Move four · the wedge
Where it starts: earthquake prediction
Of the six preparations on the build page, this one has a single deliverable, a death rate behind it, and a size a grant can hold. Its inputs are the inputs the contested solar question turns on, so it answers more than it is funded for.
One · the arithmetic
One hour, 100 kilometres, one action.
Forecasting is usually argued as a physics problem. Priced as a public health problem it looks different, because what an earthquake kills people with is mostly a building.
Earthquakes kill about 18,300 people a year on the thirty-year average, and more than three quarters of those deaths come from a structure falling on somebody rather than from the ground. Over a century that is about 1.8 million people, some 1.4 million of them indoors.
One hour of warning and 100 kilometres of accuracy is enough for one action: out of the building, into open ground. It will not evacuate a city or empty a hospital, and it does not have to. The one action it buys addresses the largest share of the toll.
A warning at that resolution sits on the order of a million lives a century.
| Step | Figure | In the mind's eye |
|---|---|---|
| Deaths per year | About 18,300 | Thirty-year average. |
| Killed by the structure rather than the ground | More than three quarters | The building is the hazard the warning acts on. |
| Over a century | About 1.8 million | Adelaide, Hobart and Darwin together. |
| Of those, indoors | Some 1.4 million | Near enough everybody in Adelaide. |
| Resolution needed | One hour, 100 km | Long enough to walk out of a building. |
| What sets the range | Compliance | How many people move, not how well the ground is read. |
Two · compliance
The binding constraint is trust, not physics.
A warning nobody acts on saves nobody. The upper bound on how many people act was measured in 2004, with no technology involved.
Simeulue is an island off Sumatra. On Boxing Day 2004 it sat about sixty kilometres from the epicentre and was among the first land the wave reached. Aceh, on the mainland, lost something like a hundred and seventy thousand people. Simeulue lost seven, from a population near seventy thousand.
There was no siren, no seawall and no message. There was a word.
Smong is kept in bedtime stories, in song, in commemorations and cut into grave markers, told there since the tsunami of 1907. The instruction: if the earthquake is strong and then the sea pulls back, do not go and look, run for the high ground. A trigger, a trap and an action, carried ninety-seven years with no institution, siren or funding behind it.
Which sets the ceiling and names the failure mode together. Compliance that high rested on an instruction already believed, and belief is what a false alarm spends. The false alarm rate is the variable the toll is most sensitive to. Oral tradition as a dated, actionable record is on the stories page.
The trigger is geophysical, the toll is governance
Deaths from collapse concentrate where construction standards go unenforced: about 83 per cent of them over three decades, in countries rated anomalously corrupt (Ambraseys and Bilham, Nature, 2011). Warning is the layer available to somebody who cannot change the building stock.
Three · the physics under it
Stressed rock behaves like a battery.
Lights before a quake, radio noise with no transmitter, ground potential shifts, warm ground in a thermal image. Each is listed as its own curiosity, and one piece of solid-state physics sits under all of them.
Most crustal rock carries dormant charge built into the mineral structure, in pairs of oxygen atoms bonded to each other with the charge locked in place. These peroxy defects are ordinary rather than rare. Under stress the bonds break and release mobile positive charge carriers, which flow out of the loaded volume into rock that was not squeezed. This is Friedemann Freund's work, and the laboratory half is reproducible: rock into a press, charge measured arriving in the untouched end.
A stressed volume of rock becomes a battery, driving current into everything around it.
What that buys is not a sixth indicator. It is one mechanism under five that are studied separately, so they should move together and in order. The correlation between them is a stronger trigger than any one alone, and nobody appears to be watching them as a set.
| Usually listed as its own indicator | What one mechanism has it doing |
|---|---|
| Ultra-low-frequency radio before a quake | Current moving through a large volume of rock, which is a very big and very slow aerial. |
| Ground potential shifts | Charge leaving one volume and arriving in another, setting up a voltage difference between the two. |
| Corona discharge and light at the surface | The charge cloud reaching the surface and breaking down the air sitting on top of it. |
| Air ionisation near the ground | Charge arriving at the rock face and ionising the air in contact with it. |
| Infrared seen from orbit | The ionised air heating and radiating, which reads as warm ground in a thermal image while the ground itself may not have warmed. |
Both halves of where it stands
The laboratory effect is not the contested part. The reach is. As a forecasting method it is not established, and earthquake lights have competing accounts that need none of this: frictional heating on the fault, piezoelectricity in quartz-bearing crust, grains grinding together. Which one does the work in a given case is unsettled.
Four · the reason this page exists
And the stack is the same stack.
The inputs are collected already, by agencies with no stake in this argument, and they are item for item the stack the solar argument is fought over. Live links on the watch page.
Solar wind speed and density. The interplanetary magnetic field. Total electron content, which is the charge state of the ionosphere read off navigation satellites. Cosmic ray flux from the ground-based neutron monitor network. Coronal hole stream catalogues. Seismic catalogues. Free, public, updated daily, and needing no new hardware to begin.
Build it for the fundable reason, and the contested answer falls out of the same feed.
| Feed | What it measures | What it serves |
|---|---|---|
| Solar wind speed and density | The plasma arriving upstream, ahead of Earth. | The one-day-lag seismic claim; the loading argument. |
| Interplanetary magnetic field | Field strength and orientation in the wind. | Whether the magnetosphere couples, and how hard. |
| Total electron content | Ionospheric charge, from navigation satellite delays. | Pre-seismic ionisation; storm response. |
| Neutron monitor network | Cosmic ray flux at the ground. | The volcanic bubble-nucleation mechanism; heliospheric shielding. |
| Coronal hole stream catalogues | Recurrent fast wind on a roughly 27-day beat. | The discriminating test below, both lanes at once. |
| Seismic catalogues | Time, place and magnitude, decades deep. | The response side of every one of the above. |
What a funder is asked for is a forecasting tool with a defined deliverable. What comes with it is a modelled record of how the Sun, the shield and the ground move against each other, which the argument on the claim has been missing.
Five · two tests it can run now
Both can fail in a named direction.
Neither needs new instruments, and both want the method registered before the data is touched. The other open tests are on the test page.
A coronal hole is a patch where the Sun's magnetic field opens into space instead of looping back, so the wind escapes fast. The stream sweeps past Earth as the Sun turns, returning about every 27 days and holding for five to ten turns. Each arrival raises geomagnetic forcing for days and drops cosmic ray flux at the ground, in a recurrent Forbush decrease, so the two proposed couplings disagree at the same event: seismicity should rise, and silica-rich volcanism should fall. Dozens of repeats of one driver stack, where one-off correlations do not.
The second test is lunar. Tidal stress peaks at new moon and at full, so a purely gravitational effect on earthquakes has to be symmetric across the month. The Moon crosses Earth's magnetotail only at full, and the geomagnetic modulation measured since the 1960s runs about four per cent below average in the seven days before full and four per cent above in the seven after.
Gravity says the month is symmetric. The magnetotail says it is not.
The method is to subtract the tidal stress prediction from a public catalogue and read the residual by lunar phase. A flat residual kills it. A lean toward the days after full is not something gravity can produce.
Six · where it stands
Contested, unfunded, and not applied for.
The position at the strength the evidence supports.
The direct version of the claim is contested. Marchitelli and colleagues published a correlation between solar wind proton density and earthquakes above magnitude 5.6 at a one-day lag in Scientific Reports in 2020. A formal Matters Arising rebuttal followed, and the USGS and mainstream seismology reject it. A tool built to test it does not need it to be true; a null result at that resolution is worth having.
The funding route that would have fitted best does not exist either. The XPRIZE natural disaster prediction earthquake prize did not launch, and the specifications still circulating for it are design figures for something unfunded.
The honest state of it
Nothing is built. No feed is ingested, no epoch analysis registered, no residual plotted. The cheapest change of position is an Australian business number, free and available the same day, and it has not been applied for.