A sustainability desk can complete a full TNFD-aligned assessment — every priority location identified, every dependency named, every score populated — and still be unable to answer the question the CFO asks next: which specific failure, along which specific route, would cost us money, and how close is it?
That gap has a name, and it is the whole subject of this post. Exposure is not risk.
what biodiversity risk actually means
Biodiversity risk is the financial risk to a business arising from its relationship with living systems, and it runs in both directions. On the dependency side, it is what happens when an ecological function the business relies on stops delivering reliably — water of a given quality and quantity, pollination, soil stability, flood attenuation, a stable local rainfall regime. On the impact side, it is the transition, liability, and reputational risk that follows from the damage a business does to nature: regulation, litigation, and licence to operate. TNFD, the ENCORE database, and EthiFinance's own research all treat both halves as in scope.
This post is about the dependency half, and specifically about the physical route by which an ecological change reaches a cashflow. That is not the same thing as operating near degraded nature. It is what happens when something you actually rely on fails.
Insurers already have precise language for this. In catastrophe modelling, risk decomposes into three terms: hazard, exposure, and vulnerability. Hazard is the shock. Exposure is what sits in the way. Vulnerability is whether the thing in the way breaks, bends, or absorbs it.
That decomposition is exactly where the nature version comes up short — and not in the way it is usually claimed. A property catastrophe model has all three modules. It has hazard, it has an exposure database, and it has vulnerability: damage functions that say what a given windspeed does to a given building type. Those damage functions took the industry decades of claims data to build, and they are the reason the output is a loss number rather than a map.
Nature finance has built the exposure layer with real rigour. Nature-risk data platforms map millions of asset locations against biodiversity, water stress, and ecosystem integrity data. That work is genuine and it is not the problem. What does not exist yet is the equivalent of a damage function for an ecological dependency: the thing that says how much service loss follows a given ecological change, and at what point that loss becomes a step rather than a slope. That missing module is the subject of this post.
exposure is not risk
The framing in this post's title is Gutierrez's. In a public essay post from May 2026 he stated it as three clauses: the number is not the diagnosis, exposure is not risk, and the missing object is the pathway.
The distinction matters because the two things answer different questions and fail in different ways.
| exposure (a location score) | the dependency pathway (the missing vulnerability layer) | |
|---|---|---|
| the question | where do our assets sit, and what condition is nature in there? | what does this cashflow lean on, through what mechanism, and how close is that mechanism to failing? |
| the object produced | a scored site, or a portfolio of scored sites | a named route: source → mechanism → flow → beneficiary |
| the unit | indices, hectares, ratings at a coordinate | edges, weights, thresholds, and a margin to failure |
| what it sees well | proximity to degradation; sector and geography concentration | remote dependencies, substitutability, and where a shock propagates |
| what it cannot say | whether the service keeps arriving | the loss in euros — it is an input to risk, not risk itself |
| how it fails | scores a site low-risk while the dependency that matters sits 2,000 km away | draws a convincing network and calls a backtest a forecast |
Both failure modes are real. A dependency graph that has not been tested against null models is a picture, not a diagnosis.
And note what the right-hand column is not. A pathway is not risk either — it is the vulnerability term, and risk still needs a hazard and an exposure alongside it. The title's logic cuts both ways, which is the point. What it indicts is the common practice of stopping at the left-hand column and reporting the result as risk.
Exposure tells you where you are standing. The pathway tells you whether the thing you are standing on keeps working. Risk needs both, and a shock.
what TNFD LEAP asks for, and what it cannot supply
The TNFD's LEAP approach is four phases: Locate your interface with nature, Evaluate your dependencies and impacts, Assess your nature-related risks and opportunities, and Prepare to respond and report.
Most critiques of TNFD get the next part wrong, and it is worth getting right before claiming any gap. LEAP already asks for the pathway. "Dependency pathway" is a defined term in the guidance, inherited from the Natural Capital Protocol: a description of how a particular business activity depends on ecosystem services and specific features of natural capital, and how changes in that natural capital affect the costs and benefits of doing business. The Evaluate phase asks organisations to understand how impact drivers and external factors influence the state of nature and therefore the ecosystem services they receive, "recognising these may be beyond the organisation's site boundary." The guidance even supplies the remote example: a clean, regular freshwater supply may depend on the health of forests far upstream. Box 15 asks organisations to consider threshold effects and tipping points, because the relationship between drivers, state, and services "will not always be smooth."
So the gap is not that TNFD forgot to ask. It asked in 2023, in the right vocabulary, with the right caveats, including the remote and non-linear cases.
The gap is that asking is not supplying — and the guidance says so itself. From the LEAP guidance, in the section on assessing changes in the state of nature:
There is not yet a comprehensive, methodologically consistent global reference source describing linkages between drivers of nature change, changes to the state of nature and changes to the availability of ecosystem services. Organisations therefore need to form their own assessment of the state of nature in locations relevant to their LEAP due diligence.
Read that as an underwriter. The framework asks each organisation to construct, on its own, the causal link from an ecological change to a service it depends on — at the resolution of a named source, with a threshold attached — and states that no consistent reference exists to do it from. What comes back in practice is a defensible inventory of dependencies at a set of locations, because the inventory is the part a desk can actually finish. The mechanism is requested and left blank.
The remote case shows the shape of it most clearly. LEAP tells you to look past the site boundary and even names the upstream forest. It gives you no instrument for identifying which upstream forest, or for quantifying how much of your water depends on it. That is the difference between a scope instruction and a method. We have written elsewhere about the supplier your map can't see and about how a place depends on another place; this post is about the layer that would let you compute either.
the CAC 40 stress test shows the seam
In 2026 the rating agency EthiFinance published a biodiversity stress test of the CAC 40. It is a serious piece of work, and worth describing accurately, because what it reaches and what it leaves out are both instructive.
The method measures how sensitive each sector is to ecosystem services using the ENCORE database, propagates that sensitivity through global supply chains with an EXIOBASE input-output model, and converts the resulting production loss into a change in probability of default through a Merton-Vasicek framework, consistent with recent European Central Bank work. Its central assumption is that for multinationals of this size, biodiversity exposure sits above all with suppliers and in the regions where those suppliers operate. This is not a "where the headquarters sits" score. It follows the shock through the value chain.
Their result across five natural capitals: water is the most financially material. Under their pessimistic scenario (SSP5-RCP8.5), worsening water stress drives roughly a 33% increase in average probability of default between 2025 and 2050. Species and habitats follow at roughly 24% and 23%. Atmosphere and soils produce no significant financial effect at that horizon — either because sector dependencies are weak, or because the economic loss-of-functionality thresholds are not reached in the scenarios modelled.
Note that last clause: thresholds are inside this model, not deferred to future work. What the model does not do is let the five capitals interact. Each is degraded and propagated independently, with no ecological feedback between them.
Then the finding that should interest an underwriter more than the headline: exposure was not the decisive variable. Credit quality was. While companies hold high ratings, the increases stay modest, which is why the index looks resilient. Simulate downgrades and the response turns sharply non-linear — one notch multiplies probabilities of default by roughly five, two notches by twenty to thirty, across all five capitals. Two companies with identical nature exposure can face very different outcomes, and the difference is the buffer, not the hectare.
Be precise about which buffer that is. Credit quality is a financial buffer: distance to default, measured on a balance sheet.
The study also produced its own most interesting number, and flagged it itself. Under the optimistic scenario, water-related default risk runs higher than under the pessimistic one, because a more sustainable agricultural transition consumes more water, particularly in temperate regions. Their reading is that scenarios have to be interpreted as structural transformations of production, not only as degrees of degradation. That is the right reading, and it is also the shape of the thing this post is about: the sharpest result in the exercise came from two systems interacting — agricultural transition and water — rather than from any single capital degrading on its own. The design captured that interaction because it arrived through the scenario inputs. What the same design cannot generate is the equivalent interaction arising from ecology, because feedback between the capitals is not modelled.
None of that is a knock on EthiFinance. It is the cleanest public demonstration of where the frontier actually sits. Reading the study publicly in September 2026, Gutierrez put it this way: the exposure envelope is now computable with real rigour, while the route from ecological impairment to buffer failure still sits below the model's resolution.
the missing object is the dependency pathway
The next step is to ask whether the same logic holds one layer down — and it is worth being explicit that this is a proposition, not something the study establishes. EthiFinance measured a financial buffer. The claim on offer is that there is a second buffer beneath it, an ecological buffer, with its own distance to failure.
The case for taking that seriously is the shared mechanism. A system can look resilient precisely because its buffers are still working. That is true of a balance sheet, and it is true of a watershed. Where it holds, the most valuable indicator is not the current level of degradation — it is the distance to the point where the buffers stop working. Gutierrez proposes modelling that ecological distance as a graph robustness margin: the smallest disruption to ecological dependencies that pushes service delivery below a specified threshold.
Two buffers, two different measurements, and only one of them currently has a method. The study measures the financial buffer, well. Nobody yet measures the ecological one at a resolution a credit committee would accept. That is the honest state of play, and conflating the two is how this argument goes wrong.
Underneath sits a definitional problem he keeps returning to, inherited from economic accounting. Natural capital is conventionally defined as a stock of assets yielding a flow of benefits. But what actually produces water reliability, pollination, and coastal protection is living network infrastructure — a web of dependencies among species, habitats, and physical flows, whose wiring determines resilience, function, and the service itself. Stock metrics tell you how much remains. The dependencies that decide whether the service keeps arriving stay invisible.
His chain for making a pathway inspectable runs: source → mechanism → flow → exposure → consequence, with evidence and uncertainty attached at every step. The purpose is not to turn a forest into a spreadsheet. It is to make the route legible enough to question, govern, and monitor. That object — the dependency graph — is the subject of the pillar post in this series.
That work is his. Fluvion, which prices a teleconnected water dependency from an upwind Amazon forest to a downwind soy harvest, and AI Ecologist, which turns a species list into a map of what holds an ecosystem together and where it breaks first, are his public engines — published with their own caveats, indicative rather than prudential grade, and honest about backtest versus forecast. We collaborate with him. We do not run his models, and nothing here should be read as a claim that we do.
from a location to a mechanism
If you run a nature assessment and want it to survive contact with a credit committee, five moves change what you get out of it.
- Keep the location work. It is the prerequisite, not the enemy. You cannot reconstruct a pathway without first knowing what you touch.
- Rank by cashflow, not by hectares. The dependency worth reconstructing is the one carrying the most revenue at the most concentrated point — usually not the biggest number on the map.
- Write the route in four terms. Source, mechanism, flow, beneficiary. If you cannot fill all four, you have an exposure, not a risk.
- Ask for the margin, not the score. How far is this mechanism from the point where it stops delivering, and what is the evidence for that distance? A model that cannot answer is describing the present, not pricing the future.
- Ask who would pay to keep the source working. Once the beneficiary is named, funding the source stops being philanthropy and becomes loss avoidance with a counterparty.
funding the source, once the pathway is named
That fifth move is where we come in, and it is deliberately the last one.
A disclosure names an exposure. A pathway names a source, a mechanism, and a beneficiary — which makes a different question available: would the beneficiary rather fund the condition of that source now, or absorb the failure later?
That is what ensurance is for. Not a payout after the loss, but capital that funds the present condition of a named natural asset before it fails. Coins fund protection across the protocol; certificates fund a specific named source directly. Both depend on the same prerequisite this whole post is about: someone has to name the pathway first.
We are frank about our own stage. Live instruments, real agents, small volumes. The pathway problem is not solved, by us or by anyone. But the ordering is clear — disclose the location, reconstruct the mechanism, then fund the source.
One clarification, because this post has spent two thousand words on the limits of stock metrics. A stock metric used as a risk metric is the thing being criticised. Stocks-and-flows accounting is a different instrument: flows are the service arriving, which is precisely what the pathway says is at risk. The pathway names which flow is exposed and how close it is to failing. The ledger is where that flow is accounted once it arrives. You need both, and neither substitutes for the other.
See the stocks and flows ledger →
frequently asked questions
what is the difference between nature exposure and nature risk?
Exposure measures proximity and interface: where your assets, suppliers, and financed activities sit relative to nature, and what condition nature is in there. Risk is what happens to your cashflows when a specific ecological function you depend on stops delivering. Exposure is one input to risk, not a synonym for it — in catastrophe-model terms, risk requires hazard, exposure, and vulnerability together. Nature finance has built exposure well and vulnerability barely at all.
what does TNFD LEAP miss?
Not what most critiques claim. LEAP — Locate, Evaluate, Assess, Prepare — explicitly asks for dependency pathways, tells organisations that the ecosystem services they depend on may originate beyond their own site boundary, and asks them to consider threshold effects and tipping points. What LEAP cannot do is supply the method, and the guidance says so: there is not yet a comprehensive, methodologically consistent global reference source describing the linkages between drivers of nature change, changes to the state of nature, and changes to the availability of ecosystem services. Each organisation is therefore asked to construct that link itself, with no consistent reference to build from. LEAP asks the right question at a resolution the available science cannot yet answer.
why do stock metrics fail on living systems?
Because a stock metric measures how much is left, and living systems fail on connectivity rather than volume. Two forests with identical canopy cover can occupy completely different positions in the network that delivers a downstream service: remove one and the service degrades, remove the other and nothing moves. Stock metrics are also poor at thresholds. Living systems tend to absorb stress with no visible signal and then change state abruptly, which means a count that looks stable right up to the transition was never a risk metric.