EMBER & CREST · property-level assessment
Every property in your book has a wildfire and flood number. Right now you cannot show where it came from.
EMBER and CREST assess a British Columbia address, or a whole book of them, and return the formula that produced each result, the weight and resolution of every input, the date it was run and the version it was run under. Across a book they also report where the exposure concentrates, by Forward Sortation Area. When someone asks how the number was reached, the answer is already in the response.
What is coming
Every regulator you answer to is converging on one question: what is the physical risk at this property, and how do you know?
The instrument differs by who supervises you. The question does not, and neither does what it takes to answer it: the hazard at the address, what the number was built from, and when it was run.
If you underwrite
OSFI’s Climate Risk Return already asks federally regulated P&C insurers for wildfire and flood exposure, including gross and net Probable Maximum Loss at 1-in-100 and 1-in-250, and OSFI has signalled that insurance exposure reporting is moving toward Forward Sortation Area granularity. From 1 May 2027, E-23 adds an inventory, a validation and ongoing monitoring for every third-party model feeding that work. For B.C.-incorporated insurers, BCFSA consults this fall on a Natural Catastrophe and Climate-related Risk Guideline, with a Model Risk Management Guideline alongside it.
If you lend
Federally regulated lenders carry that same 1 May 2027 E-23 obligation, on top of B-15’s expectation of reliable physical-risk data down to the geophysical location of the exposure. B.C. credit unions answer to BCFSA rather than OSFI: the Loan Data Report has been mandatory since 1 June 2026 and reaches all applicable fields in June 2027, the Commercial Lending Guideline has required consideration of natural-catastrophe and climate risk since 19 August 2026, and BCFSA has said Natural Catastrophe Risk Guidance for credit unions is coming this fall.
1-in-250
wildfire and flood PML already in OSFI’s P&C Climate Risk Return
Fall 2026
BCFSA consults on natural-catastrophe and model-risk guidelines for insurers
12%
of deposit-taking institutions collected flood-insurance coverage on collateral
Jun 2027
BCFSA Loan Data Report Phase 2, all applicable fields
We do not produce a PML, a loss estimate or a price, and the interface has no field for your sums insured. What EMBER and CREST produce is the property-level hazard sitting underneath that exposure, reported on the same unit the return aggregates on, with the method attached. A province-wide average answers none of it. Neither does a score you cannot take apart in front of someone who asks.
What comes back
Two assessments for one address, delivered together and never combined.
No field averages them into a single grade. A fire index and a flood index answer different questions, and one number would hide which of them moved.
Wildfire
EMBER
Fuel measured around the building at 30 m, terrain, seventy-five years of nearby ignitions, potential fire behaviour under a declared severe scenario, and suppression context. Five weighted components, each published.
Fluvial flood
CREST
Two independent terrain estimates at fixed weights. A floodplain designation and recorded flood events raise a band and never lower one. Coastal, pluvial, flash and sewer-backup flooding are named as not assessed.
Across a book
One address answers a decision. A book answers accumulation.
Send a list of properties and the response carries the concentration view a single lookup structurally cannot: how the book distributes across bands, and where the same ground is carrying more than one peril at once.
Concentration by Forward Sortation Area
The unit OSFI aggregates physical-risk exposure on, computed for you rather than left as a roll-up to do afterward. Properties, band counts, and how many are elevated on both perils, per FSA.
A cross-peril matrix, not a blend
Fire band against flood band. A property severe on flood and minimal on fire is a different problem from one moderate on both, and a combined number renders them identical, so no field combines them and there is no total anywhere.
Your own grouping, echoed back
Send a branch, a segment, a vintage or a treaty as an opaque key and it is rolled up beside the FSA view. The engine never interprets it, which is why it works for whatever your book is actually organised by.
Loss refused in the response, not merely absent
No damage, expected loss, loss ratio or PML, at any return period. The refusal is a named field, so nobody has to notice something is missing to learn it was declined.
What an assessment looks like
The number comes apart. That is the point of it.
Every component is named, weighted and tagged with the resolution it was measured at, so the total re-derives by hand. Values are withheld on this page and are included in every assessment.
| Component | Resolution | Contribution |
|---|---|---|
| fuel_exposure | 30 m land cover · 250 m fuel typing | •••• |
| terrain | 5 km | •••• |
| ignition_history | point · 1950–2025 | •••• |
| potential_behaviour | declared severe scenario | •••• |
| suppression_context | 5 km | •••• |
Fire behaviour is computed under a declared severe reference scenario, not today’s weather and not a forecast.
methodology_fingerprint •••• · assessment_fingerprint ••••
Early access
Tell us what you would assess.
EMBER and CREST are in early access with a small number of partners. Tell us the book you would run and we will come back to you about it.