Mid-Indian Ridge

Rayleigh window: from the arrival of the fastest period (45 s, 3.84 km/s) to that of the slowest (25 s, 3.32 km/s) along this path (GDM52). Shorter, slower periods arrive afterwards.

Body waves: IASP91 travel times (Kennett & Engdahl, 1991), computed with TauP for a source at 10 km. Rayleigh waves: GDM52 group velocities (Ekström, 2011). Not included: ellipticity and elevation (about a second), Earth heterogeneity (a few seconds), later phases (pP, PP, SS…).

Measurable at Uccle? Order of magnitude: standard IASPEI magnitude relations (2013) read backwards, against the median noise level of the Peterson models (1993), where the noise measured at Uccle sits; measurable when signal / noise ≥ 3. The focal mechanism and the structure along the path are ignored.

Vertical component (HHZ channel), reduced from 100 to 10 samples per second when the site is updated (Butterworth anti-aliasing low-pass at 2.5 Hz, applied in both directions), then, in the browser, detrended and band-pass filtered with a Butterworth filter in both directions, so without shifting the arrivals. Each trace is scaled to its maximum in the displayed window; that maximum, on the right, is the signal divided by the sensitivity of the instrument, without correcting its frequency response.

Signals: public archive of ORFEUS (EIDA), network BE (Royal Observatory of Belgium).

Multiple filter technique (Dziewonski, Bloch & Landisman, 1969), applied when the site is updated to the raw signal (100 samples per second): trend removed, ends tapered over 30 s, then, for each of the 81 periods from 1 to 100 s, a Gaussian filter exp(−α ((f − f₀)/f₀)²) with α = 50 and the envelope of the analytic signal; the power (squared envelope) is smoothed in time with a Gaussian whose standard deviation equals the filter’s, √(2α) T₀ / 2π (the smoothing operator of Torrence & Webster, 1999, used by Grinsted et al., 2004), and the displayed amplitude is its square root, the RMS amplitude at the filter’s resolution, free of the fades of the instantaneous envelope; one sample every 1.28 s. Up to normalisation, this computation is equivalent to a continuous wavelet transform with the Morlet wavelet (ω₀ = √(2α) ≈ 10). Amplitudes as ground velocity, corrected for the frequency response of the sensor (poles and zeros from the ORFEUS inventory), in decibels relative to the maximum; below −60 dB, the background colour.

Transverse: −E·cos β + N·sin β, with β the back-azimuth of 123°, the direction from the station to the epicentre (ObsPy convention); both horizontals are converted to velocity with their sensitivity before the rotation; both components share the same 0 dB.

The horizontal band already present before the waves arrive, strongest here around a period of 6 s, is the permanent background noise generated by ocean waves (microseisms); it does not come from the earthquake.

Computation checked against a second, independent implementation (ObsPy and NumPy) on the same signal. Signals: public archive of ORFEUS (EIDA), network BE (Royal Observatory of Belgium).

Rays computed when the site is updated by integrating the ray equations in IASP91 (velocity linear within each layer, ray parameter conserved at interfaces, first arrival kept when several rays exist), checked against TauP (ObsPy) on randomly drawn depth–distance pairs: travel time within 0.01 s, angles within 0.01°, depth along the ray within one kilometre. The fans are checked the same way, ray by ray; the travel-time curves are the tables computed by TauP.