Comparative analysis of human mobility around municipal offices in five cities and towns affected by the 2024 Noto Peninsula Earthquake using smartphone GPS data

Abstract

In the Noto region affected by the 2024 Noto Peninsula Earthquake, while a continuous decline in the resident registration population has been reported, post-disaster human mobility dynamics, including non-resident visitors, have not been systematically captured. This study analyzed changes in the estimated number of visitors observed from smartphone GPS data at five municipal office areas in the affected region: Wajima City Hall, Suzu City Hall, Noto Town Office, Anamizu Town Office, and Nanao City Hall. The analysis periods were set as the pre-disaster period, from April to September 2023, and the post-disaster period, from April to September 2025, maintaining the same month composition (n = 6 vs. n = 6) to control for seasonal factors. The post-disaster period was selected to capture mobility dynamics after the emergency response phase had largely concluded, following the complete withdrawal of the Self-Defense Forces on August 31, 2024. The analytical methods employed Welch’s t-test with Cohen’s d effect size, cosine similarity, and Ward’s hierarchical clustering to ensure reproducibility.

The results revealed statistically significant increases in the estimated number of visitors across all five locations, ranging from 2.47 to 3.20 times the pre-disaster levels, with Cohen’s d values from 2.83 to 7.38 and p < 0.01 for all locations. The day-of-week profiles maintained cosine similarity above 0.99 between the pre- and post-disaster periods, indicating structural stability. Ward’s method extracted a hierarchical structure in which Wajima City Hall, the location with the highest disaster intensity, was classified independently from the other four locations. In contrast to the declining resident registration population, which decreased by 5.6% to 12.9% across the five municipalities, the estimated number of visitors around municipal offices exceeded pre-disaster levels. These findings suggest that dynamic regional activities, which cannot be fully captured by resident registration statistics alone, may be continuing in the affected areas.