---
title: "the bet that isn't in your scenarios"
canonical_url: https://ensurance.app/guide/the-bet-that-isnt-in-your-scenarios
markdown_url: https://ensurance.app/guide/the-bet-that-isnt-in-your-scenarios.md
subtitle: a stress test reports the risk you modeled. the source your site depends on is usually the scenario nobody ran
category: nature-finance
---

# the bet that isn't in your scenarios

*a stress test reports the risk you modeled. the source your site depends on is usually the scenario nobody ran*

**What is climate scenario analysis?** Climate scenario analysis is testing how an asset, a company or a portfolio would fare under several possible climate futures, usually hazard paths tied to a higher and a lower emissions pathway, over a set horizon. It reports a loss or a change in value under each path. It is a set of what-ifs, not a forecast.

Picture the slide. A company has run its plants through a high-warming and a low-warming path to 2050, and the flood, heat and wind numbers for its largest plant come back modest. Then someone asks where the plant's water comes from: a reservoir below a forest that burned two summers ago, where the next heavy rain will wash ash and sediment toward the intake. None of the scenarios asked about the forest. (The plant is an illustration, not a named case.)

The forest above that reservoir, the slope that holds its soil and the floodplain that spreads the high water were setting the plant's hazard long before anyone scored it, and they keep doing that work whether or not anyone pays for it. [**Ensurance**](/manual/ensurance?from=guide) is how the people who depend on ground like that can pay for its keeping. It is not the forest, and it does not stop a fire or a storm.

## what climate scenario analysis does well

Scenario analysis exists because the past is a weak guide to a warming climate. Instead of one forecast, it runs the same assets through several plausible futures and shows how the answer moves. Supervisors use it to stress banks and insurers (what that can and cannot do is in [a stress test is not financial stability](/guide/a-stress-test-is-not-financial-stability?from=guide)), and companies use it for disclosure and capital planning.

A guide for adaptation and resilience finance published in November 2026 by Standard Chartered and KPMG asks banks and investors to test an investment against climate scenarios, "ideally one high-emissions and another lower, over the relevant asset life," and to ask whether it still works "if climate change intensifies beyond current projections." That is good practice. Done well, it tells a board which sites carry the most hazard and which decisions hold across paths.

## a stress number only covers the bets you ran

Every scenario result has a boundary. A **value at risk** figure (the loss a book should not exceed at a chosen confidence level over a set period) or a stress loss under a named path adds up the exposures you put in the model, under the futures you chose to run. It says nothing about the exposures you left out. A clean result can mean the risk is small. It can also mean the risk was never in the run.

That is not a flaw in the method; it is what a model is. The trouble is that the result travels further than its boundary: a slide that says "physical risk: low" gets read as a statement about the plant, when it is a statement about the plant's coordinates under the hazards someone scored. Leaving something out of the run does not take it off the books. You still hold the position. You have just stopped reporting it.

## the scenario nobody ran: the source

For physical risk, the scenario most often missing is the **source**: the water supply, the forested slope, the floodplain or the marsh that a site depends on and usually does not own. Three things make it easy to leave out.

**It sits off the site.** Scenario runs start from an asset list and a set of coordinates. The forest above the reservoir is not on the list, so its condition rarely enters the hazard layer as more than a land-cover class, and almost never as a row that changes, has an owner or could be worked on.

**It fails on its own clock.** A climate path moves the weather gradually over decades. A source can change in one season: a fire burns the slope, beetles kill a stand, a drought draws the reservoir down, a storm moves a hillside of sediment into the channel. A hotter path can make some of those more likely, but they arrive as steps, and a run that scores hazard at the site's coordinates can miss a step that happens upstream.

**It reaches the site through several doors at once.** A burned watershed can raise flood and debris risk at the plant, foul its water, close its road and empty the town its workforce lives in. A scenario set that scores flood depth at the site counts one of those doors. [The loss started outside the property line](/guide/the-loss-started-outside-the-property-line?from=guide) follows one such chain; when one source sits under many sites at once, that is [what systemic risk actually is](/guide/what-systemic-risk-actually-is?from=guide).

| question | scenario sets usually include | they usually leave out |
|---|---|---|
| Where | The registered address, or the site's own coordinates | The ground upstream or upslope that sets the site's hazard |
| What is scored | Hazard at the site: flood depth, heat days, wind speed | Condition of the source: burned, beetle-killed, drained, silted |
| How it changes | Smooth climate paths to 2030 or 2050 | Step changes on the source's own clock: fire, beetle, drought, sediment |
| How it reaches you | One risk line per peril | Several channels from one source: flood, water, debris, access, workforce |
| Who owns it | The company's own assets | Ground held by an agency, a neighbor or many small owners |

The first row is not hypothetical. A 2025 note from the Network for Greening the Financial System (NGFS), a network of central banks and supervisors, says banks often hold only a borrower's registered address, and that leaving out sites other than headquarters "grossly underestimates exposure to climate hazards." It cites a study of 177 listed firms that own 1,820 physical assets in Mexico, which found investor losses underestimated by up to 70% when asset-level information is left out; the same note records an analysis by European Central Bank and Banque de France staff finding the headquarters shortcut can be reasonable at portfolio level. Treat 70% as one sample, not a rule. And if the site is often missing, the ground above it is further away still.

The Standard Chartered and KPMG guide asks whether an adaptation investment's contribution is "not contingent on conditions that may not hold." The same question is worth asking of the plant itself: its water is contingent on a forest in poor condition.

:::johnson
**the bet you did not model is still a bet.**

A clean stress test reports the scenarios you ran. The forest, slope or floodplain your site depends on is a position whether or not it is in the run.

[read next: the watershed isn't in the asset register →](/guide/the-watershed-isnt-in-the-asset-register?from=guide)
:::

## how to add the missing bet to a climate scenario analysis

None of this needs a new model, only a longer list and one more case beside each climate path.

1. **Map the sources.** For each material site, name what it draws on and what stands between it and the hazard: the watershed above the intake, the slope above the access road, the floodplain upstream of the yard. [How to map what your property depends on](/guide/how-to-map-what-your-property-depends-on?from=guide) walks through that ring.
2. **Read their condition and their owners.** Note what has happened to each source (a burn scar, a drained wetland, a low reservoir) and who owns it.
3. **Write the source-failure case beside each climate path.** Under each path, add one case where the source fails: the slope burns, the reservoir runs low, the floodplain is cut off. Say what reaches the site, through which channel, for how long.
4. **Run it through the same loss chain.** A hazard specialist estimates what the failure does physically; a loss model turns that into damage and downtime on the same assets and events as your other cases.
5. **Report it as a position.** Say it even when the number is rough: this plant depends on a burned watershed we do not own, and the failure case adds this much interruption.

Some dependents already pay for that position. In California, Yuba Water Agency put $1.5 million into a 2018 Forest Resilience Bond pilot and then $6 million toward a $25 million second bond for about 48,000 acres of the Tahoe National Forest, in what the partners call its source watershed. No avoided-loss figure has been published.

## what this does not do

Raising, hardening and insuring the plant still do real work: they protect the building from the hazard that arrives. They do not change the ground it comes from.

Keeping a source in good condition mostly moves frequent and mid-size losses. It does not buy back the extreme tail: a record fire or flood can exceed any buffer, and top-of-the-curve cover still has its job. [The average year is not the year that breaks you](/guide/the-average-year-is-not-the-year-that-breaks-you?from=guide) sorts which losses living systems move.

We do not run scenario sets, compute stress losses or sell climate models. Hazard scores and catastrophe models from specialist suppliers are inputs, and each case still needs a hazard specialist. What ensurance adds is a way to pay for the source once you have named it: an account for the place, a certificate for the named ground, and proceeds routed to the people keeping it. A certificate starts as ensured and is meant to move toward permanent protection; ensured alone does not finish the job.

## next steps

If you hold a scenario report, read it for its boundary. To see which ecosystem services your sector leans on, start with [nature exposure by sector](/exposure?from=guide). For how a company can map and fund the services it depends on, see [ensurance for corporations](/solutions/corporations?from=guide&topic=after-the-score). Next in this series: [the watershed isn't in the asset register](/guide/the-watershed-isnt-in-the-asset-register?from=guide).

## frequently asked questions

### what is climate scenario analysis?

Testing an asset, a company or a portfolio against several possible climate futures, usually a higher and a lower emissions path over a set horizon, and reporting the loss under each. It is a set of what-ifs, not a forecast.

### what does climate scenario analysis miss?

Whatever is not in the run. For physical risk that is most often the source a site depends on, such as the watershed or floodplain upstream, which fails on its own clock and reaches the site through several channels at once.

### how do you add a water source to a climate scenario analysis?

Name the source for each site, read its condition and owner, write a case where it fails beside each climate path, and run it through the same loss model. Report the result as a position.

## sources

[Standard Chartered and KPMG, *Guide for Adaptation and Resilience Finance* (GARF 2.0), November 2026](https://www.sc.com/en/uploads/sites/66/content/docs/Adaptation-Finance-Reports.pdf) — Screening principles, p. 21: Test against climate scenarios, "ideally one high-emissions and another lower, over the relevant asset life"; an adaptation investment's contribution "not contingent on conditions that may not hold"

[NGFS, *Leveraging physical climate risk data*, information note, September 2025](https://www.ngfs.net/system/files/2025-09/NGFS_information%20note_Leveraging%20physical%20climate%20risk%20data.pdf) — Section 3.2, pp. 13–14: Banks often hold only the registered address; excluding non-headquarters sites "grossly underestimates exposure"; footnote 4 notes the headquarters proxy may be reasonable at portfolio level

[Bressan, Đuranović, Monasterolo and Battiston (2024), *Nature Communications*](https://pmc.ncbi.nlm.nih.gov/articles/PMC11217445/) — The study NGFS cites: 177 listed firms owning 1,820 physical assets in Mexico; investor losses underestimated by up to 70% when asset-level information is neglected

[World Resources Institute, new Forest Resilience Bond will finance $25 million of restoration (October 2021)](https://www.wri.org/news/release-new-forest-resilience-bond-will-finance-25-million-restoration-reduce-wildfire-risk) — Yuba II: $25 million for about 48,000 acres of the Tahoe National Forest after a 15,000-acre pilot; Yuba Water Agency's $1.5 million (2018 pilot) and $6 million commitments "in their source watershed"; funding from utilities, corporations and state agencies

## the series

1. [the score is the easy half](/guide/the-score-is-the-easy-half?from=guide) — **Read first:** What a climate risk score leaves undecided, and the five questions after it
2. [the average year is not the year that breaks you](/guide/the-average-year-is-not-the-year-that-breaks-you?from=guide) — Expected annual loss is an average; the bad year sets the capital, and living systems move some of it, not all
3. [the bet that isn't in your scenarios](/guide/the-bet-that-isnt-in-your-scenarios?from=guide) — A stress test reports the risk you modeled; the source your site depends on is usually the scenario nobody ran
4. [the watershed isn't in the asset register](/guide/the-watershed-isnt-in-the-asset-register?from=guide) — A climate risk model needs hazards, assets and damage curves; the ground that sets the hazard has no row
5. [waiting is a position](/guide/waiting-is-a-position?from=guide) — When the bad years get worse, the protection is worth more and each year of waiting costs more
6. [the cost-benefit case for the ground upstream](/guide/the-cost-benefit-case-for-the-ground-upstream?from=guide) — A worked example on a hypothetical water utility, from the loss to a payment a budget can book