11 perils. 36 federal sources.
Full transparency on every call.

Regulatory flood zones (A, AE, AO, VE, X, etc.) from the official Flood Insurance Rate Maps. SFHA boundary joined per coordinate — the DOI-defensible baseline for flood classification.

Census-tract annualized loss, expected annual loss rates, and social vulnerability — integrated into the GFI composite and the NRI endpoint family.

885K-row CONUS dataset of precipitation return periods (2yr–1000yr) at point locations. Downsampled to noaa.atlas14_uw. Powers three dedicated precipitation endpoints.

2.7M geospatially indexed flood claims from 1978–2026. Powers the full loss-history endpoint family including radius analysis, SRL/repetitive loss flags, and tract-level summaries.

Letter of Map Amendment (LOMA), Revision (LOMR), and FIRM panel compliance data surfaced in the FEMA master report endpoint.

21M surge-model points covering the CONUS coastline (24.55°N–46.22°N) at Cat1–Cat5 intensity levels. Provides the FEMA zone-X gap flag — 8 dedicated endpoints including /surge/fema_gap.

30 years of all recorded severe weather events at county/point level — the authoritative federal record for wind, hail, tornado, and lightning frequency scoring.

NRI hurricane/wind annualized loss columns used to boost the GWHI score, especially for Gulf and Atlantic coastal markets.

Separate wind magnitude and frequency layers from the NRI dataset used for wind frequency scoring and nearest-event lookup.

2024 raster dataset at 270m resolution in EPSG:5070 (Albers Equal-Area Conic). Point-in-raster lookup delivers 1–5 WHP class at the coordinate level.

NRI wildfire annualized loss rate and building exposure columns power the financial comparison endpoint and contribute to the composite report.

36,665 polygons covering SRA (effective April 1, 2024) and LRA layers statewide. Returns hazard class (Very High / High / Moderate) and layer source (SRA/LRA) at the coordinate. Validated: ~83% of CAL FIRE DINS-recorded destroyed structures fell in the top-2 FHSZ classes — the strongest available signal for CA wildfire underwriting. California only.

30,154 historical wildfire perimeter polygons filtered from the nationwide NIFC dataset — spanning fires from 1878 through 2024. Returns in_burn_perimeter flag, matched fire names/years/acreage, burns within 5/10 miles, and most recent burn year. Validated against Camp Fire (153,335 acres), Thomas Fire (281,790 acres), and Dixie Fire (963,309 acres).

ASCE 7-22 design-level PGA, Ss, and S1 spectral accelerations from the USGS National Seismic Hazard Model. Returns 2%/50yr and 10%/50yr exceedance probabilities.

Shear-wave velocity at 30m depth used for NEHRP site class assignment (A–E). Amplifies or attenuates PGA estimates based on soil conditions at the coordinate.

Active fault traces from the authoritative USGS database. Returns nearest fault name, slip rate class, and distance in meters — factored into the SEI composite score.

NRI earthquake annualized loss rate and exposure columns, joined at the census-tract level for the SEI composite and the NRI master report.

All recorded hail events ≥0.75 inches from the 30-year Storm Events database. Frequency scoring and heatmap derived from event density per county.

NRI hail annualized loss rate used as a multiplier in the HHI composite, adding actuarial weight beyond raw event frequency.

30-year tornado record with F/EF scale, path geometry, and injury/fatality counts. Frequency scoring based on event density per county, boosted by NRI annualized loss weighting.

NRI tornado annualized loss rate (trnd_alr) used to weight the THI score — critical for accurate risk in Tornado Alley markets.

24-year lightning event record from NOAA Storm Events. Frequency scoring via county-level event density, with county benchmark and heatmap endpoints.

US Census Bureau TIGER/Line coastline shapefile — one of three reconciled coastline distance sources for the GCI composite.

NOAA Coastal Unit Shoreline Program dataset providing high-resolution shoreline geometries for coastline distance calculation.

OSM-derived coastline used as third reconciliation source. Three-source median distance reduces individual dataset edge cases.

Also powers the Coastal index — surge depth at Cat1–5 layered on top of distance scoring for coastal markets. 8 surge-specific endpoints.

112,214 polygons covering 52 states and territories — point-in-polygon lookup returns rock type (carbonate, evaporite, sandstone, volcanic pseudokarst, permafrost), karst classification, geologic age, and source map unit citation. Indicates geologic potential for karst features, not confirmed sinkholes — significant collapse hazards occur in only a small subset of mapped areas.

4,376 reported incidents statewide (1909–2025) with nearest-incident lookup plus 1/3/5/10-mile ring counts. Represents reported subsidence — most records are not field-verified as true karst sinkholes, and reporting carries population-center bias.

CONUS, Alaska, Hawaii, and Puerto Rico at 90m resolution. The quantile nonlinear (n10) model captures 98.9% of observed landslides while covering 44.6% of land area — the paper's recommended model for general use. Raw value (0–81) represents the count of susceptible 10m sub-cells within each 90m output cell. GridHazard classification: 0 = Not Susceptible, 1–20 = Low, 21–40 = Moderate, 41–60 = High, 61–81 = Very High. GeoTIFFs served from Azure Blob Storage via rasterio /vsicurl/ — no PostGIS raster storage required.

County-level susceptibility summary from the USGS data release — proportion of susceptible terrain and historical landslide density (landslides/km²) per county. 3,221 counties, 100% FIPS-matched. Returned as county context on every /hazard/landslide_susceptibility call.