2.1. NEMO-PISCES ocean biogeochemical model

Historical ocean-only simulations were performed using version 3.6 of the Nucleus for European Modelling of the Ocean (NEMO). This comprehensive framework includes version 3 of the Louvain-La-Neuve Sea Ice Model (LIM) [Fichefet and Maqueda, 1997] and version 2 of the Pelagic Interaction Scheme for Carbon and Ecosystem Studies (PISCES) biogeochemical model [Aumont et al., 2015]. The PISCES ocean biogeochemical model [Aumont et al., 2015] includes key elements for the assessment of air–sea carbon fluxes and oceanic sequestration of anthropogenic carbon. This includes carbon in various non-living forms (dissolved inorganic carbon, dissolved organic carbon, particulate inorganic carbon (calcite), and particulate organic carbon) and living compartments (nanophytoplankton, diatoms, microzooplankton, and mesozooplankton). It also includes total alkalinity and essential marine macro/micro nutrients (nitrate, ammonium, phosphate, silicate and iron). No vegetated coastal ecosystems, such as blue carbon ecosystems, are explicitly represented in the model. Air–sea CO₂ fluxes follow Ocean Model Intercomparison Project protocols [Orr et al., 2017], with gas exchange dependent on the air–sea partial pressure gradient and a parameterization of the instantaneous gas transfer velocity that depends on 10 m atmospheric wind speed [Wanninkhof, 1992, Ho et al., 2006]. We implemented the widely used “eddy-permitting” (0.25° horizontal grid) global ocean configuration eORCA025, which contains 75 vertical levels, including 7 levels in the top 10 m. The same model configuration has previously been used to assess the efficacy of marine carbon dioxide removal (mCDR) interventions, including macroalgae afforestation [Berger et al., 2023] and ocean alkalinity enhancement [Kwiatkowski et al., 2023] and has been shown to simulate total and anthropogenic ocean carbon fluxes in broad agreement with data-based estimates [Kwiatkowski et al., 2023]. The simulated estimate of the global oceanic carbon sink for 2011–2020 is 9.59 GtCO₂⋅yr⁻¹, consistent with the multi-model global carbon budget estimate of 10.3 GtCO₂⋅yr⁻¹ (9–12 GtCO₂⋅yr⁻¹) [Friedlingstein et al., 2023].

2.2. Simulations

Simulations were initiated in 1958, based on data-derived climatologies and a coarser resolution NEMO-PISCES historical simulation of the anthropogenic ocean carbon stock [Terhaar et al., 2019]. The model was then run from 1958 to 2015 using atmospheric forcings provided by the Drakkar forcing set 5.2 (DFS5) [Dussin et al., 2016], based on the ERA-Interim reanalysis over 1979–2015. A historical simulation was forced with yearly averaged historical atmospheric CO₂ concentrations reconstructed from ice cores and atmospheric records [Le Quéré et al., 2015]. A parallel preindustrial control run (PI control) was also performed, in which the atmospheric forcings were identical but the atmospheric CO₂ mole fraction was held constant at the preindustrial level, defined as 287 ppm in 1870. Total air–sea CO₂ fluxes were derived from the historical simulation while anthropogenic fluxes were calculated as the difference between fluxes in the historical simulation and the parallel PI control.

An additional set of simulations (historical and PI control) was performed, incorporating atmospheric forcings from JRA55-do v1.5 [Tsujino et al., 2018] which cover the period from 1958 to near present. These simulations were run from 2006 to 2020, initiated from DFS5 simulations. The DFS5 and JRA55-do v1.5 forced simulations were finally merged in January 2014 to form two coherent historical and PI control simulations that cover the period 1958 to 2020.

2.3. Observations

Robust pCO₂ measurements are critical for estimating total carbon fluxes from observations. In this context, we used the latest version of the Surface Ocean CO₂ Atlas (SOCAT, https://www.socat.info), which contains about 43 million observations of surface pCO₂. We characterized the temporal and spatial coverage of observations in the French EEZ using the gridded product on a 1° × 1° grid. This was compared to EEZ boundaries projected on the same grid using the Python shapefile function.

EEZ boundaries, including the French territories, were defined using the Maritime Boundaries Geodatabase, version 11 (Flanders Marine Institute (2019), https://www.marineregions.org). This analysis included certain disputed areas, including those contested by New Caledonia and Vanuatu, and by Tromelin and Mauritius.

3.1. The anthropogenic open ocean carbon sink of the French EEZ

The simulated anthropogenic ocean carbon sink in the French EEZ over the period 2011–2020 was 0.21 GtCO₂⋅yr⁻¹ (Figure 1). This is 7.5% of the global EEZ anthropogenic carbon sink and 2% of the total global oceanic anthropogenic carbon sink. Due to extensive overseas territories, notably in the Pacific and Indian Oceans, French waters were the third largest EEZ carbon sink, after the United Kingdom and Australia, absorbing the equivalent of about 50% of the 418 million tonnes of CO_2,eq⋅yr⁻¹ emissions of France over 2011–2020 (territorial annual greenhouse gas emissions in 2011–2022) [Jones et al., 2023, Friedlingstein et al., 2023].

3.2. Variable regional contributions to the French ocean carbon sink

The model simulations indicate significant regional disparities in the French EEZ anthropogenic carbon sink. Southern Ocean territories exhibited the highest fluxes per unit area, with mean anthropogenic carbon fluxes ranging from 20 to 60 gCO₂⋅m⁻²⋅yr⁻¹ (Figure 2), up to 160 gCO₂⋅m⁻²⋅yr⁻¹ in winter. Notably, the Atlantic and Mediterranean coastal areas of mainland France also demonstrated relatively high area-specific anthropogenic carbon fluxes, ranging from 10 to 50 gCO₂⋅m⁻²⋅yr⁻¹, reaching 100 gCO₂⋅m⁻²⋅yr⁻¹ in winter. In contrast, the tropical and subtropical regions of the French EEZ displayed lower area-specific anthropogenic carbon fluxes, ranging from 5 to 30 gCO₂⋅m⁻²⋅yr⁻¹. Despite lower area-specific fluxes, the areal extent of some of these tropical regions, notably in the Pacific territories, means that they are major drivers of the anthropogenic carbon sink in the French EEZ. For instance, the total uptake for French Polynesia and New Caledonia in 2020 was 83.7 MtCO₂⋅yr⁻¹ and 31.8 MtCO₂⋅yr⁻¹, respectively. The combined anthropogenic carbon sequestration capacity of these two regions exceeded that of all other French territories combined.

3.3. Historical trends in the French ocean carbon sink

Our simulations show that over the last 40 years the French EEZ sink increased from 140 MtCO₂⋅yr⁻¹ to about 225 MtCO₂⋅yr⁻¹ (Figure 3a). The largest increases occurred in French Polynesia, New Caledonia, the Southern Ocean Islands and Clipperton, with the increase near proportional to the initial sink (R² = 0.96). The observed increase of the French EEZ carbon sink corresponds to that of the global EEZ carbon sink, indicative of a response to the atmospheric carbon dioxide growth rate (Figure 3a). However, our simulations revealed a decline in the growth of the French carbon sink since 2013 with a particularly low carbon sink in 2016 observed in 7 out of 21 territories. This slowdown is distinct from a slowdown of the global EEZ carbon sink (Figure 3b). As a result, the French EEZ share of the global EEZ carbon sink declined from 7.3% to 6.4%, with French Polynesia, New Caledonia, and Amsterdam and St Paul Islands territories being the main contributors to this decline.

3.4. Disentangling anthropogenic and total ocean carbon fluxes and stocks

Analysis of the anthropogenic carbon sink using the French EEZ territories as a case study, highlights a number of features that are consistent across other EEZs.

3.4.1. Net air–sea fluxes of anthropogenic carbon into an EEZ (or EEZ region) are not indicative of the change in anthropogenic carbon stocks within that EEZ

For all of the French EEZ territories, there is an imbalance between the net air–sea anthropogenic carbon flux and the change in the ocean anthropogenic carbon stock (Figure 4a). Some EEZ territories are net exporters of anthropogenic carbon to other EEZs and the high seas. For example, roughly 25% (3 gC⋅m⁻²⋅yr⁻¹) of the net-air sea anthropogenic carbon flux in Kerguelen is not retained in its waters. While in Guadeloupe, the increase in the integrated anthropogenic carbon stock is approximately 4 times what it receives in net air–sea fluxes, with 12 gC⋅m⁻²⋅yr⁻¹ supplied by waters that are transported into the Guadeloupe region. Ocean dynamics, therefore, lead to significant disparities between the carbon stocks within EEZ waters and the anthropogenic carbon uptake from the atmosphere.

3.5. Observational coverage

The NEMO-PISCES model permits the calculation of anthropogenic carbon fluxes in space and time, providing insight into carbon flux trends and the mismatch between carbon fluxes and carbon stocks. However, given the persistence of large discrepancies between data and models [Friedlingstein et al., 2023], robust pCO₂ measurements are essential for estimating total carbon fluxes from observations. Therefore, good observational coverage is crucial, as under-sampling could lead to significant biases in fluxes reconstructed from observations [Hauck et al., 2023b]. A review of the available SOCATv2023 pCO₂ measurements in French EEZ waters, showed that observational coverage remains limited (Figure 5). Compiled measurements for 2011–2021 are patchy in several French territories. This is particularly striking in French Polynesia where spatial coverage is notably low. Large portions of the territory either have no observations at all or only one month of observations. Specifically, on a 1° × 1° grid, only 14% of the French Polynesia EEZ grid cells have at least one observation, and a mere 1% has at least 2 months of observations available. The southern Indian Ocean also has limited data, with 35% of Crozet, 73% of Amsterdam and Saint Paul and 83% of Kerguelen Island covered by at least one observation. However, this coverage drops to 0%, 5.5% and 7% for those three islands, respectively, when the criterion is at least 4 months of observations per year. This lack of data is particularly problematic as this region plays an important role in carbon uptake, but only a fraction of the seasonal cycle is present in the observations. On the contrary, the Atlantic territories have good spatial and temporal coverage, with 75% of French Guiana territory covered with at least 6 months of data and 66% for both mainland France and Martinique.

4.1. The French EEZ case study

Our study reveals a significant anthropogenic carbon sink in the world’s EEZs, accounting for 3.55 GtCO₂⋅yr⁻¹ or 35% of the global ocean CO₂ sink in 2020. The anthropogenic carbon sink in EEZs can be a similar order of magnitude to national emissions: for example, the French EEZ sink corresponds to around 50% of French CO_2,eq territorial emissions, while the two largest EEZ sinks, Australia and the UK, both absorb the equivalent of around 40% of their emissions. Even more absurd, the waters of 34 countries absorb more than their own emissions, up to 1640 times more for Tuvalu. We estimate an increase in the anthropogenic carbon sink within the French EEZ over the last 40 years, from 140 MtCO₂⋅yr⁻¹ in 1980 to about 225 MtCO₂⋅yr⁻¹ in 2020. This is driven by the increase in the global ocean carbon sink in response to rising atmospheric CO₂ levels [IPCC, 2021, Friedlingstein et al., 2023]. However, we have identified a recent decrease in the French EEZ that is not reflected in the global EEZ carbon sink. This decrease may be attributable to changes in atmospheric forcing in our simulations that affect ocean dynamics and redistribute air–sea fluxes.

Using the French EEZ as a case study, we highlight the considerable diversity among EEZ territories in terms of total air–sea carbon fluxes, anthropogenic air–sea carbon fluxes and changing oceanic C_ant stocks. French EEZ territories exhibit both total uptake of carbon from the atmosphere (ingassing) and total loss of carbon to the atmosphere (outgassing). Yet despite this, over the simulated period, all EEZ territories are net absorbers of anthropogenic carbon, albeit with the magnitude of the area-specific air–sea C_ant flux differing by a factor of 6 between territories. Moreover, territories show wide disparities in the accumulation of C_ant in their waters, and this accumulation is not equivalent to the cumulative air–sea C_ant flux. This disparity is a consequence of the advection and mixing of C_ant within the ocean, which results in certain EEZ territories being net exporters of C_ant to adjacent waters, while others are net importers. Despite the temptation to incorporate ocean carbon fluxes into NDCs due to the magnitude of associated anthropogenic carbon sinks, the drivers and properties of ocean carbon fluxes, as well as the wider context of the remaining carbon budget, make this highly inadvisable.

4.2. Pitfalls of including EEZ carbon fluxes in NDCs

One of the principal issues with including EEZs in NDCs, is that ocean anthropogenic carbon uptake is already accounted for when estimating the remaining carbon budget [Lamboll et al., 2023, Rogelj et al., 2016, Matthews et al., 2021]. Consequently, the only rationale for including EEZs in NDCs lies in demonstrating that human interventions have enhanced the anthropogenic carbon sink. The transient climate warming response to cumulative CO₂ emissions (TCRE) is central to remaining carbon budget estimates and can be derived from both models and observations using a variety of methodologies. TCRE estimates invariably account for unmanaged ocean anthropogenic carbon uptake, including that which is projected to occur in EEZs. As such, our best estimates of permissible emissions to stay within 1.5 °C or 2 °C global warming targets already assume that ocean anthropogenic carbon uptake in EEZs will continue in response to biogeophysical forcing. Including total anthropogenic fluxes, as opposed to any managed additional flux, into NDCs would therefore break the coherence between the remaining carbon budget and the global stocktake.

Moreover, for consistency, any management practices found to have diminished EEZ C_ant uptake, would equally need to be declared [Bach et al., 2021]. The concept of additionality requires a standardized definition applicable across all contexts. For instance, notable disparities exist in estimating anthropogenic land-use carbon fluxes between the model-based Global Carbon Budget (GCB) and the inventories submitted by countries to the United Nations Framework Convention on Climate Change (UNFCCC). The former includes only direct anthropogenic effects, while the inventories reported by countries generally include indirect removals on that land, such as the fertilization effect caused by increasing atmospheric CO₂ concentrations [Grassi et al., 2023, Gidden et al., 2023]. Similarly, the anthropogenic ocean carbon sink operates independently of human management, primarily in response to increased atmospheric CO₂. Although ocean C_ant uptake is nearly ubiquitous, the magnitude of this flux at the air–sea interface is highly regionally variable (Figure 1) and strongly determined by ocean dynamics and carbonate chemistry. In particular, cold deep-water formation regions such as the North Atlantic and Southern Ocean are major hotspots of anthropogenic carbon uptake (Figure 1) as water masses containing C_ant are subducted. As such, both historical and future C_ant uptake in an EEZ is almost entirely determined by the natural ocean dynamics and carbonate chemistry of the waters, and the impact of climate change on these properties (in conjunction with global CO₂ emissions). Crucially, these oceanic properties are almost entirely independent of human intervention and presently beyond the control of any nation state.

Furthermore, integrating EEZ carbon fluxes into NDCs presents a challenge to fairness and equity, and may raise climate justice concerns. C_ant uptake at the air–sea interface is higher in high latitude regions, such as the Southern Ocean and the North Atlantic (Figure 1c), coincident with the EEZs of a small number of developed countries that have historically high emissions. The potential for human enhancement of C_ant uptake may additionally be greater in these regions due to generally lower sea surface temperatures (i.e., higher CO₂ solubility, [Berger et al., 2023, Feng et al., 2017, He and Tyka, 2023]. Conversely, interior ocean C_ant storage is higher in subtropical regions [Gruber et al., 2023], coincident with the EEZs of developing countries that have historically low emissions. The question of which nation could rightfully claim credit for Ocean C_ant would arise because of large differences between where uptake initially occurs and where C_ant is subsequently stored. The functioning of the ocean is global and cannot be divided into administrative units.

The transboundary nature of oceanic carbon also presents a heightened risk of potential double counting. Many EEZs are located near multiple borders; EEZ territories may be net importers or exporters of C_ant to adjacent waters, regardless of whether they ingas or outgas total carbon (Figure 4). This situation exacerbates the risks of potential double counting [Luisetti et al., 2020]. Moreover, interventions in the waters of one EEZ can cause harm in another, particularly in the context of carbon dioxide removal (CDR) projects [Tagliabue et al., 2023, Boyd et al., 2022]. Thus, the emergence of ocean CDR poses novel risks that require specific policies and legislation [Boettcher et al., 2021]. The development of such a comprehensive legal framework is imperative, but is challenging and necessitates an arduous process due to the ocean’s status as a shared resource requiring extensive international cooperation. Although existing international treaties, such as the London Protocol, have addressed transboundary environmental risks, focusing on ocean fertilization [Dixon et al., 2014], there remains a notable absence of a comprehensive legal framework dedicated to marine CDR research [National Academies of Sciences, Engineering, and Medicine, 2022]. However various legal norms exist as well as codes of conduct to help guide responsible decisions in marine CDR [Boettcher et al., 2023].

4.3. Enhancing monitoring and estimation of oceanic carbon sinks: imperatives for climate mitigation and reporting

Although the inclusion of EEZs in NDCs is misguided, an accurate estimate of the EEZ carbon sink is highly desirable given its significant contribution to the global ocean carbon sink (up to 35% in our simulations). Enhanced monitoring of EEZ carbon fluxes is essential to track long-term trends and variability, as the ocean carbon sink is vulnerable to shifts caused by both direct and indirect anthropogenic drivers. Climate change exerts a diverse and multifaceted influence on ocean carbon fluxes and storage, via temperature increases, reduced buffering capacity, and changes in circulation and marine productivity. In addition, anthropogenic impacts on runoff and river inputs [Regnier et al., 2022], bottom trawling [Atwood et al., 2024, Sala et al., 2021], coastal development, and deep-sea mining can also have substantial impacts on carbon fluxes and stocks. Predicting future changes in carbon fluxes and storage is therefore a considerable challenge, requiring the establishment of an efficient observational network to accurately constrain these dynamics. Moreover, monitoring the changes and variability is critical to delivering accurate carbon budgets to the parties to the UNFCCC, and to providing a comprehensive global understanding of carbon stocks and fluxes. This understanding assists parties in formulating mitigation targets consistent with the temperature goals of the Paris Agreement. Similar concerns have been raised for terrestrial ecosystems, highlighting the need for improved monitoring of natural and anthropogenic carbon fluxes [Gidden et al., 2023, Nabuurs et al., 2023].

Recent improvements in observational capabilities and developments in machine-learning based interpolation methods have facilitated the generation of global gap-filled, high-resolution data products of oceanic pCO₂ and ocean air–sea CO₂ fluxes [Fay et al., 2021]. These developments have improved our understanding of long-term trends in coastal air–sea carbon fluxes [Roobaert et al., 2024] and enabled model-data combination to estimate global coastal fluxes [Resplandy et al., 2024]. However, despite this progress, the quantification of coastal CO₂ flux trends from observations remains severely constrained by limited spatial coverage and/or relatively short time series duration, while large discrepancies between data and models persist [Hauck et al., 2023a, Resplandy et al., 2024]. Additionally, the anthropogenic component of the air–sea carbon flux cannot be directly observed. While local estimates of DIC_ant are possible (e.g. though the ΔC^∗ method [Gruber et al., 1996] or the transient time distribution method [Hall et al., 2002], data-based methods can only estimate global anthropogenic carbon uptake. Furthermore, model estimates generally do not account for anthropogenic impacts unrelated to atmospheric CO₂ and climate change.

In the context of reporting on potential CDR projects, establishing a baseline is critical as it is the starting point against which emission reductions or additional sinks are measured. Accurate monitoring helps avoid errors in establishing this baseline and ensures that it reflects carbon fluxes and storage in the absence of human intervention [Ho et al., 2023]. Accurate monitoring of carbon fluxes will allow for better accounting of managed carbon fluxes and avoiding methodological errors in setting baseline emissions in reporting of carbon projects, within NDCs or elsewhere [Cooley et al., 2023]. In addition, a poorly defined baseline can lead to inaccurate accounting. This can be particularly detrimental if emission reductions are overestimated, creating a discrepancy between the remaining emissions allowed to meet climate change targets and the actual remaining emissions.