# Data attribution

This data is derived from the National Flood Hazard Layer (NFHL) of the
Federal Emergency Management Agency (FEMA), U.S. Department of Homeland
Security.

  Source:       https://msc.fema.gov/portal/home (FEMA Flood Map Service Center)
                https://hazards.fema.gov/femaportal/NFHL/searchResult
  Product:      NFHL, layer S_FLD_HAZ_AR (flood hazard areas), the
                county-wide deliveries FEMA lists
  Producer:     Federal Emergency Management Agency (FEMA)
  Licence:      U.S. federal government work, public domain. No restriction
                on use or redistribution.
  Attribution:  FEMA National Flood Hazard Layer

# Not for official use

These files are a repackaging, not a FEMA product. They are not the
effective Flood Insurance Rate Map and must not be used for flood zone
determinations, insurance rating, regulatory or legal purposes: for those,
use the FEMA Flood Map Service Center. FEMA has not reviewed, approved or
endorsed these files. FEMA updates the NFHL continuously: these files are
checked against FEMA's list every day, and a delivery FEMA republished is
replaced (last check: 2026-10-03). Each row carries the date of its delivery
(`fema_update_date`, from 2002-03-04 to 2026-10-02); changes.json lists what each
day replaced.

# Coverage

2,502 county-wide deliveries in 51 states and territories: every
county-wide dataset FEMA lists that holds a flood hazard layer
(skipped.json names those that hold none). Counties FEMA has not mapped
digitally are absent, as are the community-level deliveries (a town or
city published on its own). An absent area is unmapped here, not free of
flood hazard.

# What was changed

  1. Pieces. Flood zone polygons reach hundreds of thousands of vertices.
     Each one is cut into pieces of at most 100 vertices (ST_Subdivide), so
     a point lookup tests a small polygon: 5,808,014 zones make 95,572,036
     pieces. A row is a piece, not a zone: `piece_id` numbers the pieces
     of a zone from 0, so `piece_id = 0` counts zones. `source_feature_id`
     is FEMA's FLD_AR_ID, which FEMA does not keep unique inside every
     delivery: 117,301 zones in 391 deliveries share theirs with
     another zone. Elsewhere (`dfirm_id`, `source_feature_id`) identifies
     the zone and dissolving on it gives the original outline back.
     Invalid source geometries were made valid first.
  2. Columns. `flood_zone` (FLD_ZONE), `zone_subtype` (ZONE_SUBTY), `sfha`
     (SFHA_TF), `static_bfe` (STATIC_BFE, NULL where FEMA writes -9999),
     `dual_zone` (DUAL_ZONE), `dfirm_id` and `source_feature_id`
     (FLD_AR_ID) are FEMA's values. `state` (postal code), `county` and
     `fema_update_date` come from the delivery. Three columns are not part
     of the NFHL, they are a reading of FEMA's zone and subtype added by
     Geomermaids: `risk` (1 percent flood zone, 0.2 percent flood zone,
     minimal, undetermined, water, unmapped), `floodplain` (the same as
     `sfha`) and `subzone` (the zone of a dual or AR zone).
  3. Layout. One file per delivery, state=<XX>/<DFIRM_ID>.parquet, sorted
     along a Hilbert curve in row groups of 20,480 pieces with a
     covering bbox column. counties.parquet indexes them: filter it on
     bbox or state, then read the files it names.

Coordinates are FEMA's, NAD83 (EPSG:4269), unchanged.

The pipeline that builds these files is public:
  https://github.com/gsueur/nfhl-geoparquet-workshop

This data is packaged and hosted by Geomermaids:
  https://geoparquet.geomermaids.com/
  contact: gsueur@geomermaids.com