[M-GI29] Data-driven geosciences

[MGI29-P05] Data-driven prediction of pumice drifting patterns using similarity search of the Kuroshio current axis: Application to Myojinsho, Izu Islands

*Tatsu Kuwatani¹, Hideitsu Hino², Haruka Nishikawa¹, Shotaro Akaho^3,2 (1.Japan Agency for Marine-Earth Science and Technology (JAMSTEC), 2.The Institute of Statistical Mathematics (ISM), 3.National Institute of Advanced Industrial Science and Technology (AIST))

Keywords:Drifting pumice, Data-driven, Ocean current, Natural hazards

Pumice raft drift has recently been recognized as a new disaster risk, and predicting its distribution patterns is crucial for ensuring safe maritime navigation and implementing timely coastal disaster mitigation measures. While drift simulations based on ocean current patterns have proven effective, conducting such physical simulations immediately after an eruption requires significant preparation time, human resources, and computational costs. Since eruptions are unpredictable and often small in scale, establishing a real-time prediction method that can serve as an alternative to physical simulations is an urgent issue. This study aims to develop a methodology for real-time drift pattern prediction by searching for similar patterns in daily-reported Kuroshio Current axis data, eliminating the need for physical simulations. We conducted validation tests using existing data under the assumption of a pumice eruption at Myojinsho, the Izu Islands, where the Kuroshio Current axis frequently passes nearby. The results demonstrated that the use of Dynamic Time Warping (DTW), a distance metric that measures similarity based on the overall alignment of directed paths, enables relatively accurate predictions of drift patterns over a 10-day period.