What a small bay in Wales tells us about blue carbon

”I don't wanna argue, I don't wanna fight I just wanna go dancin' with you tonight See everybody happy everybody have another round Let's sing that song one more time Before the ship goes down The water's rough it's a stormy sea Before you know it we'll be history”

‘Before the Ship goes down’ by Glenn Frey, released in 1988 on his solo album Soul Searchin’

In May 2024, a group of marine biologists in Wales watched a year of work pretty much disappear. They had sown over a million seagrass seeds across two bays on the North Wales coast as part of one of the larger restoration efforts ever attempted in Northern Europe. The 2023 cohort had established and was doing well. The 2024 cohort, planted with the same methods and the same seed stock, did not.

The cause was an unusual marine heatwave. Between 5th and 12th of May 2024, sea surface temperatures around Bae Tremadog rose 3.5°C in seven days. Intertidal sediment exposed at low tide reached 30°C. Salinity dropped sharply and wave heights spiked. The paper documenting what happened, published earlier this year by Thomas and colleagues at Swansea and Plymouth, makes for sobering reading, particularly when you notice that another planting fifty kilometres up the coast at Bae Caergybi survived the same heatwave and is now going from strength to strength.

That contrast is the part worth dwelling on. The two sites had been selected using essentially identical methods: historic evidence of seagrass presence, proximity to a donor meadow, and habitat suitability modelling that scored both as optimal. The decisive difference was something the models did not capture. Bae Caergybi sits in deeper shelf water with greater hydrological mixing and a slower long-term warming trend, so the same heatwave reached it as a 1-1.5°C anomaly rather than the 2-2.5°C event that hit Tremadog. One project survived and the other did not. Is that all the story? Well maybe pause and let's ask what this tells us, because I think the answer matters well beyond marine biology.

What the Welsh case documents at high resolution is a structural failure mode that runs through a lot of nature-based investment thinking. Habitat suitability models, climate-zone agronomic maps, insurance loss models, project-level NPV calculations: most of these are built on mean conditions and historical variance. They work well in a stationary climate. They fail, often in ways that only become visible once something breaks, when the variance itself is shifting and stressors begin to compound. The Tremadog failure was not a thermal event. It was a thermal event followed by a salinity drop followed by a wave-exposure spike, acting in sequence on plants that were too young to be resilient. None of those things individually exceeded historical tolerance ranges, but their joint occurrence did. We see such things in the markets all the time.

This is conceptually adjacent to the developing El Niño story, though I want to be careful with the linkage. The 2024 North Atlantic heatwave was not an ENSO event, rather it was driven by Atlantic blocking, anomalous winds, and reduced Saharan dust loading, so the mechanism does not transfer, even if the framing does. A strong El Niño hitting an already stressed food and fertiliser system would expose the same kind of mismatch between models calibrated on historical means and outcomes shaped by extreme tail draws (and if you hadn't heard, there's likely one coming later this year). Even a moderate event can produce non-linear damages when the buffers (fiscal space, aid envelopes, fertiliser supply, strategic grain reserves) are already drawn down. The Tremadog failure is a microcosm of that dynamic - the seedlings were not killed by an event that exceeded any single threshold. They were killed by a sequence of moderately bad draws that the design framework had no way to anticipate.

This pattern matters for investors thinking about nature recovery as an asset class because it interacts uncomfortably with the carbon-credit economics that underpin most blue carbon thesis work. A paper published earlier this year by Ward and colleagues did the quantitative work that the conservation community had been gesturing at for some time, and the numbers should give anyone pricing seagrass blue carbon pause. Using the only meadow on the planet with the right kind of paired data (the Virginia Coast Reserve LTER), they modelled four management approaches across a ten-year horizon. A six-hectare seeding project generates around $1,189 in carbon revenue over the decade. A 100-hectare conservation project, preventing one metre of sediment loss from dredging, generates up to $1.53 million over the same period. The asymmetry is between thirteen and thirty-three times. They suggest that the economic case for seagrass blue carbon turns out to live almost entirely in prevention rather than regeneration. The sediment under a meadow represents 170 to 200 years of accumulation. Replanting cannot recreate that on commercially relevant timescales, wheeras protecting it from being dredged or stirred can.

Two implications follow that I have not seen properly priced into the way blue carbon project pipelines are being marketed. The first is that the addressable market for seagrass restoration as a carbon revenue stream is significantly smaller than the headline area-based targets suggest, because the unit economics on regeneration are marginal at any realistic voluntary carbon price. The second is that the addressable market for marine sediment protection is significantly larger, because most existing project frameworks have not been designed to capture that value. The investable opportunity sits closer to a coastal version of avoided-deforestation than to afforestation, and the diligence questions are and should be different. You need to know what would otherwise happen to the sediment, not just what could be planted on top of it.

Permanence is the next problem. Ward and co-authors are explicit that their model assumes "perfect survival and persistence of meadows through time," and equally explicit that this is an assumption rather than an empirical claim. They cite the 2018 Arias-Ortiz paper showing that a single marine heatwave caused massive losses from Shark Bay's seagrass carbon stocks, the largest meadow on the planet. Verra, the main issuer of voluntary blue carbon credits, is currently mid-revision of its long-term reversal monitoring framework precisely because the permanence assumption is under pressure. The Tremadog failure is exactly the kind of event that would trigger a reversal under any sensible methodology, and unlike credit defaults in conventional asset classes, blue carbon reversal risk is heavily correlated across geographies. The 2023 and 2024 North Atlantic heatwaves hit thousands of kilometres of coastline simultaneously. A diversified portfolio of blue carbon projects spread across the UK and Northern Europe would have offered little protection. This is a non-diversifiable tail risk that is currently being priced as if it were idiosyncratic.

There is also a model-risk question that the Ward paper raises almost in passing but I think it's suggestion is worth taking seriously. Their result depends on data from a single site, because Virginia LTER is the only meadow globally where all four key carbon parameters have been measured with appropriate before/after or vegetated/unvegetated controls. Every blue carbon model for Zostera marina that anyone is relying on right now traces back to the same calibration data. The robustness of that data within its own context is high. The generalisability of it to other temperate sites is largely unknown. For a private credit investor looking at a blue carbon pipeline, this is the equivalent of pricing an entire mortgage book off the historical loss experience of a single zip code. It is not a critique of the science (the authors are admirably forthright about the gap), but it is a flag for anyone using these methodologies for capital allocation.

A few brighter notes, because the picture is not uniformly bleak. The Sievers paper that came out late last year showed that the configuration of restoration plantings matters considerably for biodiversity outcomes, and that the optimal arrangement is patchy rather than contiguous, with multi-habitat assemblies (seagrass plus mussel reefs plus boulder reefs) outperforming monoculture replants of equivalent area. This is operational, empirical support for the seascape ecology framework that Wedding and colleagues laid out earlier this year as the "5Cs" of context, configuration, connectivity, scale and culture. Taken together, they point to a real opportunity to identify projects that will deliver better ecological outcomes per dollar than the standard area-target approach which currently dominates the field because regulators and funders are scoring on hectares. Markets generally reward what they can measure, and the gap between hectares restored and biodiversity outcomes delivered is where attentive capital can probably find an edge over the next decade.

There is also a recalibration of the addressable market worth registering. Green and Unsworth's 2021 analysis of UK seagrass loss argues that mainland Britain may have lost up to 92% of its historical seagrass, against a high-certainty estimate of 44-50% loss since 1936. The methodological caveats are real and the authors are open about them, but the implication holds. Standard suitability models substantially under-price the area available for genuine recovery because they implicitly anchor to a baseline that is already deeply degraded. If the historical extent is anywhere near the upper estimate, the carbon stock that has been lost from UK seagrass meadows alone is worth somewhere around £1.8 billion at Net Zero-consistent carbon prices. Realising even a fraction of that would require regulatory architecture that does not yet exist. A separate paper from Unsworth and an international team last year flagged that current UK licensing rules actively constrain restoration by, among other things, dictating donor meadow provenance without considering thermal compatibility. That is the precise mechanism that may have contributed to the Tremadog failure: the donor meadow at Porthdinllaen was thermally suited to its own bay, not necessarily to the planting site twenty kilometres east where conditions diverged.

Pulling all of this together: the marine restoration sector sits at an awkward inflection point. The science is finally producing the kind of detailed, project-level evidence base that should let serious capital underwrite it properly. The economics, when looked at honestly, are different from the dominant pitch. Protection is significantly more carbon-efficient than regeneration. Configuration matters more than gross area. The addressable historical baseline is larger than current models recognise. The permanence assumptions underpinning voluntary carbon credit accounting are being eroded by exactly the climate-variance dynamics that make restoration projects fail in the first place. And the regulatory framework, in the UK at least, is currently constraining the upside while not pricing the downside.

For anyone trying to think clearly about physical climate risk and nature recovery, the small Welsh failure is more useful than it looks. It is a worked example of how the joint distribution of stressors matters more than the marginal one, how mean-based models miss extreme outcomes that increasingly drive realised performance, and how the asset class with the most prominent climate marketing attached is also the one most directly exposed to the same risks it purports to mitigate. None of this means restoration is a bad idea. It means we should be doing more of it, but honest about what carries the economic case (protection, configuration, and refugia siting) and what does not (assumed permanence, monoculture replants, and hectare-target chasing).

The seagrass at Bae Caergybi is still growing. Picking the right bay turns out to matter quite a lot.

References

Aoki, L. R., McGlathery, K. J., Wiberg, P. L., Oreska, M. P. J., Berger, A. C., Berg, P., & Orth, R. J. (2021). Seagrass recovery following marine heat wave influences sediment carbon stocks. Frontiers in Marine Science, 7, 576784.

Arias-Ortiz, A., Serrano, O., Masqué, P., Lavery, P. S., Mueller, U., Kendrick, G. A., Rozaimi, M., Esteban, A., Fourqurean, J. W., Marbà, N., Mateo, M. A., Murray, K., Rule, M. J., & Duarte, C. M. (2018). A marine heatwave drives massive losses from the world's largest seagrass carbon stocks. Nature Climate Change, 8(4), 338-344.

Berthou, S., Holliday, R., Pirret, J., & Wallace, E. (2024). Investigating driving factors for the UK-average temperatures in May 2024: A study of the impact of local sea surface temperatures and of climate change. Met Office.

Berthou, S., et al. (2024). Exceptional atmospheric conditions in June 2023 generated a northwest European marine heatwave which contributed to breaking land temperature records. Communications Earth & Environment, 5, 287.

England, M. H., Li, Z., Huguenin, M. F., Kiss, A. E., Sen Gupta, A., Holmes, R. M., & Rahmstorf, S. (2025). Drivers of the extreme North Atlantic marine heatwave during 2023. Nature, 642, 636-643.

Govers, L. L., Heusinkveld, J. H. T., Gräfnings, M. L. E., Smeele, Q., van der Heide, T., & Chapman, M. G. (2022). Adaptive intertidal seed-based seagrass restoration in the Dutch Wadden Sea. PLoS One, 17(2), e0262845.

Green, A. E., Unsworth, R. K. F., Chadwick, M. A., & Jones, P. J. S. (2021). Historical analysis exposes catastrophic seagrass loss for the United Kingdom. Frontiers in Plant Science, 12, 629962.

Hobday, A. J., et al. (2016). A hierarchical approach to defining marine heatwaves. Progress in Oceanography, 141, 227-238.

NOAA Climate Prediction Center. (2026, May). ENSO Diagnostic Discussion. National Oceanic and Atmospheric Administration.

Oreska, M. P. J., McGlathery, K. J., Aoki, L. R., Berger, A. C., Berg, P., & Mullins, L. (2020). The greenhouse gas offset potential from seagrass restoration. Scientific Reports, 10, 7325.

Sawall, Y., Ito, M., & Pansch, C. (2021). Chronically elevated sea surface temperatures revealed high susceptibility of the eelgrass to winter and spring warming. Limnology and Oceanography, 66, 4112-4124.

Sievers, M., Brown, C. J., Rasmussen, J. A., Nielsen, B., Steinfurth, R. C., Flindt, M. R., Banke, T. L., Gilby, B. L., & Connolly, R. M. (2025). Optimizing seagrass planting arrangements for animal benefits in a multihabitat restoration seascape. Ecological Applications, 35(4), e70163.

Thomas, O. R., Patterson, B. M., Rowland, C., Berthou, S., Worsfold, M., Yates, E., Keen, C., den Haan, S., & Unsworth, R. K. F. (2026). Marine heat waves endanger the rebuilding of marine life. Environmental Research Letters, 21, 104019.

Unsworth, R. K. F., Bertelli, C. M., Coals, L., Cullen-Unsworth, L. C., den Haan, S., Jones, B. L. H., Rees, S. R., Thomsen, E., Wookey, A., & Walter, B. (2023). Bottlenecks to seed-based seagrass restoration reveal opportunities for improvement. Global Ecology and Conservation, 48, e02736.

Unsworth, R. K. F., Sweet, M., Govers, L. L., von der Heyden, S., Vergés, A., Friess, D. A., Jones, B. L. H., Monfared, M. A. A., Steinfurth, R. C., Fariñas-Franco, J. M., Cullen-Unsworth, L. C., et al. (2025). Rethinking marine restoration permitting to urgently advance efforts. Cell Reports Sustainability, 2, 100526.

van Katwijk, M. M., Thorhaug, A., Marbà, N., Orth, R. J., Duarte, C. M., Kendrick, G. A., et al. (2016). Global analysis of seagrass restoration: The importance of large-scale planting. Journal of Applied Ecology, 53(2), 567-578.

Verra. (2023). Methodology for Tidal Wetland and Seagrass Restoration (VM0033). Verified Carbon Standard.

Ward, M., Dibble, C., Millington-Drake, M., Lynch, J., McGlathery, K., Lilley, R. J., Strong, A. L., & Wedding, L. M. (2025). Management approach matters: meeting seagrass recovery and carbon mitigation goals. npj Ocean Sustainability, 4(1), 18.

Wedding, L. M., Stuart, C. E., Govers, L. L., Lilley, R. J., Olds, A., Preston, J., Tavasi, L. E., & Pittman, S. J. (2025). Five ways seascape ecology can help to achieve marine restoration goals. Landscape Ecology, 40, 115.

Yeager, L. A., Keller, D. A., Burns, T. R., Pool, A. S., & Fodrie, F. J. (2016). Threshold effects of habitat fragmentation on fish diversity at landscape scales. Ecology, 97(8), 2157-2166.