Massive geolocation data reveal evacuation behaviour during the 2024 Noto Peninsula earthquake and tsunami

Abstract

On 1 January 2024, devastating tsunamis caused by the Noto Peninsula earthquake hit coastal areas within several minutes, but only two tsunami casualties were officially reported. Despite its importance, the cause of this unexpectedly low human loss was unclear because of the limited access to the peninsula and the presence of many visitors during the holiday, which made conducting conventional surveys infeasible. Here, we reveal evacuation behaviour during the 2024 Noto Peninsula tsunami using massive geolocation data collected from a smartphone app. By analysing these massive data, which include over 1.5 million records collected on this day, we find that the evacuation was extremely fast, occurring within 2--6 minutes after the origin time. Further analyses suggest that these fast departures were driven mainly by strong ground shaking; the fact that the tsunami occurred during the family-oriented New Year holiday was also a key factor. Additionally, the long-term analysis of the data reveals that people started returning to the coastal area 20--100 minutes after the origin time, which was long before the downgrading and cancellation of the tsunami warnings. These results highlight the utility of the innovative data-driven approach to evacuation surveys, which addresses the limitations of conventional evacuation surveys.

Figures (4)

• Figure 1: Massive geolocation data recorded during the 2024 Noto Peninsula earthquake and tsunami (16:10--16:20 JST). An animation is available on Zenodo (https://doi.org/10.5281/zenodo.14300988) zenodo. Each blue dot represents an individual position recorded at a certain time. The animation reveals notable landward movements after the earthquake occurred. In this study, 10 coastal cities (Wajima, Suzu, Noto, Nanao, Himi, Toyama, Itoigawa, Joetsu, Kashiwazaki, and Niigata) with sufficient observations were analysed. The areas for the analyses were determined on the basis of the standard geological grid mesh code (JIS X 0410), shown as white rectangles in the figure. The green star represents the epicentre of the main shock.
• Figure 2: Evacuation departure trends. The difference in distance between each location data point and the coastline were calculated, and the differences were organised into 1 min bins. The plots and bars represent the mean values and standard errors, respectively. Changes in the positive direction indicate landward movements, and those in the negative direction indicate seaward movements. The increasing trends of landward movement within 10 minutes after the origin time are considered to represent tsunami evacuation behaviour.
• Figure 3: Relation between the observed peak ground accelerations and the proportion of landward movements within 10 minutes after the earthquake origin time. Locations of municipal offices were used to calculate the distance from each city to the epicentre.
• Figure 4: Time series of the proportion of seaward trips within 6 hours before and after the earthquake origin time. Here, moving averages with a 10 minute window size are presented to show the trends.