In February 2026, CEN published CWA 18349 — the first European guidance written specifically for insurers and investors who have to decide whether an ecosystem is material to a risk. It is about thirty pages, it is free to read, and it is the most useful document in this field right now.
It also has a blank column, and the blank column is the reason this post exists. The agreement tells you when a living system is material enough to sit on a term sheet. It is quiet on who pays to keep the system in the condition the underwriting assumed.
what CWA 18349 actually is
A CEN Workshop Agreement is a consensus document, not a certifiable standard. There is no conformity assessment scheme behind CWA 18349, no accreditation, and nobody — including us — is "CWA 18349 certified." What it provides is a shared reference for judging whether a nature-based insurance or investment proposal is credible, so that an underwriter in Milan and an investment committee in London are reading the same evidence in the same order.
It was sponsored by the NATURANCE project under Horizon Europe, with the workshop secretariat run by UNI, the Italian national standards body. It is explicit that it extends rather than duplicates what already exists: the IUCN Global Standard for Nature-based Solutions for ecological quality, ICMA's Green Bond Principles and the LMA's Green Loan Principles for financial transparency, the IFC Performance Standards for environmental and social risk, and UNEP FI's Principles for Sustainable Insurance for underwriting practice.
The document has two halves. Five guiding design principles, and a shared metrics framework for monitoring, reporting, and verification.
the five principles, and the column they leave blank
| principle | what CWA 18349 asks you to prove | what ensurance adds |
|---|---|---|
| economic soundness | A documented baseline risk assessment and a with/without comparison showing risk reduction "large enough to influence underwriting, pricing, or investment decisions," with uncertainty ranges disclosed | A condition record on the named place, re-read between renewals, not only at loss — so the "with" case does not quietly go stale |
| financial viability | A lifecycle business case: capital and operating costs, monitoring, maintenance, end-of-life, who pays, who benefits, and how avoided losses or a greenium get captured | A funding rail for the operating line — proceeds routed to the agent that represents the place, rather than a one-time capital grant with no crew behind it |
| environmental integrity | A net nature-positive outcome, ecological baseline, measurable condition indicators, adaptive management with thresholds set in advance, and stewardship agreements | Condition is the thing being funded, so ecological improvement is the instrument's subject rather than a co-benefit reported afterwards |
| social value and ethical practice | Early and continuous engagement, equitable distribution of benefits and burdens, recognition of Indigenous and local knowledge, free prior and informed consent where applicable | An explicit, on-the-record recipient of funds. Consent and benefit-sharing still happen on the ground; the record makes them checkable rather than implied |
| governance, data, and assurance | Named custodianship, MRV proportional to scale, auditable and reproducible indicators, documented provenance, and spatial attribution clear enough to prevent double counting | Claims and evidence attached to a specific agent and asset, timestamped, with the routing of funds visible to anyone who wants to check |
Read the middle column on its own and it is a good piece of work. Read it next to a calendar and the gap shows up: the standard asks you to name who pays for maintenance and monitoring. It does not supply anyone to do it, or an instrument to book it through.
resilience measurement: the bar is higher than most people expect
Principle 1 is the gate, and it is stricter than the summaries suggest. It is not enough to show that a marsh reduces surge or that a thinned forest burns cooler. The reduction has to be large enough to change a price. The agreement asks for a transparent baseline, a with/without project comparison using accepted catastrophe or risk models, and core metrics an actuary recognizes: change in annual average loss, probability of exceeding a loss threshold, avoided downtime, beneficiaries protected, indicative premium effects.
Then it asks two questions that will fail most proposals. Explain why the risk reduction would not occur without the solution. And explain how risk shifting to somewhere else is avoided. That is an additionality test and a leakage test, wearing an insurance suit.
Resilience measurement, in this context, is the practice of quantifying how much a system's condition changes the losses a hazard produces. It is a different exercise from hazard measurement, which most catastrophe models already do well. Hazard measurement asks how hard the wind blew. Resilience measurement asks what the marsh, the canopy, or the reef did to the loss on the way through — and that answer moves as the ecosystem's condition moves.
what it looks like when a project clears the bar
The agreement's own annex works through a case, and it is worth knowing because it is the cleanest example of resilience measurement changing a price.
Willis, a WTW business, structured a $2.5 million parametric wildfire product for the Tahoe Donner Association in Truckee, California, in collaboration with The Nature Conservancy and the Center for Law, Energy and the Environment at UC Berkeley. The policy covers 1,345 acres of the association's forested and recreation land. Because roughly 1,520 acres had been treated since 2015 — mechanical and hand thinning, mastication, limbing, prescribed fire — the underwriting explicitly accounted for the fuel reduction. Gross premium came in around $55,000 instead of about $90,000, and the each-and-every-loss deductible at $10,000 instead of about $62,000. Roughly 39 percent off the premium and 84 percent off the deductible, because of forest management. (The launch release rounds the deductible cut to 89 percent; the technical report's own table says 84.)
That is the first policy anyone involved knows of that put fuel-reduction work into the pricing in writing. It is a genuine milestone, and it deserves to be described as one.
It is also described honestly in the agreement, which is the reason the document is worth your time. The risk reduction was not estimated with fire-spread simulations inside a probabilistic catastrophe model. It came from a simplified counterfactual: two historic fires in the Lake Tahoe region, an earlier WTW and TNC study concluding annual average loss could fall by roughly 40 percent from fuel reduction, and confirmation from the local fire district that recent fires stopped at treated ridgelines. A standards body publishing its flagship case study with the limits of the analysis printed next to the result is behaving well.
Note what the policy still pays on, though. The trigger reads burn severity from satellite imagery after a fire. The forest work changed the price. The contract still pays after the burn.
claims and evidence, in plain language
Principle 5 is where most nature finance quietly falls apart, so it is worth restating without the standards vocabulary.
A claim is an assertion about a place: this marsh attenuates surge, this watershed holds sediment, this canopy cuts peak temperature. Evidence is what somebody outside your organization can check. CWA 18349 asks for documented provenance and version control on the data, spatial attribution precise enough that two sponsors cannot both book the same acre, MRV proportional to the scale and risk of the solution, and independent review at milestones once the stakes get high enough.
Our version of that is ecosystem stocks, service flows, and measured condition — the accounting layer that makes a parcel legible to a funder, scored on present condition rather than a projected outcome. It gives an underwriter something to look at between renewals. It is not a performance guarantee, and it does not by itself prove an ecological outcome. Anyone selling you a nature dataset as proof of outcome is selling you the map.
One honest limit, stated plainly. Pricing a function is a bridge to funding it — never a claim that the function is the value. A condition score is not the worth of a shoreline, and a marsh that fails one funder's materiality screen is not less alive; it is simply not material to that mandate. Invert the order and you have stopped funding a living system and started appraising a dead one.
how to run this on one place
If you are holding an actual exposure — a shoreline, a watershed, a wildland-urban interface — the agreement is usable this week. Five steps, in order.
- Name the hazard and the exposed asset. Not "climate." A specific peril, a specific set of assets or populations, and the critical functions that amplify the loss if they fail.
- Establish the ecological baseline. Habitat condition, connectivity, hydro-morphological and water-quality indicators where relevant, with the date and method recorded.
- Build the with/without comparison. Use an accepted risk model where you can, a defensible simplified counterfactual where you can't — and say which one you used, as the Tahoe Donner analysis did.
- Name who pays between events. The standard requires you to name this party. It does not help you find one. Establishment, monitoring, and adaptive care are recurring costs; premiums, payouts, and capital grants are not recurring revenue for them.
- Name the custodian and the MRV plan. One accountable party for condition, indicators that are auditable and reproducible, revision thresholds set in advance.
Steps 1 through 3 and step 5 are the standard's territory, and it is well mapped. Step 4 is where most proposals go quiet, and it is the one that determines whether the evidence in steps 1 through 3 is still true in five years.
If step 4 is the step you are stuck on for a real place, that is the conversation worth having. Talk it through with someone who works on this, or read the adjacent argument about why insurer capital that reduces losses behaves differently from insurer capital that chases yield.
the series
Six posts on nature-based insurance — what it does well, what its structure cannot reach, and who funds the living system between events.
- nature-based insurance is still insurance — the definition and the timing trap
- a payout is not a reef — how coral reef parametric cover actually works, and what $1m cannot do
- when nature is material enough to underwrite — CWA 18349 and the evidence bar (this post)
- the book that walks away still pays — non-renewal ends the policy, not the loss
- put nature in the captive, not just the claims file — retention and a condition-funding sleeve
- a million dollars after the hurricane — scale, stated honestly
Adjacent reading: underwriting natural infrastructure like an asset class, what living infrastructure actually is, insurability is the first domino, and there is no risk transfer.