Get FEMA National Risk Index scores — get_national_risk

Downloads the complete FEMA National Risk Index (NRI) for every US census tract or county. The NRI characterizes a community's relative risk from 18 natural hazards by combining expected annual loss, social vulnerability, and community resilience. All fields published by the NRI feature service are returned, with the source's abbreviated field codes expanded into descriptive snake_case column names (see Details).

Usage

get_national_risk_index(geography = c("tract", "county"), cache_path = NULL)

Arguments

geography: The geographic summary level of the results. One of "tract" (the default) or "county".
cache_path: Optional path to a .parquet file used as a read-through cache. If supplied and the file already exists, data are read from it and no download occurs. If supplied and the file does not exist, freshly downloaded data are written there for reuse. If NULL (the default), data are downloaded fresh and not written to disk. Because the NRI is periodically revised (see Details), delete a stale cache file to force a refresh.

Value

A tibble with one row per census tract or county. Every field published by the NRI service is returned, with the source's abbreviated field codes expanded into descriptive snake_case names (see Details). Columns fall into these families:

Identifiers: geoid (a standardized, zero-padded 11-digit tract or 5-digit county FIPS join key, prepended by this function), the state_name and county_name labels, and the source version nri_version. The NRI's redundant identifier fields (nri_id, the individual state and county FIPS codes, and tractfips) are dropped in favor of geoid.
Community totals: population, value_building (building value, dollars), value_agriculture (agricultural value, dollars), and area_sq_mi (square miles).
Composite index families: risk_* (overall National Risk Index), estimated_annual_loss_* and expected_annual_loss_rate_composite_* (expected annual loss and its annualized rate), social_vulnerability_index_* (social vulnerability), resilience_* (community resilience), and community_risk_factor_value. Each family carries some combination of _score/_state_percentile/_national_percentile (percentile), _rating (rating), and _value/_value_* (absolute) columns.
Per-hazard metrics: One block per hazard, prefixed by the hazard name: avalanche, coastal_flood, cold_wave, drought, earthquake, hail, heat_wave, hurricane, ice_storm, landslide, lightning, inland_flood, severe_wind, tornado, tsunami, volcano, wildfire, and winter_weather. Within each block, suffixes denote the metric: _event_count/_annual_frequency (event count and annualized frequency), _exposure_* and _exposed_area (exposure), _historic_loss_ratio_* (historic loss ratio), _estimated_annual_loss_* (expected annual loss), _expected_annual_loss_rate_* (annualized loss rate), and _risk_value/ _risk_score/_risk_rating (hazard risk value, score, and rating).

Coastal or otherwise geographically limited hazards are NA for communities with no exposure.

Details

Data are pulled from FEMA's National Risk Index ArcGIS feature services (one service per geography) and paginated 2,000 records at a time. After janitor::clean_names() normalization, the NRI's abbreviated field codes are expanded into readable snake_case names (for example, WFIR_EALT becomes wildfire_estimated_annual_loss_total, and EAL_SPCTL becomes estimated_annual_loss_state_percentile); the ArcGIS service artifacts (OBJECTID, Shape__Area, Shape__Length) and redundant source identifier columns are dropped, and a standardized geoid join key is prepended.

Interpreting the fields depends on the suffix:

*_score, *_state_percentile, and *_national_percentile columns are values from 0 to 100: *_score and *_national_percentile rank a community against all US communities of the same type (all tracts, or all counties), while *_state_percentile ranks it against communities in the same state. Because these are ranks rather than additive quantities, they cannot be summed or averaged across geographies; to describe risk for an area spanning several tracts, request geography = "county" (or higher) directly rather than aggregating tract scores.
*_rating columns are categorical labels (e.g. "Relatively High").
Value and loss columns (*_estimated_annual_loss_*, *_exposure_*, *_exposed_area, *_event_count, and the community totals value_building, value_agriculture, population, and area_sq_mi) are absolute quantities (dollars, counts, or areas) and are additive across geographies.
Rate and ratio columns (*_annual_frequency, *_historic_loss_ratio_*, and *_expected_annual_loss_rate_*) are normalized rates rather than absolute quantities and, like the percentile columns, cannot be summed across geographies.

Field definitions are documented in FEMA's NRI Technical Documentation and data dictionary at https://hazards.fema.gov/nri/data-resources (which uses the original abbreviated field codes). Hazard coverage reflects NRI v1.20 (December 2025), in which inland flooding (the inland_flood_* columns) replaced the earlier riverine flooding hazard; the source version is carried in the nri_version column. If a future NRI release changes the schema, the underlying query surfaces the service error rather than silently truncating results.

Examples

if (FALSE) { # \dontrun{
# county-level scores, downloaded fresh
nri_counties <- get_national_risk_index(geography = "county")

# tract-level scores, cached locally for reuse across sessions
nri_tracts <- get_national_risk_index(
  geography = "tract",
  cache_path = file.path(tempdir(), "nri_tracts.parquet"))
} # }