Disaster Risk Explorer for Non-Metropolitan Counties

*Urbanization uses the USDA Rural-Urban Continuum Codes (RUCC, 2023 edition). Metro = codes 1–3 (counties in metropolitan statistical areas). Nonmetro = codes 4–9 (counties outside metro areas). 1 Counties in metro areas of 1 million population or more · 2 Counties in metro areas of 250,000 to 1 million population · 3 Counties in metro areas of fewer than 250,000 population · 4 Nonmetro, urban population of 20,000+, adjacent to a metro area · 5 Nonmetro, urban population of 20,000+, not adjacent to a metro area · 6 Nonmetro, urban population of 5,000–20,000, adjacent to a metro area · 7 Nonmetro, urban population of 5,000–20,000, not adjacent to a metro area · 8 Nonmetro, urban population under 5,000, adjacent to a metro area · 9 Nonmetro, urban population under 5,000, not adjacent to a metro area.

Data Sources

Natural Hazard Risk: FEMA National Risk Index (NRI), county-level dataset. Accessed via the ArcGIS Feature Service API. The NRI combines historical loss data, exposure estimates, and social vulnerability measures to produce composite risk scores for 18 natural hazard types across 3,232 U.S. counties and county-equivalents (50 states, D.C., and territories).
Rural-Urban Classification: USDA Economic Research Service Rural-Urban Continuum Codes (RUCC), 2023 edition (released January 2024). Based on the Office of Management and Budget (OMB) February 2023 metro area delineations and 2020 Census urban area definitions. Covers 3,233 counties.

Key Definitions

Expected Annual Loss (EAL): The average economic loss in dollars expected each year from a given hazard type, estimated by FEMA using historical frequency, probable magnitude, and spatial extent of past events. EAL combines three loss components: building value damage, population-equivalent losses (fatalities and injuries converted to dollar values), and agricultural losses (where applicable). The total EAL (EAL_VALT) sums across all 18 NRI hazard types.
Per Capita EAL: County-level EAL divided by county population (2020 Census). Expressed in dollars per person per year. This metric normalizes for county size and reveals disproportionate risk exposure in low-population counties that may not rank highly in absolute EAL.
NRI Risk Rating: FEMA’s composite classification of county-level risk. Combines Expected Annual Loss, Social Vulnerability (CDC/ATSDR SVI), and Community Resilience (U.S. Census Bureau). Five levels: Very High, Relatively High, Relatively Moderate, Relatively Low, Very Low.

Hazard Types

†All Hazards (EAL_VALT) sums Expected Annual Loss across all 18 NRI hazard types: avalanche, coastal flooding, cold wave, drought, earthquake, hail, heat wave, hurricane, ice storm, inland flooding, landslide, lightning, strong wind, tornado, tsunami, volcanic activity, and wildfire. The five hazards below are individually selectable in the dashboard as the largest national contributors to EAL.

Hurricane (HRCN_EALT): Tropical cyclone wind and storm surge losses. Includes all Atlantic and Gulf Coast hurricanes, tropical storms, and remnant systems. Does not include inland flooding from hurricane rainfall, which is captured separately under Inland Flooding.
Inland Flooding (IFLD_EALT): Riverine flooding, flash flooding, and urban flooding. Includes rainfall-driven flooding from all causes (frontal systems, tropical moisture, snowmelt). Excludes coastal surge flooding, which falls under Hurricane or Coastal Flooding.
Wildfire (WFIR_EALT): Uncontrolled fire in wildland or wildland-urban interface areas. Losses reflect building damage, population exposure, and agricultural loss from wildfire events.
Tornado (TRND_EALT): Violently rotating columns of air extending from thunderstorms to the ground. EAL estimates draw on historical tornado track data, including EF-scale intensity and path width/length.
Hail (HAIL_EALT): Precipitation in the form of ice with diameter ≥0.25 inches. Losses include crop damage, building damage (primarily roofs and siding), and vehicle damage. Hail events are sourced from NOAA Storm Events Database records.

Rural-Urban Continuum Codes (RUCC)

The 2023 RUCC classifies each U.S. county on a 1–9 scale based on metropolitan statistical area (MSA) status and degree of urbanization. The 2023 edition reflects the Census Bureau’s updated urban area threshold of 5,000 population (raised from 2,500 in prior editions).

Code	Classification	Category
1	Counties in metro areas of 1 million population or more	Metro
2	Counties in metro areas of 250,000 to 1 million population	Metro
3	Counties in metro areas of fewer than 250,000 population	Metro
4	Urban population of 20,000+, adjacent to a metro area	Nonmetro
5	Urban population of 20,000+, not adjacent to a metro area	Nonmetro
6	Urban population of 5,000–20,000, adjacent to a metro area	Nonmetro
7	Urban population of 5,000–20,000, not adjacent to a metro area	Nonmetro
8	Urban population under 5,000, adjacent to a metro area	Nonmetro
9	Urban population under 5,000, not adjacent to a metro area	Nonmetro

Methodology Notes

Map Color Scale: Logarithmic (d3.scaleSequentialLog). The extreme right-skew of EAL distributions (top 5% of counties account for ~50% of national EAL) makes a log scale necessary to reveal variation across the full range. The scale domain recomputes dynamically when filters change.
Population Data: County populations are from the 2020 Decennial Census as included in the NRI dataset (POPULATION field). Counties with zero population (e.g., certain Alaskan census areas) are excluded from per capita calculations.
County Geography: County boundaries are from the U.S. Census Bureau via the us-atlas TopoJSON dataset (10m resolution). Includes all 50 states and D.C. Connecticut uses planning regions rather than legacy counties, following the 2022 Census geographic update.
Filtering Interactions: Hazard, metric, component, and urbanization filters operate simultaneously. When a RUCC filter is active, the color scale, ranking, and summary statistics reflect only the filtered county subset. The metro/nonmetro share statistic always reflects the share of the full national total for the selected hazard and component. The state detail map uses an independent color scale rescaled to within-state variation.

Uncertainty & Limitations

The NRI does not provide confidence intervals or margins of error for EAL estimates. Instead, FEMA employs several methodological safeguards to manage uncertainty:

Bayesian Credibility Weighting: When a county lacks sufficient hazard events for a statistically reliable average, loss ratios are blended across four geographic levels (county, surrounding area, regional, national) to stabilize estimates.
Rare Event Modeling: High-consequence, low-frequency hazards (earthquake, hurricane, tsunami, volcanic activity) use probabilistic models that estimate event likelihood over extended return periods, then annualize the result. These estimates carry inherently greater uncertainty than hazards with frequent historical observations.
Rural/Urban Bias: Because EAL is driven by dollar-value exposure, urban counties with more buildings and higher property values tend to dominate total EAL rankings. A 2025 study found that the NRI may under-represent public health vulnerability in rural communities. The per capita metric and RUCC filters in this dashboard help address this by normalizing for population and allowing rural/urban comparisons.
Hazard Dependencies: The NRI does not account for correlations between hazard types. For example, hurricane and coastal flooding losses are estimated independently even though they frequently co-occur, which may lead to some double-counting or underestimation of compound risk.
Stationarity Assumption: EAL estimates are based on historical hazard patterns and assume past frequency and severity are indicative of future risk. Changing climate conditions, land use patterns, and population growth may cause actual future losses to diverge from these estimates.

EAL values should be interpreted as relative indicators of risk concentration rather than precise dollar forecasts. They are most useful for comparing risk across counties and identifying areas of disproportionate exposure.