The data

Three ways in.
Same physics.

Petrel data ships in three shapes. The free CDN sample lets you evaluate without signing up. The hosted API — metered through Stripe — is the right fit for single-address and portfolio queries. Bulk-download every vintage by contacting sales for enterprise access. Same underlying physics, three delivery shapes.

Live · free · no auth

Sample COGs · public CDN

One ~5°×5° crop per live peril, hosted publicly at data.petreldata.io. Pull with curl or open straight in QGIS / rasterio / GDAL — no signup, no API key. Right for evaluating the data shape, calibration, and integration patterns before you commit.

# Pull the Karakoram landslide sample:
curl -O https://data.petreldata.io/samples/landslide_karakoram_90m.tif

# Sample one point with rasterio (works against any sample COG):
import rasterio
with rasterio.open("landslide_karakoram_90m.tif") as src:
    score = next(src.sample([(76.513, 35.862)]))[0]
Self-serve · metered

Point-query API · single addresses + portfolios

Hosted REST endpoint for single addresses, parcels, or small polygons. JSON in, JSON out. Authenticated by API key, metered through Stripe — pay only for what you query. Designed for underwriting submission triage, due diligence, and dashboard integrations.

# Score one location (real response, Karakoram test point):
curl -H "X-API-Key: tlk.…" \
     "https://api.petreldata.io/v1/score?lon=76.513&lat=35.862"
# → {"scores": {"earthquake": 0.967, "windstorm": 0.944,
#               "frost_freeze": 1.0, "landslide": 0.094, ...}}
Sales · bulk licensing

Bulk data licenses — regional or global

License the rasters themselves: full-region cuts (the entire US, all perils) or global COGs per peril per vintage — every susceptibility / classification / uncertainty / coverage layer in one bundle, for transformation and integration inside your own stack. Delivered via signed S3 access or a private mirror in your own AWS account. Right for cat modelers, large portfolios, and customers with their own GIS stack.

# What a bundled vintage looks like once provisioned:
aws s3 sync s3://petrel-data/vintages/2026/landslide_v1.0/ ./landslide/
# → landslide_susceptibility_90m.tif, _classification_90m.tif,
#   _coverage_90m.tif, _stac_item.json, README.md, previews/...

API endpoints

The hosted point-query API is live and self-serve — authenticated by API key, metered through Stripe. Sign up free for 1,000 lookups a month, no card; upgrade in-app when you outgrow it. Every endpoint below is documented in full — schemas, real response examples, and a try-it-in-browser console — in the interactive reference. SDKs (Python, TypeScript, Go, R) are on the roadmap.

Open the interactive API reference

GET/v1/score
Single-point Surfaces score across every peril.
Query params: lon, lat, hazards (optional, default = all live). One calibrated 0–1 score per requested peril; null where a location has no data (ocean, deliberately unscored terrain).
POST/v1/score/batch
Score up to 10,000 points in one call.
JSON body of {id, lon, lat} records. Returns one row per input with all requested hazards inline.
GET/v1/return_level
Return-period intensity at a point.
The physical magnitude of the T-year event: drought SPEI-12 severity (σ) at rp5–rp100, hurricane peak gust (m/s) at rp50–rp500.
GET/v1/catalog
STAC catalog of every release.
Browse vintages, peril versions, units, and asset locations — discover versions programmatically instead of hardcoding product names.
GET/v1/layer/…
Free Layers: list, metadata, download.
Fuel, aridity, and ignition structural layers (CC-BY-NC-4.0): list vintages, fetch STAC metadata, and pull full bundles via presigned URLs.

Pricing

The point-query API prices by location lookups, metered through Stripe — pay monthly for what you actually queried, with a free tier for evaluation. Bulk vintage downloads are sold by data-volume + subscription tier under enterprise contracts (contact sales). Free CDN samples stay free, forever, no signup.

See pricing — every tier, public →

Start free — no cardBrowse free samples