ost date 19 Oct 2026 (original LinkedIn date)
Original LinkedIn version· 2,897 words
‘End of August’ by Noah Kahan from his 2026 album, ‘The Great Divide’. Let this slow-burn, opening track from his new album roll over you and soothe your brow.
Lloyds Banking Group published two reports on resilience these last couple of months. Read together, they set out a cogent argument but I think it can be taken a bit further.
The first, The Missing Line Item, asks what a business is exposed to when the natural systems it depends on deteriorate. The second, Built to Last, asks how long-life resilience infrastructure gets financed. One is about demand for resilience; the other is about the supply of capital to build it. What sits between them should be a contract, and it is the contract that neither report fully grapples with. To paraphrase a friend and colleague, Anne Reaney, if nobody can say with sufficient confidence what is being bought, by whom, and on what basis payment is due, then the whole thing risks dissolving into worthy collaboration with no investable proposition behind it.
That point matters because I remain wary of the argument that, where nature-based outcomes are uncertain, we should retreat from contracting for outcomes altogether and rely instead on collective governance, catchment institutions or voluntary contributions from beneficiaries. All of those may have a role, but none creates a payer by itself. Somebody still has to decide that a service is worth purchasing. The challenge is therefore not to dispense with outcomes, but to be rather clearer about where they sit in the commercial structure.
I think traditional infrastructure turns out to be quite helpful here. A traditional engineered flood defence may be funded because it reduces expected flood losses; a reservoir may be justified because it improves security of water supply; & a wastewater treatment plant exists because society wants cleaner rivers. Yet the contractor operating any of those assets is not normally paid according to whether the ultimate social or economic outcome is realised in every period.
The flood-defence operator is not paid according to the value of flood losses avoided that year, the reservoir operator is not paid according to whether a drought happened, and the treatment plant does not receive its fee according to whether the receiving river achieves some wider ecological state over which many other factors also have an influence. The outcome provides the reason to invest. The payment mechanism usually sits somewhere closer to what the supplier can actually deliver and control. Could that distinction be exactly what nature finance has been missing?
The demand side: a number on the dependency
The Missing Line Item, produced with the geospatial firm Earth Blox, takes individual sites and asks which ecosystem services they quietly rely on: water supply, storm buffering, soil retention, pest and disease control in a sourcing region. It then expresses the exposure as a share of site revenue or asset value in a severe but plausible year, a 1-in-20 event, dependency by dependency. They call it Nature Value at Risk.
The numbers are by no means trivial: two paper mills in the south-eastern US carry 5.5% to 6.3% of revenue at risk from loss of the storm protection that surrounding forest and wetland provide. Three manufacturing sites carry 9% to 12% from water. A water-cooled data centre in Boulder carries 13.1%, against 8.8% for its sibling in Columbus, and the difference is almost entirely explained by what the landscape around each one is doing.
There are two findings that I believe matter more than the headline figures. The first is that in the paper case, the largest exposure sits on land the company neither owns nor operates: the fibre sourcing region, where pest control alone puts 13.3% of expected fibre growth at risk. The report’s own conclusion is that managing this will require supplier engagement and sourcing decisions rather than site capex. The second is the manufacturing site the report calls Site B, where extensive plant cover in the surrounding landscape is explicitly credited with holding erosion exposure well below what the hazard would otherwise imply. The authors note that investment in the surrounding landscape can complement conventional risk management.
That sounds like the catchment argument, made by a bank, in a bank’s language. A corporate’s resilience is partly a function of land it does not control, and the value of that land’s condition can now be put on the same page as interest rates and commodity prices.
The supply side: capital follows a revenue that already exists
Built to Last is the financing story. It sets out the £11 billion annual adaptation gap from the Climate Change Committee, the case for treating resilience as a credit priority rather than a policy nicety, and three transactions Lloyds has recently financed: the £3 billion Haweswater Aqueduct replacement, the Cleve Hill solar and storage park on the flood-prone Kent coastline, and a £500 million university retrofit programme. Let’s look at what the three have in common. Haweswater has a 34-year Ofwat concession under the regulated customer-bill framework and a £300 million subordinated National Wealth Fund facility backed by a Treasury guarantee. Cleve Hill has a power purchase agreement and a Contract for Difference. The university programme has a 70% National Wealth Fund guarantee on eligible lending. In every case the asset was bankable because a regulated or contracted revenue already existed, with public credit enhancement sitting underneath it.
Nature appears in all three, but only as a co-benefit: 10% biodiversity net gain at Haweswater, a 56-hectare habitat area at Cleve Hill, hedgerows on campus; so nature is never the asset that generates the cash flow. The report’s own list of what still needs building includes “developing revenue models for resilience” that translate long-term benefits into predictable cash flows. It then offers no example of one, which should not be read as a criticism. It is simply an accurate description of where the market is.
The missing contract
Taking the two reports together, I think the gap is pretty obvious. One can now tell a corporate that a meaningful proportion of a site’s revenue depends on the condition of a catchment it does not (fully or partially) own. The other can show how institutional debt and public guarantees will finance a resilience asset once it has contracted revenue. What is clearly missing is the instrument that turns the first into the second: an agreement under which the beneficiary of upstream resilience pays for it.
I have argued before that the binding constraint in nature finance is not valuation, nor is it data, and nature VaR is a good illustration as to why. The measurement problem is increasingly tractable, by banks, with satellites, at site level. But the reason the money still does not flow is that nobody has signed anything – in short, there is no counterparty on the other side of the exposure. The difficulty, I think, is that nature makes the obvious form of contract rather uncomfortable.
It is relatively straightforward to contract for an intervention: restore this peatland, create this wetland, or reconnect this floodplain. It is considerably harder to promise that the consequence will be a precisely specified reduction in downstream flooding, not because the intervention has no effect but because rainfall, antecedent soil moisture, other land-use decisions, drainage infrastructure and a changing climate all determine how the translates into the ultimate outcome. Understandably that has led some to question whether an outcome contract is appropriate at all. But I think we can look to traditional infrastructure to suggest an answer.
What grey infras already knows
So we need to allow for an important distinction between the outcome that justifies an investment and the performance against which the provider is paid. Let’s consider a flood wall - its economic case may be based on reducing expected annual damages and lowering the probability of severe loss. But the contractor is not asked to wait for the next major flood and then prove how much damage it prevented before receiving payment. Instead, the contract defines the asset that is to be provided, the conditions under which it is expected to perform, the required standard of performance and the obligations to inspect, maintain and repair it.
There is also a distinction between the asset failing and the design basis being exceeded. A defence designed for a particular standard of protection has not necessarily failed because a rarer event overtops it, as the residual risk was there all along. That seems obvious for grey infrastructure, but we have not always afforded nature the same treatment. A wetland may be functioning exactly as intended and a downstream community may still flood. A restored peatland may be slowing and retaining water while an exceptional rainfall event overwhelms the wider catchment. If the contract treats the occurrence of flooding as evidence of non-performance, we can make a strong case that it has confused hazard exceedance with asset failure.
Note that traditional infrastructure would normally separate those risks. There is delivery risk: was the asset built and maintained to specification? There is performance risk: within the defined operating or design envelope, does it perform as promised? And there is residual hazard risk: what happens when the world presents conditions outside that envelope? The first can reasonably sit with the supplier. Much of the second can too, provided that the performance test is something the supplier can influence. The third normally cannot. That gives us a much better way of thinking about a Resilience Purchase Agreement.
What, then, is actually being purchased?
The corporate still needs an outcome to justify spending the money. If modelling shows that deterioration of a catchment exposes a data centre to a material water or flood risk, that establishes the business case for investment, but the outcome does not necessarily have to be the payment trigger. The contractual service could instead sit one step upstream. The provider might first be required to maintain a portfolio of interventions in an agreed ecological condition. That matters, but condition alone is probably insufficient. A bridge can be beautifully maintained and still fail its load-bearing requirement. So traditional infrastructure normally asks something about function as well as condition.
For nature, the equivalent might be whether the restored system is displaying the hydrological or ecological characteristics through which resilience is expected to arise: water storage, infiltration, connectivity, water-table behaviour, soil function or some other measure appropriate to the intervention. The structure then begins to look less like paying for hectares and less like asking a farmer to guarantee the weather.
Instead the business case is built around a modelled change in downstream risk. The design basis describes how the portfolio of interventions is expected to produce that change; the contracted service is the maintenance of ecological condition and functional performance; monitoring tests whether the system continues to behave broadly as anticipated; the wider outcome model is periodically recalibrated as evidence accumulates; and the buyer retains the residual risk that an event may occur outside the design envelope.
There is still an outcome contract in an important sense. The buyer is not paying for nature because nature is intrinsically nice to have; rather it is paying because it has a quantified exposure and good reason to believe that changing upstream condition will alter that exposure. But the contract no longer requires every element of that causal chain to be re-proven before an invoice can be paid. This feels much closer to the way traditional infra is normally financed.
What the metric can and cannot do
This also clarifies, to some extent, the role of Nature VaR, when it is tempting to over-use it. It is a directional estimate of exposure under a scenario, not an expected loss, and Lloyds say so plainly. It is calculated per dependency with no stated treatment of correlation between them, even though the report’s own Site A discussion is about how wildfire, erosion and water availability compound one another. And it measures the current exposure of a site, not the marginal value of an intervention upstream. So a number that prices the problem is not necessarily the number that should determine the payment and that is an important distinction.
Nature VaR belongs in the origination conversation and the credit file. It gives the corporate a reason to act and establishes an upper bound on what it might rationally spend to reduce the exposure. But the RPA payment mechanism should be built around the resilience service being provided, with the ultimate downstream outcome used to establish the value of that service and to test periodically whether the original proposition still holds. Or to put it slightly differently, the downstream outcome may be a valuation metric without having to be the payment metric. That is not an evasion for outcome-based finance, it is simply what makes it workable.
A corporate does not need a precise loss estimate to decide that a fee of a fraction of 1% of revenue is worth paying against an exposure of 12%. It needs to know that the exposure is real, that it is local, that something upstream can move it and that the thing it is paying for continues to perform in a way consistent with that proposition.
The RPA as infrastructure contract
This is the problem a Resilience Purchase Agreement exists to solve, although I now think the analogy with any environmental credit is probably less useful than the analogy with an infrastructure service contract. We now have some clearer roles and responsibilities. The land manager should be responsible for delivering and maintaining the intervention. The project vehicle should be responsible for assembling the interventions into a coherent portfolio, monitoring their condition and functional performance, and responding where evidence shows that the portfolio is not behaving as anticipated. The corporate buyer pays for that resilience service because its exposure provides the commercial rationale. And the downstream outcome remains central because without it there is no reason for the corporate to be there. But it is treated as probabilistic rather than binary.
This also leaves room for buffers, reserves and, eventually, insurance. If modelling suggests a particular change in risk, there is no necessity to monetise the full central estimate. Part can be withheld as a buffer against ordinary model uncertainty. Defined performance risks might sit within the delivery vehicle. More extreme residual risks might eventually be transferred elsewhere. What matters is that we stop asking one contractual clause to carry every form of uncertainty at once.
Nature infrastructure may be more difficult to model than a pipe or a concrete wall, but the commercial challenge is not entirely new. Infrastructure finance has spent decades deciding which risks sit with the builder, which with the operator, which with the buyer and which simply remain part of the world in which the asset operates. Nature finance needs to now start doing the same.
Why a bank should want the contract
Two details in Built to Last explain why this is now a lender’s problem as much as a landowner’s. The first is the Prudential Regulation Authority’s Supervisory Statement 5/25, which from December 2025 expects UK banks to assess the credibility of their counterparties’ adaptation plans. A corporate borrower with a signed agreement that reduces a quantified dependency has something concrete to show, whereas a corporate with a sustainability report does not. The second is the Nature Markets Framework’s target of £1 billion a year of private investment into nature in England by 2030, which explicitly names flood-risk reduction as an outcome that private capital should fund.
So the bank has a regulator asking it to look for adaptation plans, a policy framework asking it to finance nature outcomes, a metric that can locate the exposure to the site, and a financing toolkit that works once contracted revenue exists. Everything is in place except the contract, and perhaps we are now a little clearer about what that contract should and should not try to do: it should not ask a land manager to guarantee that a corporate site will never flood, bearing in mind that traditional flood infrastructure does not make that promise either. Nor should it simply pay for environmental activity, because that breaks the link between the corporate exposure and the reason for investing.
It needs instead to connect a modelled change in physical risk to a resilience service whose condition and functional performance can actually be observed, contracted and maintained. That is where the next round of work sits. It is not primarily analytical work, rather it is legal and commercial: defining the design basis, agreeing the functional performance measures, deciding where residual risk sits, building monitoring that both sides trust, and creating a delivery vehicle that a lender can underwrite.
Unglamorous, perhaps, but that is the nature of the game, and we’re already part way there. Lloyds have written the two halves of the case. The missing line item and the missing revenue model are the same thing seen from opposite ends. The line item becomes revenue the moment someone agrees to pay it, and traditional infra gives us a useful reminder of what they may actually need to agree to buy.
Referencces:
Lloyds Banking Group, Built to Last: Climate resilient infrastructure needs transformative investment (September 2026)
Lloyds Bank in association with Earth Blox, The Missing Line Item: Financial exposure hidden in environmental dependencies (2026)
Climate Change Committee, A Well-Adapted UK (2026).