Data sources and spatial harmonization

We compiled updated datasets on blue carbon ecosystem extent, annual carbon sequestration, biological carbon pump sequestration, EEZ size, carbon pricing, and socioeconomic indicators to construct country-level Blue Carbon Wealth estimates. Priority was given to recent, internationally recognized, and spatially explicit sources in order to maximize comparability across countries and ensure consistency between ecological and macroeconomic components.

EEZ sizes

Land area and Exclusive Economic Zone (EEZ) area were obtained from the Flanders Marine Institute (VLIZ) spatial dataset, “Union of World Country Boundaries and EEZs, Version 4”, updated on 4 October 2024. This dataset combines global country boundaries from ESRI’s world country database with EEZ data from Version 12 and provides standardized spatial information in square kilometers (km2) for more than 300 countries and territories worldwide (Supplementary Data 2). It serves as the geographic backbone of the analysis and ensures consistent delineation of national maritime areas across all country-level estimates.

Coastal blue carbon: Area & annual sequestration

We updated blue ecosystem area sizes using recent UNEP-WCMC datasets (Supplementary Table 1), improving upon data reported by Bertram et al.3. Sources are detailed in Supplementary Table 1, including shapefiles for mangroves, salt marshes, and seagrasses. These shapefiles were reprojected into the Mollweide projection, as recommended by Bertram et al.3, to preserve area proportions and minimize distortions. Shapefile geometries were corrected for accuracy before performing an “intersect” operation with the EEZ shapefile, using the “UNION” field to assign each polygon to a specific country or territory. We applied a “dissolve” operation by the “UNION” field to merge overlapping polygons and avoid inflated area calculations caused by duplicates. Final area calculations were conducted for each country, ensuring representation of all territories and small islands.

For countries with incomplete or outdated spatial information, we supplemented the geospatial datasets with recent literature. In the Bahamas, seagrass meadow area was updated using Fu et al.48 and Gallagher et al.49, yielding a mean estimate of 79,757 km2, with a range from 66,990 to 92,524 km2. For Mauritania, salt marsh and seagrass areas were refined using estimates reported by Pottier et al.50.

To ensure consistency with the structure of the biological carbon pump dataset, mangrove, salt marsh, and seagrass areas for the Line Group, Gilbert Islands, and Phoenix Group were aggregated and reported under Kiribati as a single entity. This harmonization was necessary to maintain alignment between the BCE and BCP components of the framework.

We then estimated country-level annual carbon sequestration from BCEs by combining ecosystem areas with updated ecosystem-specific sequestration rates. Rates were converted into tonnes of organic carbon per square kilometer per year (tC km–2 yr–1), then multiplied by national ecosystem area to derive annual carbon uptakes. Updated sequestration parameters are reported in Supplementary Table 3, together with the values used by Bertram et al.3.

Country-level biological carbon pump annual sequestration

For the Biological Carbon Pump (BCP) component, we used the country-level annual sequestration estimates reported by Berzaghi et al.12. These estimates quantify the organic carbon sequestered annually via the ocean’s biological pump, driven primarily by phytoplankton production and the subsequent export of organic matter to deeper ocean layers. The authors allocated modeled global BCP fluxes to national EEZs, enabling a spatially explicit, country-by-country assessment of sequestration potential.

According to Berzaghi et al.12, the BCP sequesters 2.81 Gt C yr–1 globally (2.44–3.53), with an estimated storage time of at least 50 years ( ± 25 years). In the present study, however, we retain only the annual country-level sequestration values from their dataset. We do not use the economic valuation proposed in their article because our objective is to integrate the BCP as a physical sequestration component within a unified Blue Natural Capital framework. In contrast to BCE sequestration, the BCP term is treated as a point estimate in the current version of the framework.

Socioeconomic and environmental indicators by country

Annual CO2 emissions for 2023 were obtained from the Global Carbon Budget 202451, as processed by Our World in Data52 and originally compiled by the Global Carbon Project. These figures cover fossil-fuel and industrial CO2 emissions, excluding land-use change, for 246 countries. It ranges from 7667 to 12 billion tCO2.

Sovereign debt levels for 2023 were taken from the World Bank’s “External Debt Stocks, Total (DOD, current US$)” dataset. To ensure comparability with other monetary variables expressed in constant 2015 US dollars, debt values were converted from current US dollars using a factor (0.77) derived from the U.S. Bureau of Labor Statistics CPI Inflation Calculator. This conversion ensures consistency with GDP and the cost-of-carbon valuation basis.

GDP data were obtained from the World Bank’s “World Development Indicators” and expressed in constant 2015 US dollars. GDP values refer to 2023 whenever available; where 2023 data were missing, the closest available pre-2023 observation was retained. Population figures correspond to the 2023 estimates reported in the “UN World Population Prospects 2024” release. These estimates were selected rather than later projections to preserve consistency with the rest of the empirical dataset and to ensure that per-capita indicators reflect the most recent observed demographic conditions.

Country group classifications follow the UNCTADstat classification framework53, which categorizes nations as Small Island Developing States (SIDS), Least Developed Countries (LDCs), Developed Economies, or Developing Economies. Our classification approach prioritizes SIDS and LDC status to avoid double-counting when countries fall into multiple categories, resulting in 58 SIDS, 37 LDCs, 69 developed economies, and 94 developing economies.

Reference year and monetary basis

Unless otherwise stated, physical and socioeconomic variables refer to 2023 conditions, whereas all monetary values are expressed in constant 2015 US dollars. Accordingly, CO2 emissions, sovereign debt, and population are aligned to 2023, GDP refers to 2023 when available (or the closest earlier year otherwise), and the GSCC is expressed in 2015 US$ per tCO2. This convention ensures that recent observed quantities are combined with a common constant-dollar valuation basis, allowing consistent comparison between Blue Carbon Wealth and macroeconomic aggregates such as GDP and sovereign debt. The reference year, units, sources, and monetary basis used for each core variable are summarized in Table 2.

Table 2 Reference year, units, source, and monetary basis of core variables used in the analysisCountry-level BCE sequestration

Country-level BCE sequestration potential is estimated as the annual amount of carbon absorbed by coastal blue ecosystems within each country. We first compile the area covered by each ecosystem type, expressed in square kilometers (km2), and then multiply these areas by ecosystem-specific mean sequestration rates to obtain annual carbon uptake.

The average sequestration rates, expressed in tonnes of organic carbon per square kilometer per year (tC km–2 yr–1), are 168 ± 7 tC km–2 yr–1 for saltmarshes, 220.7 ± 20.1 tC km–2 yr–1 for seagrasses, and 183.6 ± 14.7 tC km–2 yr–1 for mangroves. For each ecosystem in each country, annual uptake is computed as:

$${{{{\rm{Carbon\,Uptake}}}}\,({{{\rm{tC}}}}/{{{\rm{yr}}}})}_{{{{\rm{c}}}},{{{\rm{e}}}}}={{{\rm{Ecosystem\,Area}}}}\,{\left({{{{\rm{km}}}}}^{2}\right)}_{{{{\rm{c}}}},{{{\rm{e}}}}}\times {{{{\rm{Sequestration\,rate}}}}\,({{{\rm{tC}}}}/{{{{\rm{km}}}}}^{2}/{{{\rm{yr}}}})}_{{{{\rm{e}}}}}$$

where, Ecosystem Area is the surface covered by that ecosystem type (mangrove, seagrass or saltmarshes) in the country and Sequestration Rate is the average annual carbon storage rate for that ecosystem type. This provides the annual blue carbon sequestration potential for each ecosystem, expressed in tonnes of organic carbon per year (tC yr–1). Total country-level sequestration is then obtained by summing across all BCEs present in the country.

$${{{\rm{Total}}}}\,{{{\rm{Carbon}}}}\,{{{\rm{Uptake}}}}\,({{{\rm{tC}}}}/{{{\rm{yr}}}})={\sum}_{{{{\rm{all}}}}\; {{{\rm{ecosystems}}}}}{{{\rm{Carbon}}}}\,{{{\rm{Uptake}}}}$$

We assume sequestration rates follow a lognormal distribution. This choice is justified by the strictly non-negative nature of burial and sequestration rates, the strong right-skew commonly reported in blue-carbon syntheses54, and IPCC uncertainty guidance for positive skewed variables55. In practice, full distributional parameters were not always available from source studies. We therefore approximated the lognormal distributions using reported means and medians where possible. For saltmarshes, no median was available, so the median was set equal to the mean as a pragmatic approximation. The resulting country-level sequestration distributions were then obtained by bootstrap propagation of uncertainty across ecosystem-specific sequestration rates. This assumption is conservative and unlikely to materially affect propagated uncertainty given the relatively small relative standard errors of the aggregated rates.

Using this framework, the mean sequestration potential per country is estimated at 0.619 ± 0.027 MtC yr–1 (million tC per year). Country-level values range from 0.271 tC yr–1 to 17.87 MtC yr–1, reflecting substantial variation in ecosystem coverage and sequestration capacity. This approach provides a transparent and robust basis for quantifying the contribution of BCEs to climate-change mitigation.

Cost of carbon

To update the Global Social Cost of Carbon (GSCC), we adopt a methodology that remains broadly consistent with Bertram et al.3, while adapting it to our Blue Carbon Wealth accounting framework. The GSCC measures the global economic damage associated with the emission of one additional tonne of CO2. As a starting point, we use the country-level social cost of carbon (CSCC) estimates reported by Ricke et al.37, using the corrected CSCC database associated with Ricke et al.56, which provides distributions of country-specific damages under a wide range of socio-economic, climate, and discounting assumptions. We do not re-estimate these values, but rely on the published corrected CSCC database.

Each scenario corresponds to a coherent set of assumptions combining a socio-economic pathway and an emissions trajectory, denoted by SSP and RCP labels. Scenarios also differ with respect to climate response specifications and discounting rules, including either fixed discount rates or Ramsey-type formulations with alternative preference and inequality parameters. For each scenario, we construct a scenario-specific GSCC by summing the median CSCC across countries, excluding any row that already corresponds to a pre-aggregated world estimate to avoid double counting:

$${{{\rm{GSC}}}}{{{{\rm{C}}}}}_{{{{\rm{s}}}},{{{\rm{q}}}}}={\sum}_{{{{\rm{i}}}}}{{{\rm{CSC}}}}{{{{\rm{C}}}}}_{{{{\rm{i}}}},{{{\rm{s}}}},{{{\rm{q}}}}},$$

where \(s\) denote the scenario index, \({{{\rm{CSC}}}}{{{{\rm{C}}}}}_{{{{\rm{i}}}}}\) the CSCC of country \(i\), and q the 50th percentile (median).

We then build a GSCC distribution across the full scenario space using a uniform bootstrap of 20,000 draws. All scenarios are given equal probability and this choice is intentional. It provides a transparent and neutral way to represent uncertainty, without privileging any particular model specification, socio-economic pathway, or discounting convention. The resulting GSCC values are expressed in 2015 US dollars per tonne of CO2.

As commonly found in the literature57,58, the raw GSCC distribution is strongly right-skewed (Supplementary Fig. 4a). Some scenarios generate moderate values, whereas others imply very high damages. This heterogeneity reflects both irreducible uncertainty and systematic differences across scenario families. In practice, GSCC values depend not only on stochastic variation, but also on the composition of scenario labels, such as discounting regime, SSP-RCP pairing, or model specification. This structured heterogeneity is visible across SSP × RCP combinations (Supplementary Fig. 4b). This is a critical issue for Blue Carbon Wealth accounting, where GSCC enters directly as a shadow price in wealth accounts, rendering wealth indicators highly sensitive to the chosen SCC level and trajectories3,59,60.

To reduce this dependence on scenario composition, we construct an adjusted GSCC distribution. The objective is not to eliminate scenario diversity, but to remove the systematic component associated with scenario labels while retaining the residual variation that captures non-structured uncertainty. To do so, we regress on a set of scenario-label indicators. Because labels are often correlated and numerous, we estimate this relationship using ridge regression rather than an unregularized model. Ridge penalization limits overfitting, stabilizes coefficients, and ensures that only the systematic and robust component linked to labels is removed (Supplementary Fig. 4c).

$$\log \left({{{\rm{GSCC}}}}_{{{{\rm{s}}}}}\right)={{{\rm{\alpha}}}}+\underbrace{{\sum}_{{{{\rm{k}}}}}{{{{\rm{\beta}}}}}_{{{{\rm{k}}}}}{{{\bf{1}}}}\left\{{{{{\rm{label}}}}}_{{{{\rm{k}}}}}\left({{{\rm{s}}}}\right)\right\}}_{{{{{\rm{m}}}}}_{{{{\rm{s}}}}}} {{{{\rm{\varepsilon}}}}}_{{{{\rm{s}}}}},$$

where \(\alpha\) captures the common baseline level shared across scenarios, \({{{\boldsymbol{1}}}}\left\{{{{{\rm{label}}}}}_{{{{\rm{k}}}}}\left({{{\rm{s}}}}\right)\right\}\) equals 1 if scenario s belongs to label group k (e.g., discounting regime, SSP–RCP family, specification block), \({\beta }_{k}\) measures the systematic contribution of label k to \(\log ({{{\rm{GSC}}}}{{{{\rm{C}}}}}_{{{{\rm{s}}}}})\), \({\varepsilon }_{s}\) the remaining non-structured variation across scenarios, and \({m}_{s}\) the label-driven component.

We then center the estimated label component so that its mean across scenarios is zero, and subtract this centered term from \(\log ({{{\rm{GSCC}}}})\). This procedure preserves the overall mean in log space while reducing the influence of label composition on the distribution. The adjusted GSCC is therefore interpretable as the common signal shared across the scenario ensemble, net of systematic specification-driven distortions.

$${\widetilde{{{{\rm{m}}}}}}_{{{{\rm{s}}}}}={{{{\rm{m}}}}}_{{{{\rm{s}}}}}{\mathbb{-}}{\mathbb{E}}[{{{\rm{m}}}}]$$

We define the adjusted GSCC distribution as:

$${{{\rm{log}}}} \left({{{\rm{GSCC}}}}_{{{\rm{s}}}}^{{{\rm{adj}}}}\right) = {{{\rm{log}}}} \left({{{\rm{GSCC}}}}_{{{\rm{s}}}}\right) – {\widetilde{{{\rm{m}}}}}_{{{\rm{s}}}} \iff {{{\rm{GSCC}}}}_{{{\rm{s}}}}^{{{\rm{adj}}}} = {{{\rm{exp}}}} \left(\right.{{{\rm{log}}}} ({{{\rm{GSCC}}}}_{{{\rm{s}}}}^{{{\rm{adj}}}})$$

We use this adjusted distribution as our preferred GSCC estimand for valuation and accounting. The adjusted GSCC is estimated at US$ 559.6 ± 1.90 per tCO2, with a standard deviation of 268.6, and a median of 506.7. Summary statistics and benchmark comparisons are reported in Table 3. Robustness checks indicate that this adjusted distribution is more stable than the observed one (Supplementary Fig. 4d), with both the median and upper-tail quantiles remaining stable around the selected regularization level.

Table 3 GSCC distribution comparison (all values in 2015 US$ / tCO2)

This approach improves transparency by making explicit the role of scenario assumptions and discounting choices. It avoids imposing a single arbitrary SCC value, preserves uncertainty, and yields a stable distribution suitable for accounting purposes. Accordingly, we use the adjusted GSCC distribution as the main cost-of-carbon input in the rest of the paper.

However, main limitations are that GSCC estimates remain sensitive to underlying CSCC assumptions, discounting choices, and scenario design. The adjustment reduces systematic label effects but does not remove deep uncertainty or model dependence. Finally, equal scenario weighting is transparent, yet normative, and may not reflect real-world likelihoods or policy relevance across futures.

BCW estimation

We frame Blue Carbon Wealth (BCW) around two carbon pathways. First, blue carbon ecosystems (mangroves, seagrasses, salt marshes) and the open-ocean biological carbon pump (BCP; phytoplankton) deliver an annual climate-regulation flow by removing CO2 from the atmosphere and exporting it to biomass and sediments (coasts) or to deeper waters (oceanic)13,14. Second, coastal vegetated ecosystems store large stocks of organic carbon whose disturbance (e.g., drainage, clearing, erosion) can trigger one-off pulses of CO2 emissions that conservation can avoid61. The oceanic BCP, by contrast, is almost entirely a flow service (short-lived standing stocks), so its value is not represented by a durable carbon stock12. Our estimates do not include carbon stored in surficial marine sediments of the continental shelf within EEZs, which several studies treat as part of national natural capital62,63.

A flow is a quantity per unit time (tonnes of carbon dioxide per year, tCO2 yr–1) that accrues each year and contributes to climate change. A stock is a quantity at a point in time that represents an asset; it does not recur annually, but if lost it can produce an emissions pulse over years to decades64. Because these are different objects of accounting, we keep services (flows) and assets (stocks) strictly separate to avoid double counting64. In line with that guidance, we report BCW as an annual, flow-based measure of climate-regulation services and do not fold stock values into that annual number. Stocks (and stock-at-risk) are treated in asset accounts and can be valued separately using the net present value approach for ecosystem assets64, consistent with the carbon budget framework that relates cumulative emissions to temperature targets65.

We treat BCW as an annual flow measure of avoided climate damages from that year’s CO2 removals. Specifically, for country i,

$${{{{\rm{BCW}}}}}_{{{{\rm{i}}}}}=\left(\underbrace{{\sum}_{{{{\rm{j}}}}}{{\mbox{A}}}_{{\mbox{i}},{\mbox{j}}}{{\mbox{s}}}_{{\mbox{j}}}}_{{\mbox{coastal\,blue}}-{\mbox{carbon\,sequestration}}}+\underbrace{{{\mbox{BCP}}}_{{\mbox{i}}}}_{{\mbox{biological\,pump}}\,\left({\mbox{oceanic}}\right)\,{\mbox{removal}}}\right)\times 44/12\times {{{\rm{GSCC}}}}$$

where is the mapped area of ecosystem (mangrove, seagrass and saltmarsh), its net annual organic carbon sequestration rate and the country-attributed annual carbon removal by the BCP. Removals are expressed in tonnes of CO2 yr–1 by applying the molecular weight ratio of CO2 to C (44/12), and are monetize using our estimated adjusted GSCC.

The BCE component is represented as a distribution because sequestration rates are modeled probabilistically and propagated through the country-level calculations by bootstrap. The GSCC is likewise represented as a distribution, reflecting uncertainty across scenario families and model specifications. By contrast, the BCP component is introduced as a point estimate, since only country-level annual sequestration values are retained from Berzaghi et al.12 and no uncertainty distribution is propagated for this term in the present framework. BCW therefore follows a hybrid uncertainty structure. Uncertainty is propagated from BCE sequestration and the GSCC, while BCP enters deterministically. As a result, reported BCW uncertainty should be interpreted as partial uncertainty, conditional on fixed BCP estimates.

Because countries differ greatly in population size, we compute per capita indicators to enable meaningful cross-country comparisons:

$${{{\rm{BCW}}}}\; {{{\rm{per}}}}\; {{{\rm{capita}}}}_{{{\rm{i}}}}=\frac{{{{{\rm{BCW}}}}}_{{{{\rm{i}}}}}}{{{{{\rm{Population}}}}}_{{{{\rm{i}}}}}}$$

BCW per capita provides the economic value of blue natural capital at the individual level. Per-capita wealth indicators are standard in wealth accounting and sustainability assessment; tracking wealth per person aligns comparisons across countries with different demographics and with the sustainability criterion of non-declining wealth per head66,67,68. This normalization highlights each nation’s per-person contribution to climate mitigation via blue natural capital.

Because blue natural capital is tightly linked to access to, and stewardship of, marine space, we also compute EEZ per capita (total EEZ area divided by population). This offers a transparent proxy for each country’s marine asset endowment per person, which is particularly informative for SIDS and other Large Ocean States; EEZ-normalized indicators are commonly used to characterize ocean endowments and development challenges69. Including EEZ per capita alongside BCW per capita reveals cases where small populations and very large jurisdictions imply outsized per-person ocean assets.

These per capita metrics allow for a clearer and more equitable assessment of how countries leverage their blue natural capital to address climate change. For consistency across population-adjusted indicators, we also report GDP per capita and sovereign debt per capita in our analysis.

Joint regimes and overlap claims

In our dataset, some maritime areas are classified as either Joint Regimes (JR) or Overlapping Claims (OC). JRs are areas subject to formal shared governance arrangements between two or more States, whereas OCs are disputed maritime areas without such agreements. Because these areas cannot be assigned to a single sovereign country, their blue carbon wealth (BCW) is estimated separately. The database includes 21 JRs and 11 OCs.

To estimate their potential BCW, we use an econometric model fitted on observed country-level data, in which log(BCW) is regressed on log(EEZ area) and continent indicators. The model is estimated by OLS with HC3 heteroskedasticity-robust standard errors. Predicted values for JRs and OCs are retransformed to levels using Duan’s smearing estimator70 and then aggregated. Uncertainty around the aggregate is quantified by parametric simulation from the estimated coefficient covariance matrix.

As a benchmark, we also apply continent-specific observed average BCW per km2, weighted by EEZ area, to the total area of each JR and OC (Supplementary Table 7).