The frequency of Climate-induced Disasters (CID) has tripled in the last three decades, driving the World Economic Forum to identify them as the most likely and most impactful risks worldwide. With more than 70% of the world population expected to be living in cities by 2050, ensuring the resilience of urban infrastructure systems under CID is crucial. The present work employs data analytics and machine learning techniques to develop a performance prediction framework for infrastructure systems under CID. The framework encompasses four stages related to: extracting meaningful information about the impact of CID on infrastructure systems and identifying the latter's performance; investigating the relationship between different CID attributes and previously identified system performance; employing data imputation using unsupervised machine learning techniques; and developing and testing a supervised machine learning model based on the different influencing CID attributes. To demonstrate its application, the developed framework is applied to disaster data compiled by the National Weather Services between 1996 and 2019 in the state of New York. The analysis results showed that: i) power systems in New York are the most vulnerable infrastructure to CID, and particularly to wind-related hazards; ii) power system performance level depends on hazard-system interactions rather than solely hazard characteristics; and iii) a 4-predictors random forest-based model can effectively predict power system performance with an accuracy of 89%. This work is expected to aid stakeholders in developing spatio-temporal preparedness plans under CID, which can facilitate mitigating the adverse impacts of CID on infrastructure systems and improve their resilience.

在过去的三十年中，气候诱发灾害（CID）的频率增加了两倍，推动了世界经济论坛将其确定为全球最可能和最有影响力的风险。到2050年，预计将有70％以上的世界人口居住在城市中，因此确保CID下的城市基础设施系统的弹性至关重要。本工作采用数据分析和机器学习技术来为CID下的基础设施系统开发性能预测框架。该框架包括四个阶段，这些阶段涉及：提取有关CID对基础设施系统影响的有意义的信息，并确定后者的性能；调查不同的CID属性与先前确定的系统性能之间的关系；使用无监督机器学习技术进行数据估算；并根据影响CID的不同属性开发和测试有监督的机器学习模型。为了演示其应用，已开发的框架应用于1996年至2019年纽约州国家气象服务局收集的灾难数据。分析结果表明：i）纽约的电力系统是CID尤其是风力相关危害最脆弱的基础设施；ii）电力系统的性能水平取决于危害系统的相互作用，而不仅取决于危害特征；和III）的4预测随机森林基于模型能够有效地预测电力系统性能为89％的准确度。预期这项工作将帮助利益相关者在CID下制定时空防备计划，从而可以减轻CID对基础设施系统的不利影响并提高其弹性。