Multi-peril hazard API

Ten natural-hazard scores.
One API call.

Petrel scores any point on Earth’s land for landslide, avalanche, earthquake, windstorm, extreme heat, extreme cold, frost, lightning, soil erosion, and hurricane — each a calibrated 0–1 peril score on the same 90 m global grid. Present-day hazard, documented methods, no sales call.

Get a free API key — 1,000 calls/mo, no card Try it on an address

One request. Every peril. Anywhere on land.

GET /v1/score
$ curl -H "X-API-Key: $PETREL_KEY" "https://api.petreldata.io/v1/score\
?lon=7.6586&lat=45.9763&hazards=landslide,avalanche,earthquake,windstorm,\
extreme_heat,extreme_cold,frost_freeze,lightning,soil_erosion,hurricane"

{
  "lon": 7.6586,
  "lat": 45.9763,
  "scores": {
    "landslide": 0.9757225257118443,
    "avalanche": 0.9992828150273141,
    "earthquake": 0.8820001831110569,
    "windstorm": 0.9669942320017091,
    "extreme_heat": 0.19100009155552844,
    "extreme_cold": 0.06599627674184393,
    "frost_freeze": 0.8019959105197302,
    "lightning": 0.597994933927427,
    "soil_erosion": 0.0,
    "hurricane": 0.0
  }
}
Response captured live from the API, 2026-07-07 — Matterhorn massif, Zermatt (45.9763 N, 7.6586 E). Values verbatim.
Resolution
90 m native
Coverage
Global land — one grid
CRS
EPSG:4326
Format
COG + STAC + REST

Most hazard APIs stop at one country or one peril. Petrel is the same catalog everywhere — the grid in Osaka is the grid in Oslo and Ohio.

Four endpoints. The whole surface.

GET/v1/score
Every live peril for one point, in one call.
Query params lon, lat, and an optional hazards filter. Returns a calibrated 0–1 score per peril, all on one scale.
POST/v1/score/batch
Portfolio screening — up to 10,000 points per call.
JSON body of {id, lon, lat} records; one row per point with every requested peril inline.
GET/v1/return_level
Hurricane peak gust by return period.
rp50 / rp100 / rp250 / rp500 — peak gust in m/s at 90 m, from 10,000 years of synthetic tracks across six basins.
GET/v1/catalog
STAC catalog with COG asset URLs for every release.
Browse vintages and peril versions; compatible with pystac-client. Assets served from data.petreldata.io.

See how every score is built.

Every peril ships with its model, inputs, and validation documented — per hazard, per release.

Landslide

XGBoost on an 80,000-event global inventory, spatially blocked cross-validation — AUC 0.97.

Methodology →

Earthquake

GEM 2023 global seismic hazard (475-year PGA) modulated by Vs30 site amplification.

Methodology →

Windstorm

The 100-year gust from GEV fits on ERA5 reanalysis extremes, 1991–2020.

Methodology →

Method, inputs, and validation for all ten live perils →

Public pricing. No quote call.

Self-serve checkout, metered per location lookup — a batch call of N points counts as N. Above Scale, talk to us about enterprise.

Free
$0 /mo
1,000 calls / mo
No card required
Indie
$49 /mo
10,000 calls / mo
$0.008 per extra call
Growth
$199 /mo
50,000 calls / mo
$0.006 per extra call
Scale
$749 /mo
250,000 calls / mo
$0.004 per extra call

For raster teams

Every release is a Cloud-Optimized GeoTIFF with a STAC item. Free sample COGs for every live peril — pull with curl, open in QGIS, no signup.

Browse free samples

The honest FAQ.

Do you cover flood?

No — there is no live flood product today. What is live: landslide, avalanche, earthquake, windstorm, extreme heat, extreme cold, frost/freeze, lightning, soil erosion, and hurricane.

Is this a cat model?

No. Petrel ships hazard scores — plus hurricane return-period peak gusts — not loss curves or event sets.

Do you model climate scenarios?

No. Scores are present-day hazard, built from recent climatology and observations. If you need 2050 scenario projections, we are not that.

What's the demo limit?

The public demo at petreldata.io/try is rate-limited per day, no signup. A free API key lifts you to the full 1,000 calls/mo quota.

Ten peril scores. One key. Start free.

Get a free API keyScore an address first