The gap between two waves is a distance

How It Is Computed

1. The records. Each strip is the vertical broadband channel (BHZ) of one Global Seismographic Network station, networks IU and II, requested from the IRIS/EarthScope FDSN service as miniSEED and decoded from Steim-2 compression when the site is built. The decoder checks every record against the last-sample constant Steim stores in it. Each trace is band-passed from 0.05 to 2 Hz with causal second-order Butterworth sections, kept at 5 samples a second, and runs from two minutes before USGS's origin time, rounded down to the minute, for 27 minutes.
2. The gap. The P wave outruns the S wave, so the gap between their onsets grows with distance. For each station the page finds the distance whose ak135 S−P time equals the gap you picked, interpolating linearly between the table's whole degrees and solving by bisection: S(Δ, z) − P(Δ, z) = tS − tP Δ is in degrees of arc on a sphere of radius 6371.0088 km, and it is the circle's radius. Until a solution exists, circles assume a 15 km source.
3. The point. With three or more stations, the locator searches the whole sphere on a 2° grid at each of the five table depths (0, 15, 35, 50 and 100 km) for the point that minimizes Σi [ (S − P)ak135(Δi, z) − (tS,i − tP,i) ]² then refines the three best basins at 0.25° and 0.05° and finishes with Gauss–Newton steps in local kilometers, Geiger's 1912 method. The origin time is the mean of each P pick minus its ak135 P travel time. The 95% ellipse is σ²(JᵀJ)⁻¹ from the last step, scaled by √5.991, with σ² never taken below 1 s².
4. The receipt. USGS's epicenter is drawn only after your solution exists, and reported as the great-circle distance from yours.
5. The auto-picks. On a copy high-passed at 0.5 Hz, a short-term over long-term energy ratio (1 s over 10 s; Allen 1978) finds each P, and Akaike's information criterion (Maeda 1985) places the onset. A P-only location then guides a search for S within 40 s of its expected time on a long-period copy. They suggest a starting place; the receipts below say how far they land from USGS.

Why there is no magnitude

A magnitude needs each instrument's response removed, an amplitude read on a standard instrument, and a correction for distance. These traces are band-passed counts with the response left in, so a magnitude computed from them would be a guess. USGS's magnitude is on its event page, one link away from the answer.

Limits

• One velocity model for the whole planet. ak135 averages oceans, shields and subducting slabs; a wave that runs down a cold slab arrives seconds early, and the page cannot know it.
• Depth is coarse. The tables carry five depths, and at 20–90° the S−P gap barely changes with depth, so depth is the least certain number shown.
• The great quakes broke for over a minute along hundreds of kilometers of fault. The first P onset marks where rupture began, which is what USGS's epicenter marks too.
• Vertical component only. S shows more clearly on the horizontal components; the vertical carries both waves on one trace.
• No ellipticity correction (under 1 s at these distances) and no station elevation correction.

Receipts

Every row is a test in the repository; pnpm test prints the numbers.

The auto-picker misses its 80 km target on all three events, because its S picks on the vertical component land up to 40 s off. The misses are pinned in test/receipts.test.ts so they cannot grow. Picks by eye do better; that is the point of the page.