The ghost carbon in the mangrove.

The world's largest mangrove restoration issued close to half a million carbon credits, and roughly 168,000 of them are for carbon the trees never captured, because the trees never grew. That is the finding of a peer-reviewed remote sensing reassessment of the Livelihoods mangrove project in Senegal, registered on the Verra Verified Carbon Standard as VCS 1318. The number matters less than the mechanism behind it. This was not a rogue developer falsifying records. Every actor in the chain followed the methodology, passed the audits and earned their fee. The standards system did not fail to catch the ghost carbon. It produced the ghost carbon, certified it and concealed it from the buyers, while operating exactly as designed.

This account draws on the Andrieu et al. reassessment published in Environmental Challenges in June 2026, the project's registry and verification history on Verra, the Livelihoods Funds account of the programme, the BeZero rating note, the REDD-Monitor briefings, and the Hakai Magazine field reporting. Where a figure is contested or estimated, it is flagged in the text. The dollar value of the implicated credits, in particular, is an estimate, not a disclosed number.

The promise: the largest mangrove restoration ever attempted

Between 2009 and 2012 the Senegalese NGO Océanium de Dakar coordinated the planting of tens of millions of mangrove propagules across the Sine Saloum and Casamance deltas, an area recorded at 10,415 hectares. The Livelihoods account puts it at 80 million mangroves planted by more than 100,000 villagers across some 450 villages in three years. The restoration was financed by the Livelihoods Carbon Fund, created in 2008 by Danone alongside the Ramsar Convention and the IUCN, and backed by a roster of European corporates including Danone, Hermès, Michelin, Crédit Agricole, Mars and Schneider Electric. The carbon sat on top of the planting, registered under Verra's VM0033 tidal-wetland methodology, projecting on the order of 500,000 tonnes over the life of the scheme.

Two features of that pitch matter for what follows. Scale was the selling point, and scale was counted in hectares planted rather than hectares that survived. And the value rested on additionality: the claim that this mangrove would not have returned without the intervention. Both the scale and the additionality would later be the exact points the science contradicted. At the moment of sale, neither was in doubt, and that is what a buyer reasonably relied on.

The warning: the auditor saw it in 2014

The first signal that promise and reality were diverging came from the project's own verifier. In 2014, TÜV SÜD South Asia reported, while confirming the project's credits, that around a quarter of sampled plots showed near-total loss of tree cover. The same cycle named the ecological mistake underneath: the project had planted Rhizophora, a species poorly suited to the saltier, higher mudflats where much of the planting had gone, and the verifier recommended switching to the hardier Avicennia. The species switch was never carried out. The recommendation was repeated at the next verification. The crediting continued.

This is the moment that defines the case. A verifier, paid by the project, looking at the project's own plots, found the planting dying and named the reason, and nothing stopped. The verifier's job under the methodology was to confirm the rules were followed, not to halt issuance because the outcome looked poor. There was no party in the chain whose mandate, and whose pay, depended on the trees actually being alive. The developer earned on issuance, the verifier on verification, the registry on registration. The only party with an interest in fewer but real credits was the buyer, and the buyer was not in the room.

The audit: in 2021 the failure was confirmed, and the credits went up

Seven years later the divergence had not closed. A 2021 verification by TÜV NORD CERT confirmed the recommended Avicennia replanting still had not happened. Then comes the part that turns bad luck into something structural. In the same monitoring report, the project revised its soil carbon component upwards. Mangrove carbon sits in two pools: the visible biomass of the trees, and the carbon held in the waterlogged soils beneath them. The soil pool is the larger one, and the hardest to verify from outside, because it cannot be seen by satellite and is expensive to sample. By raising the soil carbon factor, the project generated roughly 40 per cent more credits than the original design anticipated.

Read the order carefully, because the order is the point. The trees were confirmed failing. The species correction was confirmed undone. And in the same window the accounting was revised toward more credits, drawing on the one carbon pool the verification chain was least equipped to challenge. Forest cover can be checked from space. Tree survival can be checked, expensively, in the field. Soil carbon is checked by the developer and confirmed on paper. The harder a quantity is to verify from outside, the more the system trusts the developer's number, and the soil pool is the hardest of all.

The harder a quantity is to verify from outside, the more the system trusts the developer's number. The soil pool is the hardest of all.

The science: the satellite put a number on the ghost

The figure that gives this case its weight arrived in June 2026, from outside the chain entirely. Julien Andrieu of the Université Côte d'Azur and colleagues combined Sentinel and Planet imagery with botanical field plots and time-series rainfall to ask the question the chain never asked: did the planted areas actually become mangrove, and where they did, was it the planting or would the mangrove have returned anyway. The answers were stark. Survival ran at roughly 18 to 20 per cent of what was planted. Thirty-six per cent of plots were total failures. Only 31.7 per cent of the project area showed mangrove cover, and of that, only around 19 per cent could be attributed to the project rather than to spontaneous regeneration. On that basis, about 168,000 credits, roughly a third of the total, are ghost carbon.

The most damning detail is that the finding did not come from nowhere. Andrieu had already shown, in Forest Ecology and Management in 2020, that 96 per cent of mangrove regeneration in the Saloum Delta was spontaneous, driven by the recovery of rainfall after the Sahel droughts, not by planting. The additionality claim at the heart of the project had been undercut in the peer-reviewed literature six years before the reassessment, by the same author, looking at the same delta. The market kept issuing through all six years. Up to 55 hectares had been planted in salty mudflats with no tidal flooding, where Rhizophora cannot establish. The planting chased area; the ecology punished it; the gap between the two is the ghost carbon.

This is not an allegation of fraud. It is a finding that the project's monitoring extrapolated from the surviving fraction to the whole, that the methodology permitted it, and that a third of the credits represent carbon the satellite cannot find. The verification chain checks the developer's model for internal consistency. It does not check the model against an independent measurement of the ground. That adversarial test is the one thing no one in the chain is paid to perform. As of June 2026, Verra had not publicly responded to the paper, and the VCS 1318 credits remained registered, with no suspension, cancellation or buffer adjustment announced.

The chain: who bought the ghost, who rated it, and why it is still live

Most of the credits were retired anonymously through the registry, which is the market's normal state and its first concealment. The documented buyers give a sense of the spread: Robeco appears in the record; Amaury Sport Organisation retired 5,500 credits against the 2014 Paris Marathon; Michelin used credits via a 2014 challenge; the organisers of the Paris 2024 Paralympic Games retired credits from the project. None is accused of bad faith. They did what a responsible buyer was supposed to do: they bought Verra-registered credits, verified by accredited third parties, under an approved methodology, with a strong community story. The registry record they received told them the serial number, vintage, methodology and verifier. It did not tell them that a quarter of sampled plots had near-total mortality in 2014, that the species correction was never made, that the soil carbon was revised up in 2021, or that a third of the carbon would later prove to be a ghost.

The ratings layer is meant to fix exactly this asymmetry, and the case shows where it stops. BeZero rated VCS 1318 at BBB, explicitly flagging significant non-permanence risk and notable over-crediting risk from potential over-estimation of soil carbon. The rating agency identified the soil carbon problem as a published flag, and the credits kept selling anyway, because nothing in the market structure required a buyer to act on it. The diligence existed. It was advisory. The integrity layer flagged the risk, and the issuance layer overrode it.

The financial magnitude is genuinely uncertain. There is no disclosed price for the implicated credits; VCS mangrove credits traded across a wide band over the period, very roughly 8 to 25 dollars per tonne, putting the ghost tranche somewhere in the low single-digit millions. That is an estimate built on a price range, not a disclosed figure. The more reliable number is the human one, documented by Hakai Magazine: of a project budget recorded at around 4.4 million dollars, roughly 5.2 per cent, on the order of 22 dollars per hectare, reached the communities who did the planting. The villagers carried the labour. The upside went up the chain.

What the ghost carbon tells us

The structural reading is the one to hold on to, because it makes this a system story rather than a scandal. No one needed to lie for the outcome to occur. The developer extrapolated within the methodology. The verifier confirmed adherence and recorded its concerns as observations. The registry issued against an approved method. The rating agency published its warnings. The buyer relied in good faith on the stamp. Each actor behaved rationally given its incentive, and the incentives all pointed one way: more issuance. The only check that would have caught the ghost carbon, an adversarial test of the counterfactual from outside the chain, was the one thing no one in the chain was paid to perform. It took a university remote sensing team, with no commercial stake, fourteen years after planting began, to do it.

The cost did not fall evenly. It cost the villagers their labour for roughly 22 dollars a hectare. It cost the Casamance and Sine Saloum deltas more than a decade in which a real restoration might have been attempted with the right species in the right hydrology. And it cost the atmosphere around 168,000 tonnes counted as removed and now sitting as residual emissions in the inventories of corporates who believe they cancelled them. The buyers thought they held the integrity-priced asset. The cruelty of the case is that nothing they could see distinguished the real credit from the ghost. The certificate looked identical on both sides of the line.

Disclosure: Rodrigo Diaz is the founder of Kyroq, the independent risk and valuation layer for carbon as a financial asset. We disclose founder business interests where editorial subject matter overlaps. Editorial decisions and source selection are independent of Kyroq commercial relationships.