Abstract
Coastal closed-circuit television (CCTV) cameras are ubiquitous yet rarely exploited quantitatively for tsunami detection. To address this gap and the scarcity of automated methods, we propose a technique that converts CCTV footage into a time-series of wave runup heights-instantaneous shoreline elevations corresponding to each incoming wave. The workflow has two main steps: (i) compute a luminance-variation (SIGMA) image in which the runup edge appears as a bright curve, and (ii) apply a color-based land-water mask to suppress dynamic noise on land. Preliminary tests under varied lighting, wave, and obstacle conditions confirmed the method's stability. Application to footage from seven CCTV cameras during the 1 January 2024 Mw 7.5 Noto Peninsula tsunami revealed one of the tsunami's dominant 300-500 s energy band and yielded root-mean-square errors of 0.094 m and 0.191 m at two sites after removing short-period (< 180 s) components-while processing ran faster than real time on a standard computer. In the cases studied, the extracted runup time-series demonstrated the potential to complement sparse offshore gauges for real-time tsunami detection and post-event analysis. Future work will target a broader range of recording conditions, refine signal separation, and validate the method on additional tsunami events with characteristics different from the Noto case.