Guides¶

What each stage does, why it is done that way, and what the choices cost.

The pipeline, and the conventions¶

How a spectrum is processed: Every step from a raw waveform to a moment magnitude, with the equation it applies and a pointer to the code that applies it.

This is also where the conventions are stated, deliberately in the sections that apply them rather than on a page of their own:

Amplitude convention: FAS is the folded \(2\lvert X\rvert\); MAGNITUDE is the unfolded \(\lvert X\rvert\), and that is the one \(\Omega\) is defined in. Reading a folded plateau as \(\Omega\) puts \(M_0\) out by two — 0.2 magnitude units.
The Parseval contract: The one energy check every estimator is held to, which is what lets multitaper, Welch, FFT and the CWT be interchangeable rather than merely similar.
Moment and magnitude: The \(M_0\) and \(M_w\) equations, the medium constants and where they come from, and the two distances that are in different units on purpose.

Choosing a transform: What each estimator does to your data, measured. On real windows the plateau moves by 7–15% between estimators — about 0.03 magnitude units — and window position alone reaches 0.08. Small against the 0.13 m.u. scatter quoted for spectral \(M_w\), but systematic rather than random, so it does not average away across stations at similar distance.
Reading picks: What arrival formats are read out of the box, how to add one, and how a pick is matched to a trace.

Publishing a dataset: Taking an event from an FDSN archive to a hash-pinned entry in the registry, so a result can be reproduced from the same bytes.