurement gaps in remote ocean regions.

Earth and environmental sciences/Biogeochemistry

Earth and environmental sciences/Biogeochemistry/Carbon cycle

Earth and environmental sciences/Climate sciences

Earth and environmental sciences/Climate sciences/Ocean sciences

Earth and environmental sciences/Ocean sciences

Earth and environmental sciences/Ocean sciences/Marine chemistry

Earth and environmental sciences/Ocean sciences/Physical oceanography

The ocean plays a critical role in regulating Earth's climate by absorbing more than a quarter of anthropogenically emitted carbon dioxide (CO₂) from the atmosphere on an annual basis^1–3. However, climate change has already started to alter the carbon uptake capacity of the ocean^1,4, thus monitoring the sea surface CO₂ content is crucial for understanding the Earth system as a whole. Although there has been a significant community effort resulting in the collection and collation of sea surface CO₂ observations^5,6 in recent decades, and methods to upscale the existing measurements^7–12 we find a significant difference between hemispheres. While the Northern Hemisphere has been regularly sampled in the recent past being the result of the community-driven measurement efforts resulting from the Ship Of Opportunity (SOOP) program^6,13, key regions in the ocean carbon and heat uptake such as the Southern Ocean remain undersampled^5,14,15 and the resulting uncertainty in air-sea CO₂ fluxes is worrisome^16,17 limiting our ability to resolve and interpret observed and modelled variations in the carbon sink^18–20. This is concerning as the Southern Ocean alone is estimated to be responsible for 40% of the marine anthropogenic CO₂ and 75% of the marine excess heat uptake^7,21.

New techniques, including new sensors on biogeochemical floats, have started to address this observational gap, but their indirect measurements of pCO₂ – calculated from pH and salinity measurements - remain uncertain^22–24. Additionally, Antarctic operations from Saildrones²⁵ have contributed to filling the measurement gaps and are suggested to improve the air-sea CO₂ flux estimates²⁶, however, thus far no continuous measurement program exists. Given the limitations of the existing observational network and the moderate success of gap-filling methods in further improving pCO₂ estimates^16,17, it is essential to explore new opportunities to fill observational gaps.

Here we show that a novel observing platform is capable of improving our estimates of the air-sea CO₂ exchange. Since 2018, the high-performance IMOCA class 60 sailboat “Seaexplorer-Yacht Club de Monaco“ has collected pCO₂ observations (hereinafter: Seaexplorer data) while competing for 129 days in round-the-world racing events, including an Antarctic circumnavigation race from November 2020 to January 2021²⁷. We show that even a single platform (“Seaexplorer”) and in the Southern Ocean a single race has a measurable effect on data-driven air-sea CO₂ flux estimates, even when considering its expected measurement uncertainty of ± 5 µatm²⁸, thus having the potential to complement and improve the existing observing system. Nevertheless, we further illustrate that high standard measurements are crucial in detecting changes in the air-sea flux and that measurement biases still pose a challenge for detecting improvements in the air-sea CO₂ flux estimates.

Global effect of adding sailboat pCO ₂ data. Figure 1a and 1b show the air-sea CO₂ fluxes calculated based on the upscaling of all available pCO₂ measurements including (ensemble 1 = E1) and excluding Seaexplorer data (ensemble 2 = E2) using SOM-FFN, a 2-step neural network method²⁹ - see methods - regularly used in the Global Carbon Budget¹ and the recent IPCC assessment³⁰. The significant impact of adding all underway pCO₂ observations from the sailboat on the air-sea CO₂ flux from November 2020 through January 2021 is further illustrated in Fig. 1c. We chose this time period from November 2020 to January 2021 since the impact was greatest there related to the Antarctic circumnavigation race where Seaexplorer participated (see black lines in Fig. 1). Interestingly, significant differences between E1 and E2 in the North Atlantic (largely positive shown in red: E1 < E2) and the Southern Ocean (largely negative shown in blue: E2 > E1) in the air-sea CO₂ fluxes are opposing each other (Fig. 1), resulting in an insignificant change when integrated globally (i.e. an annual flux difference in 2021 from 2.55 to 2.51 ± 0.4 Pg C yr⁻¹¹), which has also been suggested by ²⁷.

The E1 flux is larger than E2 in the North Atlantic but smaller in the Southern Ocean. Similar but slightly smaller differences occur when averaged over the actual years when sailboat races occurred (2018–2021 - Supplementary Fig. 1). Considering that both the North Atlantic and the Southern Ocean are predominantly carbon sinks during that time, the addition of Seaexplorer data reveals reduced carbon uptake in the North Atlantic and increased uptake in the Southern Ocean.

Differences in the flux estimates are visible across all ocean regions. However, in many regions, these differences fall within the noise of the method³¹ (see methods) and are thus not detected as significant changes (hatches in Figs. 1 and 2). This is most visible in the high-latitude ocean regions and is likely due to the poor constraint of the air-sea CO₂ flux estimate in highly heterogeneous and sparsely observed regions^16,17. Focusing on the detectable changes, irrespective of the background fluxes, the absolute magnitude of the difference between flux estimates provides a better insight (Fig. 2).

Impact of adding sailboat pCO ₂ data in the Southern Ocean. In less frequently monitored regions like the Southern Ocean, even adding Southern Ocean CO₂ measurements from a single track results in a significant difference between E1 and E2 (Fig. 2a). This aligns with the findings presented in ²⁶ based on synthetic data demonstrating that little additional pCO₂ sampling by Saildrone would potentially improve the air-sea CO₂ flux reconstructions most in the Southern Ocean (south of 35 °S). The reconstructions of our air-sea CO₂ flux density differ most significantly between 40–60 °S with peak differences of 0.77 mol C m⁻² yr⁻¹, reflecting the rate of carbon exchange between the atmosphere and the ocean per unit area, in the time period from 1982 to 2021 in the Southern Ocean (Fig. 2b and 3a). Overall, the absolute air-sea CO₂ fluxes significantly differed on average by 0.15 mol C m⁻² yr⁻¹ in the Southern Ocean (Supplementary Fig. 2), which is roughly 20% of the regional mean flux density, thus leaving a significant imprint on the regional flux.

The impact of including the Seaexplorer data in the air-sea CO₂ flux calculations is the largest within 2-degree grid cells (or roughly 200 km) of the Subantarctic Front closely followed by the Northern Boundary (Figs. 2a and 3c). Although the sailboat did not cross the Polar Front, significant differences emerge in its vicinity (Figs. 2a and 3c) due to the extrapolation of the data using the neural network algorithm. This pattern coincides with the coverage of the Seaexplorer data, as the region along the Subantarctic Front contained most Seaexplorer data with an overall 11% of the area covered by sailboat tracks when binned into a 1x1 degree grid, followed by 9% along the Northern Boundary, and 2% in the vicinity of the Polar Front.

Regionally, we find the largest differences during the Antarctic circumnavigation race between Nov 2020 – Jan 2021 exceeding 0.4 mol C m⁻² yr⁻¹ in the interfrontal region south of Tasmania and New Zealand (Figs. 1 and 2a) where the pCO₂ signal measured by sailboats is oversaturated and distinctly higher compared to the surrounding area²⁷, illustrating the importance of pCO₂ observations from frontal crossings. Although the region south of Tasmania and New Zealand is relatively well-observed for the Southern Hemisphere⁵, data close to frontal zones, regions of enhanced vertical mixing and high biological productivity, are limited³². Our results here show the high potential of sailboat pCO₂ data in improving the air-sea CO₂ flux estimate in frontal regions.

Although our results confirm the finding that regional differences in the air-sea CO₂ flux are attributed to the frontal zones in the Southern Ocean, the proposed changes south of the Polar Front by ²⁷ probably stem from noise in the methodology and not from a detectable signal. This underscores the need for signal-to-noise detection methods as presented here or alternatively, the use of synthetic data experiments using large ensembles^16,26 when comparing different air-sea CO₂ flux estimates from neural networks.

Impact of adding sailboat pCO ₂ data in the North Atlantic. Compared to the Southern Ocean, individual races in the North Atlantic are less impactful (Fig. 2 and Fig. 3c), largely owing to the already denser observing network in place where the addition of a single measurement track does not cause large significant changes in the already robust reconstruction. Nevertheless, we still observe that sailboat pCO₂ measurements have a significant regional impact since Seaexplorer data covers a total of 7% of the North Atlantic area (when binned into 1 x 1° pixels), in comparison to only 3% of the Southern Ocean area.

The air-sea CO₂ fluxes significantly differed regionally peaking at 1.26 mol C m⁻² yr⁻¹ in the North Atlantic between 1982 and 2021 (Fig. 3a), which is thus higher than the maximum flux difference of 0.77 mol C m⁻² yr⁻¹ in the Southern Ocean. However, the mean difference of 0.08 mol C m⁻² yr⁻¹ in the North Atlantic is substantially smaller than observed in the Southern Ocean (0.15 mol C m⁻² yr⁻¹) (Supplementary Fig. 2), considering the historic coverage of the SOOP program. In recent years however, we also find a reduction in North Atlantic measurements (www.socat.info; ⁵), thus even in the better observed North Atlantic the sailboat data might gain more importance.

Temporal evolution. Comparing the flux reconstructions E1 and E2 over time, we see the greatest impact of adding Seaexplorer data from 2018–2021 in the air-sea CO₂ flux estimates in the latter years of the time series (Fig. 3). About 91% of the significant differences between E1 and E2 occurred between 2018 and 2021, which is when the sailboat pCO₂ observations were measured. The pCO₂ data collected during the Antarctic circumnavigation race affects the air-sea CO₂ flux estimate only up to ca. 5 years prior to the race. This is not immediately obvious, since the applied method extrapolates information both in space and time, learning from all available observations when clustering the ocean into biogeochemical provinces and estimating the missing pCO2 values by using previously established relationships between already available pCO2 and other environmental variables within each province. However, a similar observation, where differences become smaller as we look further back in time, has been made when BGC Argo data were previously added²². This is explained by trend variables used in the method²⁹ limiting the potential of the method when extrapolating the missing pCO₂ values into the past²². As a consequence, we expect that a longer time series is required to change the interannual to decadal air-sea CO₂ flux trends. Nevertheless, with upcoming races announced (round-the-globe racing events taking place every other year) and with the increasing willingness among skippers to contribute with observations, we see a long term potential to increase pCO₂ data in remote ocean regions to overcome this limitation.

The addition of Seaexplorer data has the highest impact on austral summer, whereas it has little to no impact on austral wintertime fluxes (Fig. 3), mirroring the seasonal availability of data and illustrating the weak connectivity between seasons identified in our neural network. Therefore, sailboat measurements - unlike saildrone campaigns^15,16 - currently are unable to bridge the wintertime discrepancy between float-based and ship-based flux estimates²². Even though sailboat pCO₂ data have limited added value during harsh winter conditions in the Southern Ocean where no sailboat racing events take place, we show that sailboats support the existing observing system of Argo floats^22,23,33, Saildrones^25,34, moorings, drifting buoys, and wave gliders.

Sensitivity of air-sea CO ₂ flux to measurement uncertainty. Finally, we also tested whether potential measurement uncertainties or even measurement bias has an effect on the air-sea CO₂ flux estimate by considering a random measurement uncertainty of ± 5 µatm (simulation E3) and a constant measurement offset of 5 µatm (simulation E4) (Fig. 4). Figure 4a illustrates that a random measurement uncertainty of ± 5 µatm does not affect the air-sea CO₂ flux at a basin-wide level in the North Atlantic and the Southern Ocean, as the mean difference (E1-E3) is near zero for both basins (Fig. 4a). However, locally the air-sea CO₂ flux densities can be significantly impacted by up to 0.65 mol C m⁻² yr⁻¹ in the North Atlantic during individual months and up to 0.32 mol C m⁻² yr⁻¹ in the Southern Ocean. This highlights the importance of accounting for measurement uncertainty when investigating high-frequency and small spatial scale fluxes which will become increasingly important as we move towards marine carbon accounting, marine carbon dioxide removal and national carbon stocktake efforts^35,36.

We also test for potential limitations of the system and imperfect calibration over long offshore racing events by testing a constant measurement offset of 5 µatm. This causes a global bias up to 0.06 Pg C yr⁻¹ (E1-E4) in 2021 (Supplementary Table 1), which is on the one hand only roughly 2.5% of the global annual uptake, showing the rather small sensitivity of the air-sea CO₂ flux, to biases in a single platform, however on the other hand it is larger than the global flux change from adding Seaexplorer data of 0.04 Pg C yr⁻¹ (Supplementary Table 1). Thus our comparison highlights that flux changes from measurements from 129 days at sea remain undetectable if the measurement system does not follow the highest standards, supporting the need for a CO₂ reference network³⁷. However, while important globally, the constant measurement biases are still smaller at basin scale. The mean absolute difference in the air-sea CO₂ flux attributed to the offset is 0.03 mol C m⁻² yr⁻¹ in the North Atlantic and only 0.07 mol C m⁻² yr⁻¹ in the Southern Ocean (Fig. 4b), which is smaller than the mean differences caused by adding Seaexplorer data.

Particularly in the North Atlantic, the flux estimate proves to be more robust than the Southern Ocean flux estimate towards a potential measurement offset, again linked to the already denser observations from different platforms. This however also indicates the need for cross-calibration of measurements in remote regions, as measurement biases have a larger impact there. Locally, the air-sea CO₂ flux densities (based on E1-E4) significantly differed up to 0.76 mol C m⁻² yr⁻¹ in the North Atlantic during individual months and up to 0.38 mol C m⁻² yr⁻¹ in the Southern Ocean. The spatial distribution of significant differences between the original flux estimate and the flux estimate including a sailboat measurement offset follows the pattern of the sailboat tracks with the highest differences between 40°S and 60°S (Fig. 4c). Equipping more sailboats with a pCO₂ measurement device during the round-the-world races would help to reduce the impact of potential measurement uncertainties and increase the accuracy of our flux estimate. Considering that many studies thus far do not include or assess the impact of measurement uncertainty in their pCO₂ observations^7,38, we hereby show the importance of measurement uncertainty analyses at a regional scale to provide a more accurate estimate of high-frequency fluxes.

We quantify the impact of underway pCO₂ data from sailboats on the air-sea CO₂ exchange by comparing the air-sea CO₂ flux estimates based on pCO₂ measurements of the SOCAT database – with and without the Seaexplorer data and show that measuring pCO₂ underway and particular during round-the-world sailing events improved air-sea CO₂ flux reconstruction at regional scales, particularly in under-sampled regions like the Southern Ocean. We nevertheless further find that locally also in the better observed North Atlantic significant flux density changes occur. The flux reconstructions differ the most between 40°S and 60°S, particularly within 200 km of the Subantarctic Front during austral summertime where the largest disagreement in air-sea CO₂ flux reconstructions exist^39,40. Even though sailboat data cannot help to close the winter discrepancy between float-based and ship-based flux estimates²², due to the seasonal sampling bias, it supports the existing observational platforms (www.socat.info) and the data can be utilized to reduce air-sea CO₂ flux uncertainties, particularly in combination with various other platforms, particularly from Argo floats and Saildrones in the Southern Ocean⁴¹.

Rare underway pCO₂ observations collected close to the frontal zones changed the air-sea CO₂ flux estimate the most and can help to better understand these regions and their role in carbon uptake and the longer-term variation of the air-sea CO₂ exchange. Compared to the Southern Ocean, races in the North Atlantic were less impactful due to the historical stronger observing network there. However, since the majority of races took place there, we still see sailboat pCO₂ observations having a significant impact on regional air-sea CO₂ flux densities. Thus, our analysis suggests that a declining number of observations in the North Atlantic as we currently see (www.socat.info), may lead to a significant impact on the global ocean carbon flux estimates³⁰.

Added random measurement uncertainty (± 5 µatm) has a low impact on the overall air-sea CO₂ flux estimate due to compensating errors. However, we illustrate the importance of including measurement uncertainty locally when investigating high-resolution fluxes. A potential measurement offset of 5 µatm on the contrary results in a global flux bias larger than the detectable change due to 129 days of sailboat measurements. The impact of the measurement bias was more pronounced in the data-sparse Southern Ocean flux estimate, whereas the North Atlantic flux estimate proved to be more robust towards a measurement offset as a result of the denser existing measurement network⁵. Thus our findings indicate that the quantity of the data has a greater influence on accuracy than the data quality in densely observed ocean areas.

We show the importance of cross-calibration and having a fleet simultaneously measuring pCO₂ while closely sailing together. In this study, we detect any changes in the air-sea CO₂ flux and attribute them to the integration of sailboat pCO₂ observations. Although we now know the difference induced by the Seaexplorer data, a conclusive answer to whether and to how much the Seaexplorer data reduce the overall present-day uncertainty in the air-sea CO₂ flux is still missing and needs to be core to another study. Considering the recurrence of sailboat races, they have the potential to improve reconstructive air-sea CO₂ flux estimates on longer timescales in the future.

Surface-ocean carbon dioxide data. Two sea surface carbon dioxide datasets were used in this study: 1) pCO₂ measurements from underway shipboard and mooring data contained in the 1x1 degree gridded Surface Ocean CO₂ Atlas SOCATv2022 from 1982 to 2021⁵ and 2) underway pCO₂ measurements from the sailboat “Seaexplorer-Yacht Club de Monaco” (until 2019 “Malizia”) during offshore sailing and training events from 2018 to 2021. The former dataset includes the latter data as well, hence we artificially create a third dataset, where we exclude the Seaexplorer measurements from the SOCAT gridded dataset.

To quantify the changes in air-sea CO₂ fluxes based on the addition of Seaexplorer data, we used these 2 datasets as starting points, i.e. (1) SOCATv2022 including Seaexplorer data (E1), and (2) SOCATv2022 excluding Seaexplorer data (E2). To assess the impact of a potential measurement accuracy of ± 5 µatm we created two more datasets by (3) adding a random uncertainty of ± 5 µatm to the Seaexplorer data (E3) (similar to differences observed during measurement campaigns^42,43 where multiple pCO₂ systems were compared to the membrane system used on sailboats) and by (4) adding a constant measurement offset of 5 µatm to the Seaexplorer data (E4), which is in theory possible but less likely considering the prescribed daily two-point calibration. For a more detailed description of the measurements and the instrument setup and justification of the 5 µatm offset, we refer to²⁷.

pCO ₂ Mapping and Air-Sea CO₂ Flux Calculation. Mapped estimates of the sea surface pCO₂ were created by applying the SOM-FFN method to all four datasets. Here we provide a brief overview of the method, whereas a more detailed description including evaluation can be found in^29,44.

In the first step, a self-organizing map (SOM) clusters the ocean into 16 biogeochemical provinces based on common patterns in predictor variables. We used sea-surface temperature data⁴⁵, sea-surface salinity data⁴⁶, a mixed layer depth climatology⁴⁷, and a pCO₂ climatology⁴⁸ as predictors. In the second step, a feed-forward neural network (FFN) establishes non-linear relationships between the predictors and pCO₂ observations within each province separately. It uses these relationships to reconstruct the missing pCO₂ values within each province. The predictors for the FFN were SST, SSS, the MLD climatology as well as Chlorophyll-a (http://www.globcolour.info), and the atmospheric CO₂ concentration⁴⁹. We used a monthly climatology from 1998 to 2002 for chlorophyll-a prior to 1997, as chlorophyll-a is only available after the launch of satellites in 1997. The data for the FFN are divided into a training dataset to train on and a validation dataset used for validation within the method²⁹.

From the four reconstructed pCO₂ maps described above we estimate the air-sea CO₂ flux based on a bulk gas transfer formulation with a quadratic relationship between windspeed and transfer velocity^29,50 where we scale the mean gas transfer to a global average rate of 16.5 cm hr⁻¹ ⁵¹. We calculate the difference between the flux estimate based on SOCAT including Seaexplorer data and the flux estimate based on SOCAT excluding Seaexplorer data to quantify the impact of Seaexplorer data on the air-sea CO₂ flux. We calculate the difference between the flux estimate based on the original Seaexplorer data and the flux estimates based on Seaexplorer data with the added measurement uncertainties to assess the impact of the expected measurement accuracy on the air-sea CO₂ flux.

Signal-To-Noise-Detection. To detect statistically significant differences and reduce the impact of random errors arising from methodological choices, we use a Monte Carlo approach to reconstruct and calculate each of the air-sea CO₂ flux estimates (i.e. with Seaexplorer data, without Seaexplorer data, random error and constant offset - see above) 40 times with a varying split between the training and validation dataset to create for ensembles, i.e. ensemble E1 = Seaexplorer data, E2 = Seaexplorer data excluded, E3 = random measurement uncertainty, and E4 = fixed measurement bias. We gradually increased the number of runs and based on trial and error we found that the standard deviation between runs is nearly constant for 40 runs (Supplementary Fig. 3). To ensure the statistical significance of our results, we conducted a two-sample t-test and adjusted the resulting p-values in order to control the False Discovery Rate, i.e. the expected proportion of false discoveries among all significant results, to 5%³¹. The signal corresponds to adjusted p-values below 5%, indicating significance, while non-significant differences represent noise (Supplementary Fig. 4). The noise level is highest at the beginning of the time series as SOCAT contains few observations before 1990⁵, whereas the signal increases after 2016 as new Seaexplorer data made a difference (Supplementary Fig. 4).

Regional Focus. Finally, we set our focus on two main regions of interest, i.e. the North Atlantic, where most sailboat races took place, and on the Southern Ocean, where the longest race, the Antarctic circumnavigation race, took place. Furthermore, ⁵² has highlighted significant uncertainties in the air-sea CO2 flux in both regions. We focused on three regions in the Southern Ocean: the Polar Front, the Subantarctic Front, and the Northern Boundary according to ^51,53 including the respective areas within a 2-degree (or roughly 200 km) radius. Note that overlap between the frontal regions occurs. In the North Atlantic, we define the region as the area between 70°N, 0°, 85°W, and 20°E. The extent of the Southern Ocean is defined by south of 35°S.

To determine the data availability per region, we calculate the percentage of 1x1° pixels that were filled with Seaexplorer data at least once, regardless of the monthly availability.

Acknowledgments

We would like to thank Team Malizia and in particular skipper Boris Hermann for their engagement in collecting valuable data.

The Surface Ocean CO₂ Atlas (SOCAT) is an international effort, endorsed by the International Ocean Carbon Coordination Project (IOCCP), the Surface Ocean Lower Atmosphere Study (SOLAS) and the Integrated Marine Biosphere Research (IMBeR) program, to deliver a uniformly quality-controlled surface ocean CO₂ database. The many researchers and funding agencies responsible for the collection of data and quality control are thanked for their contributions to SOCAT.

The VLIZ ICOS work is supported by Research Foundation Flanders (FWO) contract I001821N. VLIZ ICOS data and metadata are available on the ICOS Carbon Portal (https://www.icos-cp.eu/).
TT acknowledges German federal and state funding for developing the innovation platform Shaping an Ocean of Possibilities, SOOP. JB gratefully acknowledges funding from the International Max Planck Research School on Earth System Modelling (IMPRS-ESM).

Data Availability. All data used and discussed in this article are freely available via www.socat.info. The datasets generated and/or analysed during the current study are available in the Zenodo repository, 10.5281/zenodo.10036579.

Authors’ Contributions. PL collected and quality-controlled the Seaexplorer data. JB completed all data analysis. JB and PL designed the study. JB wrote the manuscript together with PL and TT. All authors have discussed and contributed to the writing of the manuscript at all stages.

Competing Interests. The author(s) declare that they have no competing interests.

Friedlingstein, P. et al. Global Carbon Budget 2022. Earth Syst. Sci. Data 14, 4811–4900 (2022).
Gruber, N. et al. Trends and variability in the ocean carbon sink. Nat. Rev. Earth Environ. 4, 119–134 (2023).
DeVries, T. et al. Magnitude, trends, and variability of the global ocean carbon sink from 1985-2018. Glob. Biogeochem. Cycles n/a, e2023GB007780 (2023).
Le Quéré, C. et al. Saturation of the Southern Ocean CO2 Sink Due to Recent Climate Change. Science 316, 1735–1738 (2007).
Bakker, D. C. E. et al. A multi-decade record of high-quality fCO₂ data in version 3 of the Surface Ocean CO₂ Atlas (SOCAT). Earth Syst. Sci. Data 8, 383–413 (2016).
Pierrot, D. et al. Recommendations for autonomous underway pCO2 measuring systems and data-reduction routines. Deep Sea Res. Part II Top. Stud. Oceanogr. 56, 512–522 (2009).
Landschützer, P., Gruber, N. & Bakker, D. C. E. Decadal variations and trends of the global ocean carbon sink. Glob. Biogeochem. Cycles 30, 1396–1417 (2016).
Bennington, V., Gloege, L. & McKinley, G. A. Variability in the Global Ocean Carbon Sink From 1959 to 2020 by Correcting Models With Observations. Geophys. Res. Lett. 49, e2022GL098632 (2022).
Iida, Y., Takatani, Y., Kojima, A. & Ishii, M. Global trends of ocean CO2 sink and ocean acidification: an observation-based reconstruction of surface ocean inorganic carbon variables. J. Oceanogr. 77, 323–358 (2021).
Rödenbeck, C., DeVries, T., Hauck, J., Le Quéré, C. & Keeling, R. F. Data-based estimates of interannual sea–air CO₂ flux variations 1957–2020 and their relation to environmental drivers. Biogeosciences 19, 2627–2652 (2022).
Gregor, L. & Gruber, N. OceanSODA-ETHZ: a global gridded data set of the surface ocean carbonate system for seasonal to decadal studies of ocean acidification. Earth Syst. Sci. Data 13, 777–808 (2021).
Chau, T. T. T., Gehlen, M. & Chevallier, F. A seamless ensemble-based reconstruction of surface ocean pCO₂ and air–sea CO₂ fluxes over the global coastal and open oceans. Biogeosciences 19, 1087–1109 (2022).
Jiang, Z.-P., Yuan, J., Hartman, S. E. & Fan, W. Enhancing the observing capacity for the surface ocean by the use of Volunteer Observing Ship. Acta Oceanol. Sin. 38, 114–120 (2019).
Pfeil, B. et al. A uniform, quality controlled Surface Ocean CO₂ Atlas (SOCAT). Earth Syst. Sci. Data 5, 125–143 (2013).
Sabine, C. L. et al. Surface Ocean CO₂ Atlas (SOCAT) gridded data products. Earth Syst. Sci. Data 5, 145–153 (2013).
Gloege, L. et al. Quantifying Errors in Observationally Based Estimates of Ocean Carbon Sink Variability. Glob. Biogeochem. Cycles 35, e2020GB006788 (2021).
Hauck, J. et al. Sparse observations induce large biases in estimates of the global ocean CO2 sink: an ocean model subsampling experiment. Philos. Trans. R. Soc. Math. Phys. Eng. Sci. 381, 20220063 (2023).
McKinley, G. A., Fay, A. R., Eddebbar, Y. A., Gloege, L. & Lovenduski, N. S. External Forcing Explains Recent Decadal Variability of the Ocean Carbon Sink. AGU Adv. 1, e2019AV000149 (2020).
DeVries, T., Holzer, M. & Primeau, F. Recent increase in oceanic carbon uptake driven by weaker upper-ocean overturning. Nature 542, 215–218 (2017).
Mongwe, N. P., Vichi, M. & Monteiro, P. M. S. The seasonal cycle of pCO₂ and CO₂ fluxes in the Southern Ocean: diagnosing anomalies in CMIP5 Earth system models. Biogeosciences 15, 2851–2872 (2018).
Frölicher, T. L. et al. Dominance of the Southern Ocean in Anthropogenic Carbon and Heat Uptake in CMIP5 Models. J. Clim. 28, 862–886 (2015).
Bushinsky, S. M. et al. Reassessing Southern Ocean Air-Sea CO2 Flux Estimates With the Addition of Biogeochemical Float Observations. Glob. Biogeochem. Cycles 33, 1370–1388 (2019).
Gray, A. R. et al. Autonomous Biogeochemical Floats Detect Significant Carbon Dioxide Outgassing in the High-Latitude Southern Ocean. Geophys. Res. Lett. 45, 9049–9057 (2018).
Williams, N. L. et al. Calculating surface ocean pCO2 from biogeochemical Argo floats equipped with pH: An uncertainty analysis. Glob. Biogeochem. Cycles 31, 591–604 (2017).
Sutton, A. J., Williams, N. L. & Tilbrook, B. Constraining Southern Ocean CO2 Flux Uncertainty Using Uncrewed Surface Vehicle Observations. Geophys. Res. Lett. 48, e2020GL091748 (2021).
Heimdal, T. H., McKinley, G. A., Sutton, A. J., Fay, A. R. & Gloege, L. Assessing improvements in global ocean pCO₂ machine learning reconstructions with Southern Ocean autonomous sampling. Biogeosciences Discuss. 2023, 1–35 (2023).
Landschützer, P., Tanhua, T., Behncke, J. & Keppler, L. Sailing through the southern seas of air–sea CO2 flux uncertainty. Philos. Trans. R. Soc. Math. Phys. Eng. Sci. 381, 20220064 (2023).
Lauvset S. et al. SOCAT Quality Control Cookbook. (2018).
Landschützer, P. et al. A neural network-based estimate of the seasonal to inter-annual variability of the Atlantic Ocean carbon sink. Biogeosciences 10, 7793–7815 (2013).
Canadell, J.G., P.M.S. Monteiro, M.H. Costa, L. Cotrim da Cunha, P.M. Cox, A.V. Eliseev, S. Henson, M. Ishii, S. Jaccard, C. Koven, A. Lohila, P.K. Patra, S. Piao, J. Rogelj, S. Syampungani, S. Zaehle, and K. Zickfeld. Global Carbon and other Biogeochemical Cycles and Feedbacks. In Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 673–816, doi: 10.1017/9781009157896.007 (2021).
Storey, J. D. A direct approach to false discovery rates. J. R. Stat. Soc. Ser. B Stat. Methodol. 64, 479–498 (2002).
Chapman, C., Lea, M.-A., Meyer, A., Sallée, J. & Hindell, M. Defining Southern Ocean fronts and their influence on biological and physical processes in a changing climate. Nat. Clim. Change 10, 1–11 (2020).
Johnson, K. S. et al. Biogeochemical sensor performance in the SOCCOM profiling float array. J. Geophys. Res. Oceans 122, 6416–6436 (2017).
Meinig, C. et al. Public–Private Partnerships to Advance Regional Ocean-Observing Capabilities: A Saildrone and NOAA-PMEL Case Study and Future Considerations to Expand to Global Scale Observing. Front. Mar. Sci. 6, (2019).
United Nations Framework Convention on Climate Change. Paris Agreement. (2015). Retrieved from https://unfccc.int/sites/default/files/resource/parisagreement_publication.pdf
Lebling, K., E. Northrop, C. McCormick, and E Bridgwater. Towards Responsible and Informed Ocean-Based Carbon Dioxide Removal: Research and Governance Priorities. Report. Washington, DC: World Resources Institute. World Resources Institute (2022).
Wanninkhof, R. et al. A Surface Ocean CO2 Reference Network, SOCONET and Associated Marine Boundary Layer CO2 Measurements. Front. Mar. Sci. 6, (2019).
Rödenbeck, C. et al. Data-based estimates of the ocean carbon sink variability – first results of the Surface Ocean pCO₂ Mapping intercomparison (SOCOM). Biogeosciences 12, 7251–7278 (2015).
DeVries, T. et al. Decadal trends in the ocean carbon sink. Proc. Natl. Acad. Sci. 116, 11646–11651 (2019).
Ritter, R. et al. Observation-based trends of the southern ocean carbon sink. Geophys. Res. Lett. 44, 12,339-12,348 (2017).
Djeutchouang, L. M., Chang, N., Gregor, L., Vichi, M. & Monteiro, P. M. S. The sensitivity of pCO_2 reconstructions to sampling scales across a Southern Ocean sub-domain: a semi-idealized ocean sampling simulation approach. Biogeosciences 19, 4171–4195 (2022).
Arruda, R., Atamanchuk, D., Cronin, M., Steinhoff, T. & Wallace, D. At-sea intercomparison of three underway p CO 2 systems: Intercomparison of pCO 2 systems. Limnol. Oceanogr. Methods 18, (2019).
Olivier, L. et al. Wintertime process study of the North Brazil Current rings reveals the region as a larger sink for CO₂ than expected. Biogeosciences 19, 2969–2988 (2022).
Landschützer, P., Gruber, N., Bakker, D. C. E. & Schuster, U. Recent variability of the global ocean carbon sink. Glob. Biogeochem. Cycles 28, 927–949 (2014).
Reynolds, R. W. et al. Daily High-Resolution-Blended Analyses for Sea Surface Temperature. J. Clim. 20, 5473–5496 (2007).
Good, S. A., Martin, M. J. & Rayner, N. A. EN4: Quality controlled ocean temperature and salinity profiles and monthly objective analyses with uncertainty estimates. J. Geophys. Res. Oceans 118, 6704–6716 (2013).
de Boyer Montégut, C., Madec, G., Fischer, A. S., Lazar, A. & Iudicone, D. Mixed layer depth over the global ocean: An examination of profile data and a profile-based climatology. J. Geophys. Res. Oceans 109, (2004).
Takahashi, T. et al. Climatological mean and decadal change in surface ocean pCO2, and net sea–air CO2 flux over the global oceans. Deep Sea Res. Part II Top. Stud. Oceanogr. 56, 554–577 (2009).
Dlugokencky, E. J., Thoning, K. W., Lan, X. & Tans, P. P. NOAA Greenhouse Gas Reference from Atmospheric Carbon Dioxide Dry Air Mole Fractions from the NOAA GML Carbon Cycle Cooperative Global Air Sampling Network. (2021).
Wanninkhof, R. Relationship between wind speed and gas exchange over the ocean. J. Geophys. Res. Oceans 97, 7373–7382 (1992).
Naegler, T. Reconciliation of excess 14C-constrained global CO2 piston velocity estimates. Tellus B 61, 372–384 (2009).
Rustogi, P., Landschützer, P., Brune, S. & Baehr, J. The impact of seasonality on the annual air-sea carbon flux and its interannual variability. Npj Clim. Atmospheric Sci. 6, 66 (2023).
Park, Y.-H. et al. Observations of the Antarctic Circumpolar Current Over the Udintsev Fracture Zone, the Narrowest Choke Point in the Southern Ocean. J. Geophys. Res. Oceans 124, 4511–4528 (2019).

No competing interests reported.

SupplementaryBehncke.docx

Download PDF

Journal Publication

published 09 Feb, 2024

Read the published version in Scientific Reports →

Editorial decision: Revision requested
09 Jan, 2024
Reviewers agreed at journal
11 Dec, 2023
Reviews received at journal
09 Dec, 2023
Reviewers agreed at journal
05 Dec, 2023
Reviewers agreed at journal
30 Nov, 2023
Reviewers invited by journal
08 Nov, 2023
Editor assigned by journal
01 Nov, 2023
Editor invited by journal
01 Nov, 2023
Submission checks completed at journal
01 Nov, 2023
First submitted to journal
20 Oct, 2023

You are reading this latest preprint version

A detectable change in the air-sea CO2 flux estimate from sailboat measurements

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

Discussion

Methods and Material

Declarations

Acknowledgments

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1