Scaling laws of human mobility persist during extreme floods

TL;DR

This study demonstrates that human mobility patterns, including the relationship between visitor density, distance, and visitation frequency, remain largely consistent during extreme floods, revealing the resilience of underlying scaling laws.

Contribution

It provides the first analysis of human mobility scaling laws during extreme floods, showing their persistence and proposing a mechanistic explanation for observed behaviors.

Findings

Visitor density relationships remain robust during floods.

Marginal density over frequency shows time-invariant power-law behavior.

Aggregated density over distance exhibits complex decay patterns.

Abstract

Although a number of studies have investigated human mobility patterns during natural hazards, mechanistic models that capture mobility dynamics under large-scale perturbations, such as extreme floods, remain scarce. Leveraging mobile phone data and building upon recent insights into universal mobility patterns, we assess whether the general structure of population flows persists during the extreme floods that struck Emilia-Romagna, Italy, in 2023. Our analysis reveals that the relationship between visitor density, distance, and visitation frequency remains robust even under extreme flooding conditions. To disentangle the effects of distance and visitation frequency, we define two aggregated visitor densities: the marginal density over frequency and the aggregated density over distance. We find that the marginal density over frequency exhibits a time-invariant power-law exponent, indicating resilience to flooding disturbances. In contrast, the aggregated density over distance displays more complex behavior: an exponential decay over biweekly periods and a power-law decay over a monthly interval. We propose that the observed power law emerges from the superposition of exponential distributions across shorter timescales. These findings provide new insights into human mobility scaling laws under extreme perturbations, highlighting the robustness of visitation patterns and suggesting avenues for improved mechanistic modeling during natural disasters.