Hsu Chia-Wei, Mostafavi Ali
Urban Resilience.AI Lab, Zachry Department of Civil & Environmental Engineering, 3136 TAMU, College Station, TX, 77843-3136, USA.
Sci Rep. 2025 Apr 7;15(1):11897. doi: 10.1038/s41598-025-95909-8.
Resilience curves have been widely used for conceptualizing and representing specific aspects of resilience behavior during hazard events; however, their use has often remained conceptual with limited data-driven characterization and empirical grounding. While broader community resilience encompasses multiple social, economic, and infrastructure dimensions, targeted analyses of specific systems can provide valuable insights into resilience patterns. Empirical characterizations of resilience curves provide essential insights regarding the manner in which differently impacted systems of communities absorb perturbations and recover from disruptions. To address this gap, this study examines human mobility resilience patterns following multiple weather-related hazard events in the United States by analyzing more than 2000 empirical resilience curves constructed from high-resolution location-based mobility data. These empirical resilience curves are then classified into archetypes using k-means clustering based on various features (e.g., residual performance, disruption duration, and recovery duration). Three main archetypes of human mobility resilience are identified: Type I, with rapid recovery after mild impact; Type II, exhibiting bimodal recovery after moderate impact; and Type III, showing slower recovery after severe impact. The results also reveal critical thresholds, such as the bimodal recovery breakpoint at a 20% impact extent (i.e., function loss), at which the recovery rate decreases, and the critical functional threshold at a 60% impact extent, above which recovery rate would be rather slow. The results show that a critical functional recovery rate of 2.5% per day is necessary to follow the bimodal resilience archetype when impact extent exceeds 20%. These findings provide novel and important insights into different resilience curve archetypes and their fundamental properties. Departing from using resilience curves as a mere concept and visual tool, the data-driven specification of resilience curve archetypes and their properties improve our understanding of the resilience patterns of human systems of communities and enable researchers and practitioners to better anticipate and analyze ways communities bounce back in the aftermath of disruptive hazard events.
恢复力曲线已被广泛用于概念化和表示灾害事件期间恢复力行为的特定方面；然而，它们的使用往往仍停留在概念层面，数据驱动的特征描述和实证基础有限。虽然更广泛的社区恢复力涵盖多个社会、经济和基础设施维度，但对特定系统的针对性分析可以提供有关恢复力模式的宝贵见解。恢复力曲线的实证特征提供了关于社区中不同受影响系统吸收扰动并从破坏中恢复的方式的重要见解。为了弥补这一差距，本研究通过分析从高分辨率基于位置的移动性数据构建的2000多条实证恢复力曲线，研究了美国多次与天气相关的灾害事件后的人类移动性恢复力模式。然后，这些实证恢复力曲线根据各种特征（例如，残余性能、中断持续时间和恢复持续时间）使用k均值聚类被分类为原型。确定了人类移动性恢复力的三种主要原型：I型，在轻度影响后快速恢复；II型，在中度影响后呈现双峰恢复；III型，在严重影响后恢复较慢。结果还揭示了关键阈值，例如在20%影响程度（即功能损失）时的双峰恢复断点，此时恢复率下降，以及在60%影响程度时的关键功能阈值，超过该阈值恢复率将相当缓慢。结果表明，当影响程度超过20%时，每天2.5%的关键功能恢复率对于遵循双峰恢复力原型是必要的。这些发现为不同的恢复力曲线原型及其基本特性提供了新颖而重要的见解。与仅仅将恢复力曲线用作概念和可视化工具不同，数据驱动的恢复力曲线原型及其特性的规范提高了我们对社区人类系统恢复力模式的理解，并使研究人员和从业者能够更好地预测和分析社区在破坏性灾害事件后反弹的方式。
