Where to Live for Generational Survivability: US County Map

How the scores are calculated

The common method

Every factor works the same way. We take real, published numbers for each county, put all US counties in order from worst to best, and split the list into five equal groups. The best fifth of counties gets a 5, the next fifth a 4, and so on down to 1. So a score is always relative to the rest of the country: a 3 means "middle of the pack nationally," not "moderate in absolute terms."

Ranking by position (rather than fixed cut-off values) keeps the five groups the same size even when the raw data has lots of ties. If a county is missing data for a factor, it gets a neutral 3 for that factor, and its detail panel says so.

Where a factor uses several measures (Water and Natural hazards), each county is ranked on each measure separately, those rankings are averaged with equal weight, and the average is split into fifths.

The combined score is the weighted average of the four 1–5 scores (Water 25%, Soil 20%, Natural hazards 40%, Practical 15%), stretched so the lowest county in the US is 0 and the highest is 100. The map colours split counties into national fifths on that combined score.

Water security (25%)

Source: WRI Aqueduct 4.0 (World Resources Institute, 2023), via Esri's Living Atlas.

How: three equally weighted measures:

1. Water depletion today: how much of the area's renewable water is used up each year (consumed and not returned), on WRI's 0–5 scale. 0 means under 5% is used up; 5 means over 75%, or a desert area with almost no water to use.
2. Water depletion by 2080: the same measure under WRI's high-emissions scenario (SSP5-8.5, matching the hazard projections), which accounts for changes in both rainfall and demand.
3. Groundwater trend: how fast the water table is falling, in centimetres per year (for example, the Ogallala aquifer under the High Plains). Trends WRI rates as statistically insignificant count as zero.

Aqueduct is mapped by river basin, not county, so each county's value is the average over about 100 evenly spaced points inside its boundary (an area-weighted average).

Why depletion and not "water stress": Aqueduct's better-known water-stress score counts all water withdrawn, including power-plant cooling water that goes straight back into the river. That makes the humid Midwest look as stressed as Atlanta (Madison, WI scores 3.7 of 5). Depletion counts only water actually used up, which separates the genuinely dry West from places with plenty of water.

Limits: Aqueduct does not cover Hawaii or the territories (they get a neutral 3). It measures the regional water balance, not a single well, water rights, or water quality.

Soil & food (20%)

Sources: USDA 2022 Census of Agriculture, plus projected crop yields from Hsiang et al. 2017, Science (Climate Impact Lab).

How: two equally weighted measures:

1. Farmland today: cropland acres as a share of the county's land area (land area measured from the county boundary itself, spot-checked against published figures to within about 1%).
2. Yields by 2080–2099: the projected change in yields of the four major US crops (corn, soybeans, wheat, cotton) under high emissions, from the peer-reviewed county-level study. Heat hits the southern and central Corn Belt hardest (central Iowa about −40%), while the northern Plains and Northwest gain (eastern Washington and Montana about +25%). Counties that don't grow these crops use the farmland measure only.

Limits: cropland share shows land actually being farmed, not a soil test, and irrigated desert farmland counts the same as rain-fed prairie. The yield projections cover four field crops, not vegetables, orchards, pasture or livestock, and assume no new adaptation such as different crop varieties.

Natural hazards (40%)

Sources: FEMA National Risk Index (current); FEMA's National Risk Index: Future Risk (published December 2024, later taken offline by FEMA; used here from the CC BY 4.0 mirror in Esri's Federal User Community); NOAA's LOCA2 county climate projections used in the Fifth National Climate Assessment; and the USDA Forest Service's Wildfire Risk to Communities (April 2026 release).

How: four equally weighted measures, each sized to the county or household so big counties are not penalised just for being big:

1. Risk today: FEMA's own national percentile of expected yearly losses as a share of what is exposed, across 18 hazards (floods, wildfire, tornadoes, earthquakes, hurricanes, hail, winter storms and more), adjusted for how vulnerable and how resilient the community is. 0 is the lowest-risk county in the US, 100 the highest.
2. Losses by ~2085: start from FEMA's dollar estimate of today's yearly losses for each hazard. For the five hazards FEMA projected (coastal flooding, drought, hurricanes, wildfire, extreme heat), grow them by FEMA's late-century, higher-emissions multiplier (2× = double). Hold the other 13 at today's level, since no projection exists. Divide by everything exposed (buildings, crops, and residents, which FEMA values at $11.6 million each). Shown as dollars lost per year for every $1 million exposed.
3. Extreme heat in the 2080s: projected days a year at or above 95°F and nights a year that stay at or above 75°F (when homes and bodies can't cool down), under high emissions. FEMA's dollar figures barely count heat, so it is measured directly. Not available for Alaska and Hawaii, which use the other three measures.
4. Wildfire risk to homes: the Forest Service's national percentile for how likely a typical home in the county is to be exposed to and damaged by wildfire. Spreading FEMA's wildfire dollars over everything in a county hides this risk in places like Bend, Oregon and Boulder, Colorado.

A county that is risky now and getting worse scores lowest; one that is safe now and stays safe scores highest.

Why not FEMA's headline risk score: FEMA's best-known county score counts total dollars at risk, so it rises with population and property values. Dane County, WI ranks in FEMA's top 10% that way, but in the safest 10% once losses are measured against what is exposed.

Limits: only some hazards have future projections, so where tornadoes, river floods or earthquakes dominate, the future part is mostly today's risk carried forward. Heat is measured as air temperature; humid heat (heat index or wet-bulb temperature), which is most dangerous in the Southeast and Ohio Valley, isn't available county by county, so those areas may look somewhat safer here than they are.

How this compares with other studies

We checked these scores against two independent county-by-county climate assessments that cover the lower 48 states. Numbers are rank correlations: 1.0 would mean identical rankings, 0 no relationship.

*Our Soil & food factor uses Hsiang's crop-yield projections, so the combined-score comparison with that study is not fully independent. Before heat and wildfire were added, our hazard score agreed at 0.56 and 0.43.

Agreement is strong but not total, and the differences are mostly deliberate:

• Mountain West counties (Colorado's Front Range and high country, northern New Mexico) score worse here, because we count earthquakes, landslides and wildfire risk to homes and Front Range water depletion. Both studies cover climate-driven risks only.
• The Ohio Valley and parts of the Southeast score better here, because the other studies weigh humid heat, which we can't yet measure county by county.
• Water supply barely appears in either study, so our Water factor adds something they don't cover rather than repeating them.

The top-ranked areas here (the lower Great Lakes: Michigan, Ohio, western and central New York, northern Indiana, Pennsylvania, plus Vermont) match the places climate researchers most often name as relative "climate havens." Buffalo, Burlington and Ann Arbor all land in the top 10% of counties.

Practical (15%)

Sources: USDA ERS Rural-Urban Continuum Codes (2023), plus the EPA Superfund National Priorities List.

How: the USDA code runs from 1 (part of a metro area of 1 million or more) to 9 (rural, small towns, far from any metro). We treat the middle as the sweet spot: code 5 (non-metro, 20,000+ urban residents, not next to a metro area) scores best, and each step toward either big-city or very remote costs points. Each Superfund site in the county takes off a little more (the penalty grows more slowly after the first few). Counties are then ranked into fifths.

Limits: the "middle is best" rule is a judgment call: it stands in for affordable land that still has hospitals, stores and jobs within reach. It is not a measure of actual land prices. Superfund sites were matched to counties by name (about 95% matched; most misses are in Connecticut, which replaced its counties with planning regions in 2022).

Other data

• County & state boundaries: us-atlas (US Census TIGER/Line, simplified).
• Cities: Plotly's us-cities-top-1k dataset (US Census population figures). A city shows its county's scores.
• Basemap: map tiles © Esri (World Light Gray), HERE, Garmin; map data © OpenStreetMap contributors.

Built from a one-time, periodically refreshed data pull, not a live feed. US only for now; Canada is not yet included.