Sixteen live.
One scale.

Every peril below is a calibrated 90 m global surface, modeled with the method that fits its physics. Every peril ships in Petrel Surfaces — a 0–1 calibrated score, a 5-class risk band, a per-pixel uncertainty layer, and an input-coverage mask. Per-return-period intensity rasters (Petrel Returns) are live for hurricane, windstorm, extreme heat, extreme cold, drought, storm surge, wildfire, and hail; stochastic event sets for the remaining extreme-value perils are on the roadmap.

How to get it: a free ~5°×5° sample per peril is on the public CDN today (no signup). The hosted point-query API is live and self-serve, metered through Stripe — get a free API key. Full bulk vintage downloads are sold under enterprise contracts — contact sales. Per-hazard model cards land with the public docs portal in 2026.Q3.

Live

Karakoram — sharp ridge & valley structure pops at 90 m

LNDv1.1 · 2026-04

Landslide

Machine learning · XGBoost · spatially blocked CV · AUC 0.97

Machine-learning susceptibility trained on 80,000+ documented landslides worldwide - terrain, geology, soil, and rainfall.

What ships

Surfaces — calibrated 0–1 probability per pixel.

Inputs: DEM slope/aspect, TWI, landcover, rainfall, NASA-GLC inventory (80K events)

Landslide hazard page →

French & Italian Alps — Chamonix corridor at peak

AVLv1.1 · 2026-04

Avalanche

Machine learning · XGBoost · spatially blocked CV · AUC 0.99

Machine-learning susceptibility from terrain and snow-climate, trained on observed avalanche events.

What ships

Surfaces — calibrated 0–1 probability per pixel.

Inputs: DEM slope/aspect, mean snowfall, winter temperature, terrain ruggedness

Avalanche hazard page →

Western US — Sierra Nevada, Southern California chaparral and the Great Basin at 90 m

WLFv1.0 · 2026-09

Wildfire

Physics · Cell2Fire fire-spread simulations on a CC-BY global fuel map, emulated to 90 m, anchored to 23 years of NASA MODIS burned area (2001–2023) with a per-cell local history anchor and calibrated so the global level matches the 2014–2023 mean

Physics-based wildfire likelihood: fire-spread simulations emulated to 90 m and anchored to 23 years of MODIS burned area. Tested against independent fire inventories — MTBS (US) AUC 0.88, NBAC (Canada) 0.92, EFFIS (Europe) 0.82 — with pixel-level AUC 0.95 against 20 years of 500 m MODIS burn history and rank agreement 0.52 with the USFS FSim model. Burn probability is relative to MODIS-detectable burned area, so small fires are under-detected; built-up land scores 0 by construction — use the exposure band for addresses; placement inside a 0.25° cell has modest skill (within-cell AUC 0.61); and it is a long-run annual probability, not a post-fire state — a just-burned area is not lowered. The downloadable rasters carry the ANNUAL probability; the API /score endpoint returns the 50-year probability by default (?horizon=50|100|250|500, or ?band=susceptibility for the annual value) — same product, two bases, so compare like with like. Refreshed annually against the burned-area record, cut each spring once the prior year of burned-area files is complete.

What ships

Surfaces — six 90 m global bands on land 60°S–75°N: annual burn probability per pixel (0–1; built-up, water, bare and ice read 0), exposure (highest annual burn probability within 2 km — what an address in a town beside wildland sees), a global five-class cut and a region-relative one, uncertainty, and coverage (non-burnable / burnable / cropland). Returns — probability of at least one burn within 50 / 100 / 250 / 500 years (rp50–rp500, 0–1).

Inputs: ESA WorldCover 2021 + Petrel fuel crosswalk v1.4 (Scott & Burgan FM40 classes), Petrel Ignition Layer v0.3, ERA5 fire-season weather climatology, Copernicus GLO-30 terrain, NASA MODIS MCD64CMQ burned area 2001–2023 (anchor), Cell2Fire simulation campaign (28,498 fires)

Wildfire hazard page →

San Francisco Bay to Salinas — high-PGA corridor along the San Andreas

EQKv1.1.1 · 2026-07

Earthquake

Physics · GEM PGA × Vs30 site amplification

Ground-shaking exposure from the GEM global 475-year seismic hazard model, refined for local soil amplification.

What ships

Surfaces — 0–1 susceptibility derived from the 475-year GEM PGA modulated by Vs30 amplification. Full-resolution (GSHHG-f) coastline: barrier islands now scored.

Inputs: GEM 2023 seismic hazard, USGS Vs30, soil-class amplification factors

Earthquake hazard page →

North Sea & UK — extratropical storm belt, coast-to-inland gradient

WNDv1.1.1 · 2026-07

Windstorm

Extreme value · GEV on ERA5 monthly fg10 maxima

The 100-year extreme wind gust, from 30 years of ERA5 climate reanalysis.

What ships

Surfaces + Returns — 0–1 score derived from the 100-year gust, plus per-return-period peak gust (rp10/25/50/100/200, m/s). Full-resolution (GSHHG-f) coastline: barrier islands now scored.

Inputs: ERA5 monthly gust extremes 1991–2020

Windstorm hazard page →

Indo-Gangetic plain to the Tibetan plateau — elevation lapse in action

HEATv1.1.1 · 2026-07

Extreme heat

Extreme value · GEV on ERA5 mx2t + elevation lapse downscale

The 100-year extreme high temperature, from 30 years of ERA5 climate reanalysis, downscaled to 90 m.

What ships

Surfaces + Returns — 0–1 score derived from the 100-year Tmax, plus per-return-period annual-max Tmax (rp5/10/25/50/100, °C). Full-resolution (GSHHG-f) coastline: barrier islands now scored.

Inputs: ERA5 monthly max temperature 1991–2020, DEM for elevation lapse correction

Extreme heat hazard page →

Russia / Mongolia / China — Siberian extreme to milder steppe

COLDv1.1.1 · 2026-07

Extreme cold

Extreme value · GEV on inverted ERA5 t2m + elevation lapse downscale

The 100-year extreme low temperature, from 30 years of ERA5 climate reanalysis, downscaled to 90 m.

What ships

Surfaces + Returns — 0–1 score derived from the 100-year Tmin, plus per-return-period annual-min Tmin (rp5/10/25/50/100, °C). Full-resolution (GSHHG-f) coastline: barrier islands now scored.

Inputs: ERA5 monthly min temperature 1991–2020, DEM

Extreme cold hazard page →

Europe — Alpine altitude band against the Mediterranean

FRZv1.1.1 · 2026-07

Frost / freeze

Climatology · Monthly frost-month climatology

How often freezing conditions occur through the year, from 30 years of ERA5 climate reanalysis.

What ships

Surfaces — 0–1 susceptibility derived from frost-month frequency per year. Full-resolution (GSHHG-f) coastline: barrier islands now scored.

Inputs: ERA5 monthly t2m 1991–2020

Frost / freeze hazard page →

Lake Maracaibo, Venezuela — densest lightning on the planet

LTNv2.0 · 2026-09

Lightning

Observational · NASA LIS/OTD/ISS flash density × ECDF normalization

Observed total flash density from NASA satellite optical detection (LIS/OTD/ISS), 1995–2023.

What ships

Surfaces — 0–1 susceptibility derived from annual TOTAL flash density (intra-cloud and cloud-to-ground, flashes / km² / yr) on the NASA LIS/OTD/ISS satellite record. v2.0 changes the measured quantity, not just the calibration: v1.x counted cloud-to-ground only, so absolute densities and class boundaries both shift and v1.x values must be re-derived, not rescaled. A score of 0.000 means no flash was detected there across the record — measured, not missing.

Inputs: NASA LIS/OTD/ISS satellite flash climatology

Lightning hazard page →

Loess Plateau, China — classic erosion landscape, hill-by-hill detail

EROv1.1.1 · 2026-07

Soil erosion

Process model · RUSLE (R · K · LS · C · P)

Long-term annual soil loss from the RUSLE model - rainfall, soil, slope, and land cover. Ice, rock, and open water are left unscored.

What ships

Surfaces — 0–1 susceptibility derived from annual erosion rate (t/ha/yr). Full-resolution (GSHHG-f) coastline: narrow coasts now scored where soil exists.

Inputs: ERA5 rainfall erosivity, SoilGrids K-factor, DEM slope-length, Copernicus Global Land Cover C-factor

Soil erosion hazard page →

US East Coast — high coastal gust exposure fading inland, Florida to New England

HURv1.1 · 2026-07

Hurricane

Physics · Event-driven: STORM v4 + CLIMADA Holland H08

Tropical-cyclone peak-gust exposure from 10,000 years of synthetic storm tracks across all six ocean basins.

What ships

Surfaces + Returns — 0–1 susceptibility and per-return-period peak gust (rp50/100/250/500, m/s).

Inputs: STORM v4 10,000-yr synthetic tracks × 6 basins (Bloemendaal 2020, CC0), CLIMADA TropCyclone wind-field renderer, Natural Earth land mask

Hurricane hazard page →

SPEI-12 rp100 drought severity · California + SW USA, 90 m

DRTv2.1 · 2026-07

Drought

Extreme value · L-moment GEV, region-of-influence RFA (~400-cell climate-similarity pools)

Drought-severity (SPEI-12) susceptibility and return levels from 85 years of ERA5 climate reanalysis, 1940-2024. Polar and hyperarid cells are left unscored.

What ships

Surfaces + Returns — 0–1 drought susceptibility (N2-normalized rp100 |SPEI-12|, v2.1) plus per-return-period |SPEI-12| magnitudes (rp5/10/25/50/100, σ), region-of-influence regional fit.

Inputs: ERA5 monthly precipitation + 2m temperature 1940-2024, Beck 2018 Köppen-Geiger 1km, Thornthwaite 1948 PET, climate_indices Pearson III SPEI-12

Drought hazard page →

Naples — Vesuvius & Campi Flegrei: per-phenomenon exposure collage

VLCv1.0 · 2026-08

Volcanic Eruption

Physics · Per-volcano physical hazard footprints (lahar / lava / PDC / tephra / collapse) × completeness-corrected magnitude-frequency eruption rates

Long-horizon volcanic exposure: per-volcano physical hazard footprints combined with magnitude-frequency eruption rates into 50-year exceedance probabilities. A screening climatology, not an eruption forecast.

What ships

Surfaces — P(≥1 eruption phenomenon reaches the pixel within 50 years) at 50 / 250 / 500-year horizons, plus a five-way per-phenomenon breakdown (lahar, lava, PDC, tephra, collapse). A real exceedance probability, not a tuned score.

Inputs: Petrel Volcano Catalog (1,608 volcanoes, VEI + eruption rates), Copernicus 90 m DEM, ERA5 winds (tephra dispersal)

Volcanic Eruption hazard page →

Sendai / Sanriku — the Tohoku 2011 validation coast

TSUv1.1 · 2026-09

Tsunami

Physics · PTHA18 offshore probabilistic hazard reuse, propagated to 90 m with coastal amplification and overland attenuation; regional source branches (NEAMTHM18, GAR15) plus non-seismic sources

Coastal tsunami screening: offshore probabilistic hazard reused from PTHA18 and propagated to 90 m. A screening climatology for portfolio triage — explicitly not evacuation planning, life-safety zoning, or engineering design.

What ships

Surfaces — P(inundation ≥ 0.5 m) at 50 / 250 / 500-year horizons (500-year is the default planning basis), plus 500-year inundation depth in metres, a five-class zoning cut, uncertainty and coverage. Scored on ocean-connected land below 50 m elevation and within 30 km of the shoreline — the inland edge follows terrain rather than a fixed line, and land below sea level (polders, subsided plains) is scored rather than excluded.

Inputs: PTHA18 + GAR15 (Geoscience Australia), NEAMTHM18 (INGV), NGI, GEBCO 2024 bathymetry, SRTM15+, Slab2, NOAA NCEI historical run-up database

Tsunami hazard page →

US Gulf Coast — the Katrina 2005 calibration coast

SRGv1.0 · 2026-09

Storm Surge

Physics · GTSM hydrodynamic storm-tide return levels (COAST-RP) on a bare-earth elevation composite (GEDTM30 v1.2 + DeltaDTM v1.1 over Copernicus); tropical from the STORM 10,000-year synthetic catalog, extratropical from ERA5, de-duplicated at event level against IBTrACS

Coastal storm-surge screening from a hydrodynamic solve, so wind direction, fetch, basin geometry and tide-surge interaction are inside the numbers rather than approximated. Still-water only — no wave setup or runup — on a frozen 1980-2017 climatology with no sea-level rise, and flood defences are not modelled, so below-datum land reads as undefended. Screening for portfolio triage, not engineering design.

What ships

Surfaces — P(inundation) at 50 / 250 / 500-year horizons (500-year default) with a five-class cut, uncertainty and coverage. Returns — still-water inundation depth in metres at rp50/100/250/500, on the same return-period grid as Hurricane so wind and surge combine return-for-return.

Inputs: COAST-RP storm-tide return levels (Dullaart 2021), DeltaDTM v1.1 bare-earth elevation (Pronk 2024, from Copernicus WorldDEM-30), STORM synthetic cyclone catalog, ERA5, IBTrACS

Storm Surge hazard page →

North American Plains — 100-year MESH hail size (mm)

HALv1.0 · 2026-10

Hail

Physics · NASA PMHAILCLIM satellite hail-signature climatology (TRMM/GPM passive microwave) sets where hail forms; one global function of the ERA5 environment (freezing level, elevation, 700–500 hPa lapse rate, 0–6 km shear), fitted once on US ground radar (NOAA MRMS MESH) and six Australian Bureau of Meteorology radars, sets how often that becomes MESH ≥ 25 mm; one measured constant sets the severity tail — position, region and continent are never inputs

Global hail screening on a 0.05° (~5 km) grid carrying ~28 km of information, with a measured transfer error of ×/÷1.65 on level and ×/÷1.93 on severity at every pixel. Sizes are MESH — the ground-radar estimate of the largest hailstone aloft, not a measured stone on the ground (above ~45 mm the typical reported ground stone is smaller). Tested against radar only in the US and Australia; the coverage flag reads 0.5 where the function extrapolates (deep tropics, high plateaus), and cells filled from observed convection are untested against radar. The 50-year probability saturates near 1 across the great hail regions (US Plains, Argentina, South Asia, central Africa) — use the Returns line to rank them. No detail below 0.25° is claimed; nodata beyond ±68° latitude, where the satellite anchor does not observe.

What ships

Surfaces — P(radar-defined hail, MESH ≥ 25 mm, at a point within 50 years) with a five-class cut, uncertainty and a coverage flag (1.0 inside the fitted range, 0.5 where the function extrapolates). Returns — MESH hail size in millimetres at rp10/25/50/100/250/500 with rp100 low/high bands, plus annual exceedance probabilities for MESH ≥ 25 / 51 / 76 mm; 0 means below 25 mm, not no hail.

Inputs: NASA GHRC PMHAILCLIM (Bang & Cecil 2019), NOAA MRMS MESH 2020–2026, Bureau of Meteorology radar volumes 2010–2025, ERA5, lightning + ERA5 convective environment for sparse satellite cells

Hail hazard page →

Free layersPETREL LAYERS

Global context rasters that feed the hazard models — free to browse and download with attribution. Not risk scores; the raw ingredients.

On the roadmap

Nine hazards to round out the 25-peril catalog. Designed and queued items have a config and methodology doc in source; untouched items are scoped but not started.

See pricing →