---
title: how an intervention lowers the loss
canonical_url: https://ensurance.app/guide/how-an-intervention-lowers-the-loss
markdown_url: https://ensurance.app/guide/how-an-intervention-lowers-the-loss.md
subtitle: "expected annual loss falls only when a specific action changes the hazard or the damage; a model shows by how much, and a saving needs someone to book the difference"
category: nature-finance
---

# how an intervention lowers the loss

*expected annual loss falls only when a specific action changes the hazard or the damage; a model shows by how much, and a saving needs someone to book the difference*

Expected annual loss is the probability-weighted average of what a set of possible events would cost a set of buildings, roads or plants in one year. It is a model output, not a bill: each event's chance, times the damage it does to what is exposed, added up. An intervention lowers it only if it changes one of those terms. The marsh in front of a street, the thinned forest above a town and the floodplain left open upstream can each shape how hard a storm or a fire arrives. To the extent it does that work, it does it whether or not anyone pays for it. Showing how much it lowers the loss takes four steps, usually done by different people, and each has a limit worth naming.

:::johnson
**the saving is the difference between two runs.** Same storms, same buildings, with and without the work. A percentage borrowed from "better condition" is not a saving.

[See the four steps →](#the-four-steps-and-who-does-each)
:::

## what expected annual loss means

A catastrophe modeler builds a long list of possible events: thousands of simulated storms or fires, each with an annual probability. For each event it estimates the hazard at each building (water depth, wave height, flame length), and a damage function turns that into lost value. Multiply each event's loss by its probability and add them up. That sum is the **expected annual loss** (EAL), also called average annual loss (AAL).

EAL is an average: most years see less, a few far more. It is not the premium. A carrier's price adds expenses, the cost of reinsurance and a return on the capital it holds for the bad year ([how insurance premiums are calculated](/guide/how-insurance-premiums-are-calculated?from=guide)).

## the four steps, and who does each

| step | the question | who usually does it | the honest limit |
|---|---|---|---|
| 1. Physical mechanism | What does the action change: runoff, fuel, slope, wave energy? | Hydrologists, engineers, fire scientists, ecologists | A mechanism at one site does not carry to another by default |
| 2. Hazard-to-loss model | How does that move the hazard or vulnerability term, same events, same buildings? | Catastrophe modelers | Only as good as the event set, exposure data and damage functions |
| 3. Expected benefit | EAL without the project minus EAL with it | Modelers, then benefit-cost economists | A counterfactual estimate, not proof a future loss was stopped |
| 4. Recognition | Who books the difference? | Underwriters, lenders, utility and city finance staff | Not a lower price until a contract, filing or budget accepts it |

### step 1: the physical mechanism

Thinning and prescribed fire can lower the fuel load and make it harder for fire to climb into the canopy. Marsh mounds put ground where open water was and can slow waves before they reach the houses. A reconnected floodplain can lower the crest downstream.

Each mechanism needs its own model. U.S. Geological Survey (USGS) post-fire debris-flow assessments estimate debris-flow likelihood and volume for given rainfall after a burn: part of the hazard, not a treatment's effect. The InVEST (Integrated Valuation of Ecosystem Services and Tradeoffs) sediment model leaves landslides out. An erosion model is not a debris-flow model, and swapping one for the other is how a good-looking number goes wrong.

### step 2: the hazard-to-loss model

Run the same events against the same exposure twice, without and with the project, changing only the term the mechanism changes. A marsh that lowers wave height at the door changes the hazard term. Hardening the house changes the vulnerability term. Never assume a percentage because the land is "in better condition." Condition is a reason to look, not a result.

The building side has claims data: in Hurricane Sally (2020), FORTIFIED homes had claim frequency 55% to 74% lower, one storm in one state, per a University of Alabama study as reported. FORTIFIED is the Insurance Institute for Business & Home Safety (IBHS) standard for roofs ([does mitigation lower insurance premiums](/guide/does-mitigation-lower-insurance-premiums?from=guide)). Land and house work are complements; an honest model carries both.

### step 3: the expected benefit

EAL without the project minus EAL with it is the expected annual avoided loss. It is a counterfactual, and field data can calibrate it. The Swiss Re Institute found that Florida areas with the best-protected coastal habitat had about half the National Flood Insurance Program (NFIP) claim frequency in frequent, less severe storms (2009 to 2022). A UC Davis study reported that fuel treatments on about 300 fires (11 states, 2017 to 2023) avoided about $2.8 billion in damage and more than 4,000 buildings lost. Field comparisons and model-based estimates like these test a mechanism. They do not prove that one project will stop one loss.

Over a term, the annual figure becomes a present value set beside life-cycle cost: the next post, [the quote beside the cost of doing nothing](/guide/the-quote-beside-the-cost-of-doing-nothing?from=guide).

### step 4: recognition

A modeled reduction becomes money only when someone books it: a carrier's credit or priced policy, a lender's terms, a budget line. Sarah Conway, formerly WTW's (Willis Towers Watson) ecosystem resilience lead, argues reduced risk often fails to become a lower premium or financing cost; in her reading, ecosystems more reliably keep a place insurable than they lower its price. Say which of the two your case is in.

## worked cases

### graveline bay: a whole loss model that stopped at the price

About 60 acres of marsh, 10 large and 45 small mounds, were finished on the north side of Dauphin Island, Alabama, in April 2023. Moffatt & Nichol modeled waves and water levels for 2-year to 100-year storms (step 1). In The Water Institute's insurance pilot, Munich Reinsurance America estimated average annual loss for 30 to 40 nearby homes on Cadillac Avenue (step 2). Modeled AAL fell 14% in Zone 1, 9% in Zone 2 and 18% in Zone 3 (step 3). The tables are not uniform: in Zone 3's 2-year storm, modeled wave height rose from 0.03 ft to 0.10 ft.

Step 4 did not happen. The team chose not to quantify a premium, agents had no process to quote the marsh, and models may not take in a new elevation model for upwards of five years. Graveline is a rare built project modeled through step 3 for specific houses; its result is modeled loss, not a lower bill.

### northern sierra and tahoe donner: from a model to one placement

After a 2021 study by The Nature Conservancy (TNC) and WTW put ecological forestry into Northern Sierra wildfire catastrophe models (modeled AAL down 40% to 60%), the Tahoe Donner homeowners' association in California placed a policy in April 2025 on 1,345 acres of its land, with a premium 39% lower and a deductible 89% lower than the sponsors' stated counterfactual without credit for the forest work. Step 4 happened there, once: one association's placement on its own land, not homes, and not a filed or typical discount.

### hurricane sandy wetlands: built on the industry's own models

Narayan and colleagues (2017), using catastrophe models from RMS (Risk Management Solutions) and Guy Carpenter, estimated that coastal wetlands avoided about $625 million in direct damage across the Northeast during Sandy. In Ocean County, New Jersey, the model put the marshes' effect at 16% lower annual flood losses on average, and up to 70% lower at low elevation. Insurers' own tools produced it. Even so, we found no insurer crediting wetlands or mangroves in homeowners' premiums.

### mangroves: the method at global scale

Menéndez, Beck and colleagues (2020) ran physical flood models and expected-damage functions with and without mangroves, and put their flood protection above $65 billion a year. At a 20 km grid, that is a global ledger, not a property's rate. Beck's caveat holds: a mangrove nearby does not prove any one property's loss falls.

### hec-fda: the public template

The U.S. Army Corps of Engineers' HEC-FDA (Hydrologic Engineering Center's Flood Damage Reduction Analysis) software combines hydrology, hydraulics and economics, computes expected annual damages, and compares the without-project condition against each alternative plan. That is steps two and three in a public tool.

## who else works on the chain

No single firm does every step. The Landbanking Group (TLG) runs a chain from hazard to physical exposure, vulnerability and dependency, then to cash flows, and says ecological condition can change the hazard or the vulnerability term depending on the mechanism. It puts estimable consequences into cash flows first and applies its Physical Risk Adjustment (PRA), a judgemental shadow price, only to the residual. Its Nature Equity paper proposes contractual payments to stewards for measured outcomes; a SEKEM case study reports purchases of measured improvements. PRA reaches the [cost of capital](/guide/ensurance-and-your-cost-of-capital?from=guide).

First Street, now part of MSCI (originally Morgan Stanley Capital International), includes existing natural defenses in its flood model and claims its adaptation module covers measure costs and avoided losses for asset-level measures. OmniGeo describes pre- and post-mitigation change detection for insurers. Conway writes on step 4, why reduced risk may not become a lower price. From the public material we read, we could not confirm a completed case running from an upstream intervention to a payor's bill for any of them. That is a limit of what we found, not a claim about their work.

## where ensurance sits

Ensurance does not run catastrophe models or [pay claims](/guide/what-ensurance-cannot-do?from=guide). It is built to connect what the chain leaves apart: the source parcel that does the work, the cost of securing it, and the bill of whoever's loss it lowers. The work it is built to fund is loss control.

Today you can [read a place's expected annual loss by peril](/place?from=guide), live from the Federal Emergency Management Agency (FEMA) National Risk Index at census-tract scale. That is a lead, not one property's loss or any intervention's effect. Our own intervention-attributable avoided-loss calculation (money at risk now minus money at risk with the upstream parcels kept) is planned, not built. Until it is, any saving has to come from the partners who run step 2 and the payor who books step 4. No ensurance saving has been realized yet. A payor applies the [avoided cost test](/guide/an-avoided-cost-is-the-benefit-you-can-book?from=guide).

If you price risk, ask whether your model can see the work ([insurers](/solutions/insurers?from=guide&topic=lower-the-loss)). If you fund mitigation, ask for the with-and-without run before the grant ([governments](/solutions/governments?from=guide&topic=lower-the-loss)). If you run an intake below a burn scar, your avoided operating cost may be the cleanest number here ([utilities](/solutions/utilities?from=guide&topic=lower-the-loss)). For a specific place, [talk to someone who can help](/contact?from=guide&topic=lower-the-loss).

## frequently asked questions

### what is expected annual loss?

Expected annual loss is the probability-weighted average yearly loss from a set of modeled events: each event's annual probability times its damage to the exposed assets, summed. It is a model output that depends on the event set, exposure data and damage functions.

### does a healthier ecosystem automatically lower expected annual loss?

No. It falls only if a specific action changes a hazard or vulnerability term, and a run with and without the project shows by how much.

### is a lower modeled loss the same as a lower premium?

No. Graveline Bay's modeled AAL fell 9% to 18% by zone, and no premium was quantified. A premium also carries expenses, reinsurance and capital.

### who calculates the avoided loss from a nature project?

Usually several: engineers and ecologists, a catastrophe modeler, an economist, and the carrier, lender or budget owner who books it.

### does ensurance calculate the avoided loss today?

Not yet. It reads tract-level expected annual loss by peril; attributing avoided loss to a specific intervention is planned, not built.

## sources

[The Water Institute: Graveline Bay insurance pilot](https://thewaterinstitute.org/graveline-bay-insurance-pilot) — About 60 acres of marsh; modeled AAL for nearby properties down 9% to 18%

[The Water Institute: Graveline Bay insurance pilot findings, June 2026](https://water-institute.files.svdcdn.com/production/Projects/Project-Findings-Graveline-Bay-Insurance-Pilot.pdf?dm=1783346302) — Zone results 14%, 9%, 18%; Zone 3 2-year wave height 0.03 to 0.10 ft; premium not quantified; model updates "upwards of 5 years"

[Moffatt & Nichol: Graveline Bay marsh restoration](https://www.moffattnichol.com/project/graveline-bay-marsh-restoration) — Mound design and wave modeling

[WTW and TNC: wildfire resilience insurance, 2021](https://wtwco.com/en-US/Insights/2021/07/wildfire-resilience-insurance-quantifying-the-risk-reduction-of-ecological-forestry-with-insurance) — Ecological forestry in catastrophe models; modeled AAL down 40% to 60%

[The Nature Conservancy: Tahoe Donner wildfire resilience insurance, April 2025](https://www.nature.org/en-us/newsroom/ca-wildfire-resilience-insurance0/) — Cover of $2.5 million over 1,345 acres; premium 39% and deductible 89% lower than the sponsors' counterfactual

[WTW: Willis and The Nature Conservancy launch wildfire resilience insurance, 2025](https://www.wtwco.com/en-cm/news/2025/04/willis-and-the-nature-conservancy-launch-first-of-its-kind-wildfire-resilience-insurance) — Placement announcement naming Globe Underwriting's analysis

[Narayan et al., Scientific Reports 7:9463, 2017](https://doi.org/10.1038/s41598-017-09269-z) — Wetlands avoided about $625 million in Sandy damage; Ocean County annual flood loss down 16% on average

[Menéndez et al., Scientific Reports 10:4404, 2020](https://doi.org/10.1038/s41598-020-61136-6) — Mangroves' flood protection above $65 billion a year, with and without scenarios

[UC Santa Cruz: mangroves and property protection, 2025](https://news.ucsc.edu/2025/10/mangroves-property-protection/) — Catastrophe-model collaboration with Moody's RMS; effects vary by location

[U.S. Army Corps of Engineers: HEC-FDA](https://www.hec.usace.army.mil/software/hec-fda/) — Expected annual damages, without-project against alternative plans

[USGS: post-fire debris-flow hazard assessment](https://www.usgs.gov/programs/landslide-hazards/science/postfire-debris-flow-hazard-assessment-data) — Likelihood and volume of debris flows for rainfall scenarios

[InVEST sediment delivery ratio user guide](https://storage.googleapis.com/releases.naturalcapitalproject.org/invest/3.20.0/userguide/en/sdr.html) — Scenario-based sediment model that excludes landslides

[Swiss Re Institute: natural habitats can reduce flood losses, 2025](https://www.swissre.com/institute/research/topics-and-risk-dialogues/climate-and-natural-catastrophe-risk/coastal-and-flood-protection-natural-habitats.html) — About half the NFIP claim frequency in the best-protected Florida areas, frequent storms

[UC Davis: forest fuel treatments and wildfire damage](https://www.ucdavis.edu/climate/news/forest-fuel-treatments-reduce-wildfire-spread-prevent-28-billion-damage) — About $2.8 billion in damage and more than 4,000 buildings avoided across about 300 fires; publication date not confirmed

[Insurance Journal: FORTIFIED roofs in Hurricane Sally, 2025](https://www.insurancejournal.com/news/southeast/2025/05/28/825141.htm) — As reported, claim frequency down 55% to 74% (University of Alabama study for the Alabama Department of Insurance, ADOI)

[Sarah Conway, The Risk & Resilience Ledger: why risk reduction rarely lowers the price](https://riskandresilienceledger.substack.com/p/why-risk-reduction-rarely-lowers) — Why reduced risk may not become a lower premium or financing cost

[The Landbanking Group: physical risk pricing](https://www.thelandbankinggroup.com/physical-risk-pricing) — Hazard to exposure, vulnerability and dependency; Physical Risk Adjustment on the residual

[First Street: flood adaptation database](https://help.firststreet.org/hc/en-us/articles/360050275794-The-Nation-s-first-flood-adaptation-database) — Natural defenses included in the flood model

[OmniGeo](https://www.omnigeo.ai/) — Pre- and post-mitigation change detection for insurers

[FEMA National Risk Index](https://hazards.fema.gov/nri/) — Census-tract expected annual loss by hazard

## the series

1. [how to save money with ensurance](/guide/how-to-save-money-with-ensurance?from=guide)
2. [what ensurance saves you: the four bills](/guide/what-ensurance-saves-you?from=guide)
3. [how an intervention lowers the loss](/guide/how-an-intervention-lowers-the-loss?from=guide)
4. [the quote beside the cost of doing nothing](/guide/the-quote-beside-the-cost-of-doing-nothing?from=guide)
5. [ensurance and your cost of capital](/guide/ensurance-and-your-cost-of-capital?from=guide)
6. [split the bill for what lowers it](/guide/split-the-bill-for-what-lowers-it?from=guide)

Sister series: [how to lower insurance premiums](/guide/how-to-lower-insurance-premiums?from=guide). The other side of the ledger: [how to make money with ensurance](/guide/how-to-make-money-with-ensurance?from=guide).