How Calchis Computes Calchis Hazard Scores

Hazard scoring is the foundation of catastrophe modeling. Before you can estimate losses, you need to know what physical intensity a location will experience from a given peril. Calchis computes Calchis Hazard Scores on a 0–10 scale for six US perils: earthquake, hurricane, wildfire, flood, hail, and tornado.

Every score is fully decomposed. We show the components, the weights, and the source data. This is a deliberate choice.

Why Transparency Matters

The incumbent cat modeling platforms — AIR (Verisk), RMS (Moody's), and CoreLogic — treat their scoring methodologies as proprietary black boxes. This made sense in 1990 when data was scarce and modeling was a genuine competitive moat. It makes less sense in 2026 when the underlying data is freely available from federal agencies and the peer-reviewed literature is indexed and searchable.

More importantly, insurance professionals need to defend scores to their compliance teams, regulators, and reinsurers. A score that says "7.8 out of 10" without showing how it got there is useless in a regulatory filing. A score that says "7.8 based on: historical CAT3+ track frequency, NOAA SLOSH surge zone classification, and coastal proximity — with each component and its data source shown" is defensible.

Scoring Architecture

Each peril score is the sum of weighted components. Components are derived from three data layers:

Historical frequency. How often has this peril affected this location in the observational record? For hurricanes, we use IBTrACS (1842–present, North Atlantic basin) to count CAT3+ tracks near each location. For earthquakes, we use the USGS earthquake catalog with Gutenberg-Richter frequency analysis. For wildfires, NIFC perimeters (1984–present) provide fire frequency by region.

Hazard zone classification. Federal agencies publish authoritative hazard zone maps. FEMA flood zones (A, AE, V, X) classify flood risk. NOAA SLOSH model outputs classify storm surge exposure by hurricane category. USGS National Seismic Hazard Maps provide probabilistic ground motion estimates. These zones are the bedrock of hazard assessment — they represent decades of scientific work and billions of dollars in modeling investment.

Physical proximity and exposure. Distance to fault lines (USGS Quaternary Fault Database), distance to coastline (NOAA National Shoreline), distance to wildland-urban interface (University of Wisconsin SILVIS Lab WUI data). These are geometric calculations against authoritative federal datasets.

Climate Adjustment

Historical frequency alone is insufficient because the climate is changing. Hurricane intensification rates in the Atlantic basin have increased measurably since 1980. Wildfire seasons in the western US are longer and more severe than the historical average. Flood return periods computed from 20th-century data underestimate 21st-century flood risk in many US watersheds.

That is the argument for adjustment. Here is what we have actually built against it, which is less: nothing, today. Until September 2026 the California wildfire score carried a 1.4× multiplier on large-fire frequency attributed to Abatzoglou & Williams (2016). The paper reports burned area roughly doubling; 1.4× was our inference from it, not a number the paper states, and the loss model never applied it. We retired it. Every peril is now scored on the unadjusted historical record, which understates a warming trend, and each score says so with a climate_adjusted flag that is false.

We would rather publish the gap than imply a grid that is mostly empty. The mechanism is real and it generalizes: each factor lives in the model parameters database with a citation to the source paper, and when a newer paper updates an estimate the factor is versioned — superseded, not deleted, so the full parameter history is preserved. What is missing is not the machinery but the per-peril, per-region literature review to fill it. Every score returns a climate_adjusted flag, so you can tell which case you are reading without taking our word for it.

What Hazard Scores Are Not

Hazard scores are not risk scores. Risk requires exposure data — what is the replacement value of the structure at this location? What construction type? What year built? Until property-level exposure data is integrated (via ATTOM or comparable sources), Calchis labels all scores as "Calchis Hazard Scores" and includes a disclaimer: "Property exposure data not included. Not a substitute for professional actuarial assessment."

This distinction matters to actuaries. Conflating hazard and risk is a credibility-destroying mistake that we refuse to make.