Damage depends on magnitude, depth, distance and what is standing — the table gives what a shallow quake near people typically does. Pick a magnitude above and its row lights up; click any example to jump to it on the map. Death tolls are the commonly quoted figures and estimates vary, sometimes by a factor of two or more.
ML = log10 A − log10 A0(δ) — the log of the peak swing A (mm) on a Wood–Anderson seismometer, minus the swing a reference quake would give at that distance. Built for Southern California and saturates near 7, so nobody uses it for big quakes any more.Mw = ⅔ (log10 M0 − 9.1), with M0 = μ · A · D in N·m — rigidity (about 30 GPa) × rupture area × average slip. This is the number quoted today, and it does not saturate. Newspapers still say “Richter”.log10 E = 1.5 M + 4.8 (joules). One step is ×101.5 = 31.6 in energy; two steps are ×1,000.log10 N = a − b M, with b close to 1 — each step down is about ten times as many quakes. The page takes b from the catalogue’s own counts at M6 and M7.log10 L = −2.44 + 0.59 M km, log10 A = −3.49 + 0.91 M km², log10 D = −4.80 + 0.69 M m — fitted to M4.8–8.1, extrapolated above that.Δ = tS−P · VPVS / (VP − VS), about 8 km for every second of gap in crustal rock.The same energy does very different things depending on where it lands. Haiti, 2010: magnitude 7.0, a shallow rupture beside a city of unreinforced concrete — estimates run from about 100,000 to over 300,000 dead. Anchorage, 2018: magnitude 7.1, similar depth — no deaths, from stricter building codes, wood-frame housing and rehearsed response. The magnitude scale measures the quake; the casualty count measures the quake and the building.
Mostly not. Charles Richter’s 1935 scale used one type of instrument, and its numbers stop climbing around magnitude 7 — a magnitude 9 reads like a magnitude 8 on it. Since the 1970s agencies quote the moment magnitude, Mw, which was chosen to agree with Richter’s numbers for medium quakes and keep going. So when the news says “Richter 9.1” for Tōhoku, the number is an Mw.
The scale is logarithmic. A magnitude 7 shakes the ground with ten times the amplitude of a 6 and releases about 32 times the energy; it takes about 1,000 magnitude 5s for the energy of one magnitude 7. The numbers are small because the physics is large.
No. The USGS is plain that nobody has ever predicted a major earthquake’s date, place and size. What is possible is a probability — the map is the best argument for it, because the same belts have been shaking for decades. Aftershocks are the exception: Båth’s law puts the largest aftershock about 1.2 magnitudes below the mainshock, and Omori’s law has their rate fall roughly as 1/time.
Shading counts every magnitude 6 or larger earthquake in the USGS catalogue from 1973 to 2025 (53 years, explosions removed) in each 2° × 2° cell, about 220 km across. It is a map of how often the ground has produced large quakes, not of danger: danger needs people and buildings on top. It starts in 1973 rather than 1900 because the early catalogue is thin in remote areas and would over-shade the places that happened to have seismographs. Fifty-three years is also short against the centuries between the great quakes of slow faults — Cascadia, which last broke in 1700, and the New Madrid zone are nearly blank here and are not safe. Plate boundaries are the PB2002 model (Bird, 2003).
Energy uses radiated seismic energy from the Gutenberg–Richter relation, an empirical average with a scatter of about a factor of three; the strain energy a quake releases is several times larger, and most of it becomes heat. TNT and bomb equivalents compare that radiated energy only: a real explosion couples a small fraction of its energy into the ground, so the 2020 Beirut blast, roughly a magnitude 5 in energy, registered nearer 3.
The seismogram panel uses the Hutton–Boore (1987) distance correction for Southern California, valid from
roughly 10 to 600 km, and assumes a Wood–Anderson amplitude: the gap-to-distance step uses
VP = 6.0 and VS = 3.5 km/s, so a gap is 8.4 km a second. Elsewhere it is a good
estimate, not a measurement. Rupture sizes are extrapolated above M8.1 and are for crustal faults; subduction
megathrusts run longer. The power figure divides the energy by the rupture length over 2.8 km/s. Verified in quake_model.py.
Between the first jolt (the P wave) and the heavier shaking behind it (the S wave), count the seconds and multiply by 8: roughly how many kilometres away it began. Ten seconds is about 80 km. The gap exists because P waves travel about 6 km/s and S waves about 3.5, and it grows with every kilometre they run.
A quake has a single magnitude and a different intensity at every place. Intensity (the Modified Mercalli scale, I to XII) describes what happened to people and buildings, and it falls with distance, rises on soft ground, and is higher near the rupture. The same magnitude 6 can be a rattle at 100 km and a catastrophe at 5.
Most damaging quakes are shallow, under about 30 km. A deep one, several hundred km down, spreads its energy over a huge area before it reaches the surface and is felt weakly: the 2011 Fiji magnitude 6.5 in the catalogue was 558 km deep.
A great quake sets the whole planet vibrating for weeks, and moving that much rock changes the Earth’s spin slightly: NASA estimated the 2011 Tōhoku quake shortened the day by about 1.8 microseconds.
Releases in which this page changed, newest last. Derived from the archived copy of every release, not from notes written afterwards — so it reflects what actually shipped. Site-wide passes are left out; they are in the full changelog.