[lispandfound]

Summary

• Adds sources.ry0_distance(faults, point), implementing Ry0 = max(0, −U, U − L) using the GC2 U coordinate. Accepts a single Fault/Plane or a sequence of them, and is vectorised over points (output shape point.shape[:-1]).
• For multi-fault ruptures, raises a RuntimeWarning (the transform isn't really well-defined there) and takes the rupture extent as the span of the traces along the GC2 U axis, i.e. [min segment origin, max(segment origin + segment length)]. For a single fault this reduces exactly to [0, L].
• Factors the segment-length / U-origin computation out of gc2_distances.multi_trace_rx_ry into multi_trace_u_origins.

Multi-fault GC2 fix

Plotting Ry0 for Darfield showed U was not monotonic along strike: west of the rupture, U increased, so Ry0 stayed 0 for 30+ km past the trace ends. Spudich & Chiou (2015, pp. 5–6) reverse the point order of each discordant trace (e_j = (p_{n+1,j} − p_{1,j})·â with a different sign from E = Σ e_j) before computing T and U. We only flipped per-segment direction vectors when computing the nominal strike, so discordant traces (Greendale NW-SE and the fault near Hororata, for Darfield) contributed rx/ry with the wrong orientation. This PR:

• adds gc2_distances.strike_corrected_traces, which decides discordance per trace and reverses the trace points;
• uses the corrected traces in multi_fault_rx_ry_distance and ry0_distance;
• takes the nominal strike as the sum of corrected end-minus-start vectors (Eq. 11) and places the U origin at the back end of it (p. 6).

Along an E–W profile through Darfield (northing 5174 km, eastings 1480–1580 km), U went from 10.5 13.8 16.6 15.6 22.2 17.9 27.6 … to a monotonic -19.8 -12.2 -4.8 1.6 9.8 18.0 27.7 37.2 45.2 52.7 60.4.

Darfield Ry0

Ry0 around the 7-fault Darfield (Beavan) rupture, shared by all 29 realisations in darfield_rupture_gen_validation:

Tests

New tests in tests/test_rx_ry.py:

• Ry0 matches max(0, -U, U - L) for a plane and a multi-plane fault; vectorised results match a per-point loop for single- and multi-fault ruptures
• No warning for a single fault; RuntimeWarning for multi-fault ruptures
• Discordant traces are reversed as a whole (not per segment); T and U are invariant to a discordant trace's point order; U = 0 at the back end of a westward-striking system

tests/test_rx_ry.py and tests/test_sources.py pass locally (116 tests).

🤖 Generated with Claude Code

[lispandfound]