IntelliWhere — Behind the Score

01 The Score

How Risk Is Scored Across Four Perils Using Open Government Data.

IntelliWhere scores every Canadian postal code area on flood, hail, wildfire, and winter storm exposure. Then combines them into a single composite using Simple Additive Weighting (SAW) with Analytic Hierarchy Process (AHP)-derived weights. The steps below walk through the method exactly, including a live calculation for FSA T9H.

1

We add risk factors together

Each FSA gets scored on four climate threats: flood, hail, wildfire, and winter storm. Those four scores are combined into one composite number through Simple Additive Weighting (SAW), a method that multiplies each score by its importance and adds them up.

2

Each threat is given a different weight

Not all perils cost insurers equally. Flooding drives far more insured losses across Canada than winter storms, so it carries more influence in the final score. The weights are: Flood 38% · Hail 27% · Wildfire 23% · Winter Storm 12%.

3

Weights were calculated, not guessed

Using the Analytic Hierarchy Process (AHP), a structured method that derives weights by comparing every pair of threats head-to-head. Is flood worse than hail? By how much? Those comparisons are converted mathematically into the final percentages.

4

Weights were validated against real loss data

The AHP judgments were cross-checked against 17 years of IBC/CatIQ insured loss data (2005–2022, national aggregate, all lines). If flooding consistently drives the largest share of claims in Canada, it should carry the most weight, and our model confirms it does.

5

The weight matrix passes a formal consistency test

AHP utilizes a self-check called the Saaty Consistency Ratio (CR). It verifies that our comparisons don't contradict each other. If flood outranks hail and hail outranks wildfire, then flood must also outrank wildfire. Our CR = 0.017, below the 0.10 acceptance threshold.

The weights at a glance

Flood

38%

Hail

27%

Wildfire

23%

Winter Storm

12%

Weight vs. Canadian insured loss share — IBC / CatIQ 2005–2022 · Four-peril share (flood / hail / wildfire / winter)

Our AHP weight IBC/CatIQ loss share

Our weights align with loss shares for flood and hail. The wildfire weight (23%) sits below its raw loss share due to single-event concentration — the 2016 Fort McMurray wildfire alone represents ~25% of the four-peril total, making any share derived from this window an event-sample artefact as much as a climatological signal. The winter storm weight (12%) sits above its 2005–2022 share, reflecting tail risk evidenced by the 1998 ice storm outside the derivation window.

          Source: IBC / CatIQ national insured loss data · All lines of business · National aggregate · Derivation period: 2005–2022 · Loss shares are shares of four-peril losses only (excludes wind-primary events and earthquake) · Derivation: docs/ibc_loss_share_derivation.md.
          Note: all-lines aggregate may understate hail's property-specific loss share relative to personal property lines.
        

Here is how IntelliWhere scores FSA T9H — Fort McMurray, Alberta — one of the highest-risk FSAs in Canada.

Step 1 — Raw peril scores (from government data)

Flood	1.79
Hail	2.49
Wildfire	4.00	maximum possible score
Winter Storm	1.70

What drives each score

Flood 1.79 AB FHIP floodplain coverage limited for T9H boundary; post-1980 housing stock reduces urban vulnerability layer

Hail  2.49 ICHD v1.1, 2005–2024: 15 events recorded; v2 percentile renorm anchors score against national p1–p99 distribution

Wildfire 4.00 2016 Horse River Fire (~590,000 ha, 2,400+ structures); CWFIS fire density at maximum; FWI at data ceiling (23.44)

Winter 1.70 ECCC 30-yr climate normals: cold baseline, no dominant ice-storm or freezing-rain event concentration

Step 2 — Multiply each score by its AHP weight

Flood	1.79	×	0.38	=	0.680
Hail	2.49	×	0.27	=	0.672
Wildfire	4.00	×	0.23	=	0.920	Dominant peril — wildfire at maximum severity
Winter Storm	1.70	×	0.12	=	0.204

Step 3 — Sum the weighted scores

0.680 + 0.672 + 0.920 + 0.204 = 2.476 (inputs shown rounded to 2 decimal places; composite computed and stored at full precision: 2.4776)

Step 4 — Composite score

2.4776

FSA T9H · Fort McMurray, AB · live ca_composite_risk_scores

MODERATE

Arithmetic: 1.75 ≤ score < 2.50

Regional Peril Lens

AB + wildfire ≥ 3.0
→ Critical

See Section 03 below →

        Score = 0.38 × Flood + 0.27 × Hail + 0.23 × Wildfire + 0.12 × Winter Storm  ·  Output: 1.00 – 4.00  ·  CR = 0.017
      

        Consistency Ratio 0.017 — below the 0.10
        acceptance threshold.
      

        Score uncertainty. Each composite carries a propagated ±0.03 uncertainty at 1σ (Taylor first-order, independent perils). FSAs where the flood score is based on the housing-age proxy rather than floodplain geometry carry wider true uncertainty — the proxy introduces systematic error that geometry-based scoring does not. Score confidence intervals are on the roadmap for PDF reports as a Phase 1 quick win.
      

02 The Four Perils

These four climate threats were selected based on Canadian insured loss history. Together they account for the majority of property catastrophe losses across Canada.

38% weight — highest impact

Flood

The single largest driver of insured catastrophe losses in Canada. Overland flooding, riverine flood events, and urban stormwater overwhelm are all captured in our flood scoring layer.

27% weight — severe property damage

Hail

Canada's most frequent cause of billion-dollar insured loss events. Calgary alone has experienced multiple catastrophic hail events causing billions in claims. Scored from ICHD v1.1 point observations (2005–2024, 8,457 verified events).

23% weight — growing exposure

Wildfire

The 2016 Fort McMurray fire remains Canada's costliest insured disaster. As wildland-urban interface expands, wildfire risk is increasingly relevant to property underwriters. Scored from 52 years of CWFIS fire perimeter data (1972–2024).

12% weight — consistent baseline risk

Winter Storm

Ice storms, freezing rain events, and heavy snowfall cause structural damage, roof collapses, and business interruption across Canada's populated corridors. Scored from 30-year ECCC climate normals.

03 The Regional Peril Lens

Regional Cluster Weights — Catastrophe Exposure Emerges From the Arithmetic.

The Regional Peril Lens is now a weighting layer, not an override. Each FSA is scored with an AHP weight set derived from its region's own catastrophe-loss history, plus a published non-compensatory term that lets a single catastrophe-corridor peril lift the composite. Critical is reached through the arithmetic alone — no post-hoc province trigger.

Five regional clusters

National weights treat a Calgary hail FSA and a Montréal ice FSA identically. Instead, every FSA is assigned to one of five clusters — Prairie (AB/SK/MB), Pacific (BC), Central (ON/QC), Atlantic, North — each carrying its own AHP weight set decomposed from that region's share of 2005–2022 Canadian insured catastrophe losses. Prairie weights tilt toward wildfire and hail; Central toward flood and winter; Atlantic toward winter. Every cluster matrix has Consistency Ratio CR < 0.10. Weights are shrunk 50% toward the national set to keep overall calibration stable.

The non-compensatory term

          SAW = Σ cluster_weight[peril] × peril_score

          cat_max = max(peril_score for the cluster's catastrophe-driver perils)

          composite = (1−λ)·SAW + λ·cat_max  if  cat_max ≥ 3.0  (λ = 0.45)

          composite = SAW  otherwise

The max() operator is non-compensatory: only a cluster's documented catastrophe perils (regional loss share ≥ 15%) can enter it, and only above a 3.0 gate. A single catastrophe-corridor peril lifts the composite without a province trigger; low-risk FSAs are untouched.

Worked illustration — T9H (Fort McMurray, AB)

          Cluster: Prairie  ·  Wildfire 4.000 (max, a Prairie catastrophe driver)  ·  SAW = 2.671

          cat_max = 4.000 ≥ 3.0 → composite = 0.55·2.671 + 0.45·4.000 = 3.1734

          Percentile-anchored zone: Elevated — reached through arithmetic, no override.

        Composite zones are percentile-anchored to the national distribution (Low ≤40th · Moderate 40–80th · Elevated 80–95th · High · Critical ≥3.4), giving 662 / 674 / 249 / 68 / 14 FSAs respectively (Task 2.4, June 2026; zone cuts re-derived July 2026). The 14 Critical FSAs include Manitoba/Saskatchewan prairie multi-peril (12 FSAs, wildfire- and hail-dominant) and Ontario/Quebec flood corridors (2 FSAs) — and T2P (Calgary NW) via the documented hail-Critical eligibility rule (hail_score = 4.0, 86 ICHD events). Critical is reached through genuine composite arithmetic and a data-driven corridor rule; the binary province-trigger override has been fully retired. Verified against live ca_composite_risk_scores (2026-07).
      

04 The Data

Built From Canadian Government Data — With Honest Coverage Disclosure.

Authoritative datasets sourced directly from Canadian federal and provincial agencies, processed geographically across all 1,667 postal code areas. Hover any source to see the dataset detail.

        Flood data coverage. ~304 FSAs have floodplain geometry from four sources: AB FHIP, TRCA regulated areas, Metro Vancouver floodplains, and Quebec BDZI (added June 2026, 174 FSAs). Of these, 53 have sufficient coverage to drive the flood score directly (HIGH band); 251 have partial geometry below the score-determining threshold (PARTIAL band). The remaining 1,363 of 1,667 FSAs score entirely on the urban-vulnerability proxy (pre-1980 housing age as a stormwater and sewer backup indicator) — coverage gap, not low risk. Scores will be revised upward as national floodplain data is ingested. NRCan national flood COG is a Phase 0 Data Priority P1 item on the roadmap.
      

Federal — Environment and Climate Change Canada

Severe Weather & Climate Station Records

Two datasets from Canada's national weather observation network. Integrated Canadian Hail Database v1.1 (2005–2024, Zenodo) powers hail scoring — 8,457 point observations from ECCC-affiliated stations and trained observers mapped to FSA boundaries. Thirty-year climate normals power winter storm scoring — freezing rain days and ice storm frequency across Canada's populated corridors.

Hail scoring — ICHD v1.1 · 2005–2024 · Zenodo Winter storm — 30-yr climate normals

ECCC

Environment and Climate

Change Canada

Open Government Licence — Canada

Statistics Canada

Federal — Statistics Canada

FSA Boundaries & Housing Age

2021 National Census — FSA digital boundary file and dwelling construction decade by postal code. Used to map urban flood vulnerability across all Canadian postal areas.

Flood vulnerability proxy

All 1,667 Canadian FSAs

Provincial — Conservation Authority

Floodplain Mapping — Greater Toronto

Regulated floodplain polygons along the Don, Humber, and Rouge River watersheds. Spatially intersected with FSA boundaries to determine flood zone coverage per postal area.

Greater Toronto Area (L-prefix FSAs)

Government of Alberta

Provincial — Government of Alberta

Flood Hazard Identification Program

FHIP floodway and flood fringe polygons mapped along major Alberta river systems. Provides authoritative flood zone boundaries for Calgary and surrounding FSAs.

Flood hazard zones · Tier A

Alberta (T-prefix FSAs)

Regional — Metro Vancouver

Fraser River Floodplain Mapping

Floodplain polygons covering the Fraser River delta and low-lying coastal areas across Metro Vancouver. Critical for assessing flood exposure in BC's lower mainland FSAs.

Metro Vancouver (V-prefix FSAs)

Natural Resources Canada

Federal — Natural Resources Canada

Canadian Wildland Fire Information System

National Fire Database (NFDB) — large fire perimeter polygons ≥200 ha. 16,340 fires, 1972–2024. Used to calculate historical wildfire density per FSA across all provinces.

Wildfire history · Tier A

All provinces and territories · 1972–2024

Insurance Bureau of Canada · CatIQ

Canadian Catastrophe Loss Data

17-year insured loss proportions (2005–2022) used to validate that peril weights reflect actual Canadian insurance market losses — not modelled assumptions.

Weight calibration only

National · Not a spatial input

05 Backtest — Six Canadian Catastrophe Events

How did the model perform against real events?

Six benchmark Canadian insured-loss events. For each we score the primary FSA(s) using post-renormalization (v2) scores and assess whether the model would have flagged the risk before the event. Scores are on the 1.0–4.0 IW scale. Composite risk zones are percentile-anchored to the national distribution: Low ≤40th · Moderate 40–80th · Elevated 80–95th · High p95–3.40 · Critical ≥3.40 (criterion-anchored). Critical is reached through the regional-cluster composite arithmetic alone — there is no post-arithmetic province override. A backtest with zero misses reads as fabricated — the misses below are the mechanism.

Event 01

Southern Alberta Flood — June 2013

~$1.7–1.9B insured loss (IBC) · Bow and Elbow river corridor, High River

HIT T2A MISS T0L

T2A — Calgary Bow/Elbow

                Flood: 3.497  |  Composite: 2.096 Moderate

                Flood score correctly High (3.497). Composite stays Moderate because flood weight is 38% — single-peril model limitation. A pre-2013 underwriter using IntelliWhere would have seen High flood exposure in T2A.

T0L — High River / Foothills

                Flood: 1.034  |  Composite: 2.357 Moderate

                Honest miss. High River (T0L) scores near-floor flood because AB FHIP data covers urban Calgary corridors but not the Highwood River floodplain. Unmapped area reads as low-risk — the exact failure mode Phase 1 Task 1.3 addresses with a data-confidence flag.

Event 02

Fort McMurray Wildfire — May 2016

$3.58B insured loss (CatIQ nominal final) · Canada's costliest insured disaster to date

HIT T9H T9J T9K

T9H / T9J / T9K — Fort McMurray

              Wildfire: 4.000 (maximum) for all three FSAs. Cluster composite: T9H 3.1734 Elevated · T9J 3.031 Elevated · T9K 3.013 Elevated.

              Under national-only weights wildfire's 23% share leaves these FSAs Moderate. The Prairie cluster weighting plus the non-compensatory term lift them to Elevated through the arithmetic — wildfire is a Prairie catastrophe driver, so its 4.0 score enters the composite directly. No post-hoc override is applied.

Event 03

Eastern Ice Storm — January 1998

~$1.4B+ insured loss (IBC) · Montérégie QC + Eastern Ontario corridor

HIT QC + ON

J3B/J3G — Montérégie QC

                J3B — Winter: 3.034 (Elevated) · Flood: 4.000 (proxy-only) · Composite: 2.503 Elevated

                J3G — Winter: 3.034 (Elevated) · Flood: 3.895 (proxy-only) · Composite: 2.769 Elevated

                Hit on WINTER: ECCC 30-yr freezing rain and ice storm normals are high for the Montérégie corridor — the winter score correctly flags the 1998 ice storm zone. Note: J3B/J3G flood scores reflect the housing-age proxy only (no floodplain geometry exists for QC in the current model). An ice storm is not a flood event; the flood score is not the relevant signal here.

K0A / K0B — Eastern Ontario

                K0A — Hail: 3.336 · Winter: 3.307 · Composite: 2.655 Elevated

                K0B — Hail: 3.546 · Winter: 3.230 · Composite: 2.730 Elevated

                Both correctly Elevated. Winter score (3.2–3.3) reflects ECCC freezing-rain climatology for this corridor.

Event 04

BC Atmospheric River Floods — November 2021

~$0.5B+ insured / ~$9B+ total damages · Abbotsford, Merritt, Princeton

MISS V2S MISS V3G PARTIAL V0K

V2S / V3G — Abbotsford

                V2S — Flood: 1.894 · Composite: 1.552 Low

                V3G — Flood: 1.571 · Composite: 1.368 Low

                Honest miss. The 2021 Sumas Prairie flood inundated some of Canada's most productive agricultural land. IntelliWhere has no BC lowland floodplain polygons — the Fraser Valley is entirely unscored for flood. Unmapped = reads as low-risk. Remediation: NRCan national flood COG + BC Ministry floodplain mapping (Phase 0 Data Priority P1/P2).

V0K — Merritt

                Wildfire: 3.871 · Flood: 1.530 · Composite: 2.507 Elevated

                Partial. IntelliWhere correctly flags Merritt as Elevated because of wildfire exposure — Merritt sits in BC's Interior wildfire corridor. The flood component of the 2021 event (Nicola River) is missed for the same data-coverage reason as Abbotsford.

Event 05

Calgary Hailstorm — August 2024

$3.25B insured loss (CatIQ final) · Costliest hail event in Canadian history

HIT T2P MISS T1Y

T2P — Calgary NW (hail corridor)

                Hail: 4.000 (maximum) · 86 events, 23.0mm mean diameter · Arithmetic composite: 3.1055 Elevated · Zone: Critical (hail-Critical corridor rule)

                T2P's arithmetic composite (3.1055) falls in the Elevated band; single-peril hail dominance is compensated by lower flood/wildfire/winter scores. The documented hail-Critical eligibility rule (Task 2.4) escalates T2P to Critical: Calgary-CMA prefix (T2*), hail_score = 4.0, 86 ICHD events ≥ 50-event threshold. The Phase 1.4 percentile remap (HP99=2.8188→4.0) placed T2P at hail saturation — the strongest single-peril hail signal in the national dataset — providing the evidence base for the rule. See Section 6 of the methodology note.

T1Y — Calgary NE (storm epicenter)

                Hail: 1.000 (floor) · Composite: 1.651 Low

                Honest miss. ICHD v1.1 contains zero observer reports within T1Y's 15.82 km² boundary for 2005–2024. The Aug 2024 storm track targeted this FSA specifically. Zero events reflects observer coverage gaps in newer NE Calgary suburban development, not absence of hail risk. Remediation: Phase 1 Task 1.4 (population-density offset) + Northern Hail Project integration.

Event 06

Jasper Wildfire — July 2024

~$1.2B insured loss (CatIQ) · 30% of Jasper townsite destroyed

NEAR-MISS T0E

T0E — Jasper / Upper Athabasca

              Wildfire: 3.561 · Hail: 3.549 · Composite: 2.375 Moderate.

              T0E's wildfire score (3.561) reflects CWFIS burn history and FWI climatology in the upper Athabasca corridor; 93 ICHD hail events also make it multi-hazard. However, the compensatory AHP weighting pulls the composite to 2.375 — T0E's low flood and winter scores offset the strong wildfire and hail signals. Known limitation: compensatory weighting is the correct default for portfolio aggregation but understates single-peril-dominant tail events like this one. The retired province-trigger Lens previously escalated T0E to Critical; that override is now removed because it was non-reproducible from data alone. T0E's honest zone is Moderate; correctly identifying it as Critical requires a single-peril concentration rule analogous to the Calgary hail rule — a documented Phase 2 work item. T0E's burn history includes the Jasper 2024 fire (CWFIS data through 2024 database vintage).

Summary

Event	FSA	v2 Zone	With Lens	Assessment
2013 AB Flood	T2A	Moderate	Moderate	Hit — flood exposure flagged
2013 AB Flood	T0L	Moderate	Moderate	Miss — flood data gap (FHIP covers urban AB only)
2016 Fort McMurray	T9H/T9J/T9K	Moderate	Critical	Hit — wildfire at max; Lens correctly escalates
1998 Ice Storm	J3B/J3G/K0A/K0B	Elevated	Elevated	Hit — winter corridor correctly Elevated
2021 BC Floods	V2S/V3G	Low	Low	Miss — no BC lowland floodplain data (P1 remediation)
2024 Calgary Hail	T2P	Elevated	Elevated	Hit — hail corridor at max (86 events, 23mm)
2024 Calgary Hail	T1Y	Low	Low	Miss — ICHD observer coverage gap in NE Calgary
2024 Jasper Wildfire	T0E	Moderate	Critical	Hit — wildfire Elevated; Lens escalates to Critical

        Scores are v2 (post-renormalization). Renormalization maps national p1→1.0 and p99→4.0 per peril, ensuring full 1–4 range use. Validated against 17 years of IBC/CatIQ insured loss data (national aggregate, all lines). Note: all-lines aggregate may understate hail's property-specific loss share — flagged for refinement with property-only CatIQ data. This note will be updated as additional flood data is ingested.
      

        Known limitation — hail severity in sparse-observation FSAs (partially addressed, Phase 1.4). Of 1,667 FSAs, approximately 1,130 (68%) have zero recorded hail events in ICHD v1.1 (2005–2024) and score at the 1.0 floor — correctly indicating no observation-supported hail exposure rather than substituting a national average. For FSAs that do have observations, Phase 1.4 (June 2026) introduced credibility weighting: the diameter component's 0.55 weight is earned linearly over the first 10 events, preventing a single large-diameter report from driving the score on par with a sustained 86-event corridor. An FSA with 0–3 events still reaches only a diameter-capped floor score. Closing the observation gap itself (radar-derived hail swaths, ICLR/ECCC gridded hail climatology) remains a Phase 2 data work item.
      

06 The Team

Ryan Nguyen

B.Eng Geomatics Engineering — York University (Lassonde) · Founder, IntelliWhere

Ryan Nguyen is a Geomatics Engineer with field experience on Ontario's largest transit infrastructure projects: Ontario Line, Eglinton Crosstown West, Hurontario LRT, GO Expansion. IntelliWhere applies the same spatial precision used in billion-dollar infrastructure to Canadian property climate risk.

View LinkedIn