Where you could actually live.
Most survival maps are a picture of where bombs land. This one opens on a different question: heating load, hours of daylight at the winter solstice, frost-free days, solar resource, and whether enough people live nearby to sustain a dentist. Then it lets you test that place against ten catastrophes. Several of the best refuges here are miserable places to live, and the model says so out loud.
Strategic nuclear exchange
Grade ARanked refuges
Score / 100Score your own location
Type a US ZIP code, pick a world city, or drop coordinates. ZIP codes resolve on your device against a US Census file that ships with this page, so nothing is looked up online. The model runs the same physics against your actual position: distance to the nearest counterforce target, downwind fallout corridor, latitude-adjusted crop failure, and how long the nearest metro's population takes to reach you on foot.
What actually keeps you alive
How each number is produced
A product in this category earns trust by showing its work and marking the places where it is guessing. Every scenario carries an evidence grade.
| Grade | Meaning | Scenarios | Basis |
|---|---|---|---|
| A | Modeled from published physics or epidemiology | Nuclear exchange, Metro strike, Natural pandemic, Carrington event, Supervolcano, Climate breakdown | Glasstone & Dolan blast scaling, Xia et al. soot-injection famine curves, Carrington-class GIC latitude bands, VEI-8 ash dispersal |
| B | Inferred from a historical analogue | Engineered bioweapon, Grid collapse / EMP | 1918 and 2020 transmission data extrapolated; Ukraine 2015 and Texas 2021 grid failures scaled up |
| C | Speculative. Entertainment, not planning. | AI takeover, Alien invasion | No empirical basis exists. Scored on generic obscurity, network disconnection and overburden. Labeled as such everywhere it appears. |
Where every number comes from
The full technical specification is in the method paper. This is the short version, in the order people actually ask.
Is this a prediction?
No. Every score is conditional: given that an event happens at a stated severity, what happens to a prepared household at that location. The atlas says nothing about how likely any event is, and any product that claims to know that is guessing.
What is measured, what is computed, what is judged?
Measured: site depths, capacities, coordinates and access, from public facility records. Computed: blast and fallout radii, protection factors, daylight hours, heating and cooling degree days, growing season, solar resource, population pressure and distance to the coast. Judged: six factor scores per site (water, food, off-grid capability, stability, maritime buffering, air sealing), assigned by hand on a 0 to 100 scale. The judged ones are the weakest part of the model and the method paper says so.
Where do the blast numbers come from?
Glasstone and Dolan, The Effects of Nuclear Weapons, the standard unclassified reference. Overpressure radii scale with the cube root of yield; thermal scales with yield to the 0.41 power because the atmosphere absorbs some of it. A 15 kt device gives a 1.70 km radius of severe blast damage, which is smaller than most people assume and is why sheltering beats evacuating.
Where does the famine modeling come from?
Xia and colleagues, Nature Food, 2022. They modeled six soot-injection scenarios and their effect on crops, fisheries and livestock. A 150 teragram US-Russia exchange cuts global calories by around 90% in years three and four. Australia and New Zealand came out best, which is why the southern hemisphere dominates the nuclear panel here.
Why is the climate panel more pessimistic than others I have seen?
Because most maps still use the 35 C wet-bulb figure from a 2010 theoretical paper. Penn State later put real people in a climate chamber and found that young, healthy adults lose thermal compensation at roughly 31 C wet-bulb in humid conditions. Four degrees lower moves the uninhabitable band a long way poleward. This atlas uses 31.
How is livability calculated differently?
The threat panels multiply, because they are gates you either pass or do not: blast, then fallout, then famine, then order. Livability averages, because thermal comfort and daylight are qualities of a place rather than gates. But any single dimension below 25 caps the total, because one fatal flaw is not averaged away by seven good ones.
Where do heating load and daylight come from?
Daylight is exact solar geometry: the sun's declination at the solstice against your latitude, no data needed. Heating and cooling degree days are modeled from published zonal mean temperatures, adjusted for elevation at 5 C per thousand meters, distance from the coast, and maritime warming above 50 degrees latitude. They are estimates, not station records, and will be off by 10 to 20 percent for any specific address.
Why does the model want some people nearby?
Because a dentist, a welder, a school and a buyer for what you make all require a population, and a household cannot supply any of them. The threat panels penalise density. The livability panel wants a middle band. That disagreement is real and it is the most useful thing in the product.
What about the alien and AI panels?
They carry Grade C and say "speculative, entertainment, not planning" in their own first line. There is no dataset. They exist because people ask for them and because labeling them honestly is what makes the Grade A label mean something.
What is it worst at?
Four things. Survival stages are treated as independent when they are correlated, which biases composites optimistic. Site factor scores are expert judgment, not data. Fallout uses a single wind vector rather than a real wind field. And the grid is 1.25 degrees, so a cell center can sit 70 km from the point you actually care about. The method paper ranks all of these by how much they would change an answer.
Do you make money if I buy something?
Where a link is paid it says so at the link, every time. No partner can influence a site score or a scenario weight, because the scoring is the only thing here worth anything. If the best answer is a product nobody pays us for, that is the one shown.
What should I actually do first?
Water storage and a filter, then a way to heat one room without the grid, then ninety days of food you already eat. That order holds in nine of the ten scenarios here and in the four-day power cut that is far likelier than any of them. Nothing on this page should move you toward a bunker before you own a survey meter and a full water butt.