Time-compressed composite-earthquake inversion for multi-event rupture representation and rupture-growth-tendency analysis in the 2023 Kahramanmaraş, Türkiye sequence

Abstract Regional tectonic loading on a connected fault system can often be treated, to first order, as a slowly varying background process. Here we test the feasibility of representing several distinct earthquakes on the same fault system as ordered source units of one complex composite event. The method compresses inter-event origin-time separations but does not time-warp the waveform samples within each event. For the proportional mapping, C1, C2, C34 and C5 are placed at 0, 6.94, 355.7 and 419 s, superposed and inverted on a common 600 × 200 km plane discretised into 30 × 10 subfaults of 20 × 20 km. Under the proportional temporal mapping, the direct C12345 composite inversion is compared with the time-shifted sum of the separately inverted C1, C2, C34 and C5 source models, yielding a source-time-function correlation of 0.865. A 19-case sensitivity suite gives a median spatial correlation of 0.710, a median source-time-function correlation of 0.965 and a median centroid shift of 7.49 km, indicating that local patch geometry is less stable than the first-order spatial and temporal organisation. A constrained convolutional image model, trained on the 0–700 s slip-image sequence and combined with adjacent-front, depth and increment caps, is then used to examine slip-increase and rupture-growth tendency over 700–1000 s. The resulting zones show partial retrospective proximity to later M ≥ 5.0 earthquakes. The present results support the feasibility of a unified time-compressed multi-event representation.