 Solution

Distinguish the duration of one phase pulse from the P-to-S separation. Figure 4 shows northern pulses with most visible motion over roughly 3–4 s each, compared with roughly 4–5 s or more in the south. Reading the low-amplitude tails gives approximately 11.5–15.2 s for the northern P pulse, 17.7–20.9 s for the northern S pulse, 14.2–19.0 s for the southern P pulse, and 22.3 s to beyond the right plot boundary near 27 s for the southern S pulse. Thus the latter duration is censored by the plotting window; it should not be assigned an exact ending at 27 s. The positive-to-negative extremum separations are only about 1 s in the northern pulses, 1.3 s in the southern P pulse and 1.4 s in the southern S pulse; those much shorter numbers are not full source durations. Smooth tails have no unique onset or ending at the precision of this drawing.

A finite-source rupture has an intrinsic duration $T$ that differs from the apparent pulse duration through rupture directivity. In an ideal unilateral line-source model, neglecting local slip rise time, instrument filtering and path broadening,Here $c$ is the chosen phase wave speed and $L_{\parallel }$ is the rupture's projection along equal-and-opposite takeoff directions. Averaging corresponding full pulses in that model suggests an intrinsic timescale of roughly 4–5 s. This is an order-of-magnitude graph estimate, not an exact rupture duration; the figure does not supply a seismic source time function definition, receiver responses, phase deconvolution, or geometry sufficient to determine one uniquely. A finite local slip rise time would also contribute to the observed width.

In particular, neither 13 s nor 19 s is a rupture duration: those numbers locate phase motion on the absolute clock. The northern P–S gap of 6 s is a propagation delay rather than the rupture duration.