Abstract
Direct observations of fault slip during earthquakes are rare but provide unique constraints on rupture processes. We analyze CCTV video recordings of the 2026 𝑀𝑤 6.8 Kumamoto and 2018 𝑀𝑤 6.4 Hualien earthquakes. At Kumamoto, the video resolves a 1.6±0.2-s slip pulse with a peak slip velocity of ∼1.3 m s−1 and an estimated slip-weakening distance of ∼0.7 m. Its onset time implies an average rupture speed of ∼2.1 km s−1 and a propagating slip-pulse width of ∼3.4 km. The video-derived slip velocity also closely resembles nearby fault-parallel ground velocities. The Kumamoto slip duration is similar to that observed during the 2025 𝑀𝑤 7.7 Myanmar earthquake despite their different magnitudes and rupture dimensions. In contrast, the Hualien case highlights the difficulty of isolating slip-pulse characteristics where surface deformation is distributed. These observations demonstrate the potential and limitations of video recordings for directly constraining slip-pulse characteristics. Plain Language Summary Videos that capture an earthquake rupture breaking the ground provide a rare opportunity to observe how faults move during earthquakes. We analyze recordings of the 2026 Kumamoto (Japan) and 2018 Hualien (Taiwan) earthquakes. The Kumamoto video reveals a short slip pulse lasting about 1.6 s, despite the earthquake being much smaller than the 2025 Myanmar earthquake, where a similarly short slip duration was observed. This suggests that local slip-pulse duration is not controlled by earthquake magnitude alone, but may instead reflect local rupture geometry and stopping processes. The close agreement between video-derived fault motion and nearby seismic recordings also supports the use of near-fault ground motions to infer local fault-slip behavior. In contrast, the Hualien case illustrates how distributed near-surface deformation can complicate the interpretation of video-derived fault motion.