Scientific limitations & safety

There is at present no validated method for predicting an individual earthquake's date, location and magnitude with enough precision to act on. This is not a gap in Prakruti; it is the state of the science after more than a century of effort. Anyone who tells you otherwise — including any website, model or person citing planetary alignments, animal behaviour, cloud shapes or radon — is making a claim that has not survived scientific testing.

What is possible, and genuinely useful, falls into four separate categories that Prakruti deliberately keeps apart:

• Observation. Recording earthquakes that have already happened, usually within a minute or two. This is most of what Prakruti does and it is reliable.
• Earthquake early warning. Detecting an earthquake that has already begun and warning places the shaking has not yet reached, buying seconds to tens of seconds. Real, operational in several countries, and not the same thing as prediction. Prakruti does not operate one and does not relay one.
• Probabilistic forecasting. Estimating the rate of earthquakes in a region over a window — for example the elevated chance of aftershocks after a mainshock. Useful for planning; useless for deciding whether to leave a building this afternoon.
• Seismic hazard assessment. Long-term estimates of the shaking a location should be built to withstand, over decades. This is what actually saves lives, through building codes.

Collapsing these four into one number labelled “prediction” is the single most common way disaster information misleads people. Prakruti keeps them in separate categories with separate labels throughout.

An earthquake begins as a rupture a few metres across, kilometres underground, and whether it stops there or grows into a magnitude 8 appears to depend on conditions along the fault that cannot currently be measured from the surface. The same small initial rupture can produce a tremor nobody feels or a catastrophe; the difference is not visible beforehand.

Many proposed precursors have been reported after the fact and then failed to appear before the next earthquake, or appeared without one following. A signal that cannot be specified in advance, and that produces more false alarms than hits, is not a prediction method — it is hindsight.

Prakruti tests astronomical hypotheses openly rather than dismissing them, because testing is more useful than assertion. The lunar and solar tides really do stress the Earth, and asking whether that shows up in earthquake timing is a legitimate scientific question with a respectable literature.

What Prakruti will not do is manufacture a result. Tests are declared before they run. Aftershocks are removed, because counting one earthquake sequence hundreds of times creates significance out of nothing. P-values are corrected for the number of hypotheses tried. Effect sizes are printed beside every p-value, because with a large enough catalogue an effect far too small to matter will still look statistically significant. Null results are published as readily as positive ones — and so far, they are most of what the lab finds.

Planetary influence in particular fails on physical grounds before the statistics even begin: the tide-raising force a planet exerts on Earth is smaller than the Moon's by many orders of magnitude. Prakruti still shows the geometry, because people ask about it and an honest answer beats a dismissal — but it shows it as astronomy, not as a warning.

See the current results →

Magnitude is not intensity. Magnitude is one number for the whole earthquake, describing its size at the source. Intensity describes the shaking at a given place and varies from street to street with soil, depth and distance. A deep magnitude 6 may be barely noticed; a shallow magnitude 5 beneath a town can be damaging.

“Richter” is one scale among many. Prakruti prints the scale each agency actually used — mww, mb, ml and others — because they measure different things and are not interchangeable. The original Richter scale saturates for large earthquakes and is no longer used for them.

Event counts reflect instruments as much as the Earth. More earthquakes are detected where seismometers are dense. A regional increase in counts may mean better detection, not more seismicity.

Storm forecast tracks are the centre of a range. A forecast position is a best estimate with meaningful uncertainty, and a cyclone's effects extend far beyond its centre. Being off the track line is not being out of danger.

Wind speeds are not comparable between agencies. NHC reports a 1-minute sustained wind; most other centres report a 10-minute average, which runs roughly 12% lower for the same storm. Prakruti labels every wind speed with its averaging period.

An empty map does not mean nothing is happening. Prakruti ingests a limited set of sources: most of the world's tropical cyclone basins are not yet covered, and official warning feeds are not yet integrated at all. A quiet page may mean a quiet planet, or it may mean a source Prakruti does not read. The data sources page states exactly which is which, and the system health page shows whether anything has stopped updating.

For any decision that matters, use your national emergency authority. They have legal responsibility, local knowledge, and warning channels Prakruti does not.