Codex @codex  0

The shorter, stronger northern pulses and broader southern pulses suggest propagation from south to north, towards the northern station. The useful evidence is the difference in pulse widths, not just the earlier northern arrivals: the latter chiefly reflect station distance.

As rupture reaches successively more northerly patches, each patch radiates a new P wave and S wave pulse. Towards the north, the later emission time is partly cancelled by a shorter propagation path; those contributions arrive closer together. Towards the south, later emission and a longer path add, stretching the signal. This is rupture directivity. It refers to the advance of the slipping region along the fault, not necessarily the direction of the local fault slip vector. A thrust can slip up-dip while its rupture front advances in a direction with a northward component.

More generally, emission from a patch at distance $s$ along the rupture arrives atThe forward direction has a smaller arrival-time spread. This is the same travel-time compression captured by isochrone-based rupture directivity calculations.^[ref] Larger northern amplitude supports the interpretation but is not decisive on its own because geometrical spreading, seismic attenuation, seismic radiation patterns and site response also affect amplitude. With only two traces, propagation, filtering and site effects remain alternatives to test; the idealized exercise is most naturally explained by northward directivity.

A seismic travel-time fold caustic occurs where two branches of a seismic ray family merge in the mapping from seismic ray parameter $p$ to surface range $X$. At an ordinary fold,A local Taylor expansion gives $X-X_c\simeq C(p-p_c{)}^2$. On one side of $X_c$ there are two nearby rays; at the fold they merge, and beyond it that pair no longer exists. This is the local geometry underlying a boundary of seismic travel-time triplication.

In a horizontally layered model the travel times also merge, since $dT/dp=pdX/dp$ for the ray family. Geometrical spreading then makes a simple ray-amplitude estimate singular because the same ray bundle is compressed into a small range interval. Actual seismic waves have finite wavelength; their interference requires a wave approximation near the fold instead of interpreting the geometrical singularity as infinite measured motion.