How the model works
Fallout Forecast answers one question: if radioactive dust were thrown into the air over a city right now, where would today's weather take it? Here is everything that goes into the answer, and where the answer stops being reliable.
In one paragraph
We release about 24,000 computer particles into the cloud an explosion would leave, and move each one forward a minute at a time for twelve hours using NOAA's forecast. Each particle is pushed by the wind at its own height, shaken by turbulence, moved by rising or sinking air, and pulled down by its own weight. Near the ground it sheds some of its dust; in rain it sheds more. We add up what lands and what passes at breathing height on a grid, turn that into radiation dose, and draw it.
The weather
The forecast is NOAA's Global Forecast System, the same one behind most weather apps. A new one comes out four times a day; we load it within minutes and run again every hour. From it we use, for every hour and every 16 miles or so across the lower 48 states:
| What | Why it matters |
|---|---|
| Wind at 12 heights | From 33 feet to 9,000 feet. Wind turns and strengthens with height, so the top of a cloud and the bottom go different ways. |
| Friction velocity | How hard the ground is stirring the air. Sets the strength of turbulence. |
| Surface heat flux | Whether the sun is heating the ground (air churns, dust mixes upward) or the ground is cooling (air goes still, dust stays in a thin layer). |
| Mixing height | The lid on the churning layer. Dust below it spreads through it; dust above it stays aloft. |
| Vertical motion | Large-scale rising and sinking air. |
| Rain rate | Rain pulls dust out of the air and puts it on the ground, sometimes far more than anything else does. |
| Roughness and terrain | Rough ground slows the wind near the surface. Heights are measured from the ground, not sea level. |
The cloud
We do not model a device. We model the cloud a blast leaves behind, using a standard relation between the size of a charge and how high its cloud rises, and the layering used by Lawrence Livermore's HotSpot code, a tool emergency planners use. The ten classes of the ER scale differ in two things only: how high the cloud goes, and how much caesium-137 is in it.
ER-1 is simulated with about 2 curies and a cloud top of 200 feet. Each class up holds about three times as much, to 20,000 curies at ER-9. ER-10, which has no upper end, is simulated at 6,000,000 curies and a cloud top of 2,950 feet to show the extreme. ER-7 uses 2,300 curies, the figure in the federal government's planning scenario. The scale page lists every class and shows each one simulated.
We assume all of it becomes airborne dust. That is deliberately pessimistic. In tests, much of the material in an explosion comes out as fragments that fall within a few hundred feet. Real contamination would very likely cover less ground than we show.
The particles
- Sizes. From half a micrometre to 150, in five groups. A fifth of the activity is in particles small enough to breathe in. Every particle gets its own size and so its own fall speed, from a fraction of an inch a minute to several feet a second.
- Turbulence. A random walk whose strength comes from the friction velocity, heat flux and mixing height. Strong in a sunny afternoon, weak on a clear night, nearly absent above the mixed layer.
- What the forecast can't see. The forecast grid is 16 miles wide. Gusts and eddies smaller than that are added as a slowly wandering push, sized by how much the wind differs between neighbouring grid points.
- The city itself. Within about six miles of the centre the model makes the ground rougher and never lets the air get as still as it does over open country, because cities are rough and warm.
- Landing. Close to the ground a particle loses dust at a rate set by its fall speed and by how hard the air is being stirred. In rain it loses more, faster for bigger particles.
From dust to dose
Dose is in millisieverts (mSv). We add four things: breathing the cloud as it passes, standing in it, standing on contaminated ground afterwards, and breathing dust stirred back up. Coefficients are the published ones for caesium-137 (ICRP Publication 72 and Federal Guidance Report 12). Ground dose over the first year allows for rain and wear slowly carrying the caesium down into the soil.
The numbers assume a person outdoors, in one place, the whole time, with no cleanup. Indoors the dose is lower: roughly half in a wood-frame house, a fifth or less deep inside a large building.
| Dose | What it compares to |
|---|---|
| 0.1 mSv | A chest X-ray. Measurable with instruments, no effect on health. |
| 1 mSv | A third of the natural background everyone receives in a year (about 3 mSv). |
| 5 mSv | More than a normal year. Below any level at which people are moved. |
| 20 mSv in the first year | The Environmental Protection Agency's guide for relocating people until an area is cleaned. |
| 10 to 50 mSv in four days | The EPA's guide for sheltering or evacuating. |
| 100 mSv | The level above which a small increase in lifetime cancer risk has been measured in large populations. |
Radiation sickness begins around 1,000 mSv received quickly. From ER-1 to ER-9 nothing on these maps delivers that quickly: the highest figures are totals over a year of staying on the most contaminated ground. That is why experts say the blast, not the radiation, is what would cost lives. Only ER-10, simulated far beyond any dirty bomb, reaches such doses, and only close to the release.
What this cannot tell you
- Streets and buildings. Wind is channelled by streets and swirls behind buildings. A 16-mile forecast grid cannot see that. Within a few blocks of a release the real pattern would differ from ours.
- The real material. We assume caesium-137 as a fine powder because planners do. Other materials behave differently.
- The forecast being wrong. A wind forecast a day ahead is usually close on direction and timing, and sometimes is not. That is why we run again every hour and tell you when the answer changes.
- A real event. If something ever happened, officials would measure the ground with instruments within hours. Measurements replace models. Follow them, not us.
Sources
- Weather: NOAA/NCEP Global Forecast System, 0.25 degree, hourly.
- Cloud geometry: the explosive-release model documented in the HotSpot Health Physics Codes user's guide, Lawrence Livermore National Laboratory.
- Planning scenario: National Planning Scenario 11, Radiological Attack, US Department of Homeland Security.
- Protective action guides: PAG Manual, US Environmental Protection Agency, 2017.
- Dose coefficients: ICRP Publication 72; Federal Guidance Report No. 12.
- Turbulence: Hanna (1982) boundary-layer scales; Holtslag and Boville (1993) mixing profile; the method follows the FLEXPART and HYSPLIT particle models.
- Resuspension: Maxwell and Anspaugh (2011).
- Maps: OpenStreetMap contributors through OpenFreeMap; terrain from Mapzen and USGS; imagery from USGS.