CAREER: Theory-Guided Statistical Framework for Advancing Learning from Post-Windstorm Engineering Assessments

这项教师早期职业发展（Career）资助将研究新的方法，以促进从风暴后侦察数据中学习。飓风和龙卷风等风暴每年给美国造成数十亿美元的经济损失，其中大部分是由于建筑物的性能问题。作为回应，研究人员收集了越来越多的数据集，记录了风暴后建筑物的状态。这些数据有可能推动基础科学和工程实践的进步，从而加强建筑物和社区的恢复能力，并减少未来的损失和其他影响。捕获风暴性能数据的强大能力远远超过了目前从这些数据中学习的能力，这些数据通常是不完整的、有偏见的，不适合有效地发现和应用知识。该项目将开发一个强大的、以理论为指导的统计推断框架，用于从风暴后数据中学习，这将改变尺度，以了解和预测风暴损害，特别是对低层建筑。这些进展将促进制定和实施更有效的减轻风暴风险和更有力的教育战略，并进一步为更有效和更智能的灾后侦察方法提供信息。我们会建立一个互动的外展平台，将研究成果传递给公众，并提高公众对影响风暴表现的关键因素的认识。将建立一个新的研究生和本科生组织，以促进灾害研究界的跨学科合作，培养新一代的工程师、社会科学家和政策制定者，他们对灾害和减轻灾害风险有更全面的了解。该项目的数据将在自然灾害工程研究基础设施（NHERI）数据仓库（https://www.DesignSafe-ci.org）中存档并公开提供。这笔拨款支持美国国家科学基金会（NSF）在国家减少风暴影响计划（NWIRP）中的作用。建筑物的风暴性能是一系列复杂的相互作用因素的函数，这些因素涉及气象学、工程学、公共政策和社会经济学，而这些因素并没有得到全面的理解。本研究的具体目标是将传统数据科学与基础理论和专家知识相结合，创建一个理论指导的统计推理框架，使高维风暴后侦察数据能够有效地发现知识。该项目将利用由美国国家科学基金会支持的结构极端事件侦察网络收集的来自最近风暴的高质量风暴后数据集，并使用额外的数据层和人机技术进行丰富，形成开发和试点新框架的强大测试平台。该框架将建立在概率图形模型的基础上，该模型允许已建立的理论和专家知识来定义已知的风暴表现因素及其基本相互关系，同时将重点放在因果推理上，而不是黑盒预测。最终，该研究将能够全面了解已知风暴性能因素的相对贡献，确定以前未知或低估的因素，并针对由实地观测支持的新研究领域。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This Faculty Early Career Development (CAREER) grant will investigate new methodologies to advance learning from post-windstorm reconnaissance data. Windstorms, such as hurricanes and tornadoes, continue to cost billions in economic losses each year in the United States, much of which is due to the performance of buildings. In response, researchers collect increasingly vast datasets documenting the post-windstorm state of buildings. These data have the potential to drive advancements in both fundamental science and engineering practice that will strengthen the resilience of buildings and communities and can reduce future losses and other impacts. The robust capabilities for capturing windstorm performance data vastly outweigh current capabilities for learning from this data, which are typically incomplete, biased, and ill-suited for efficient discovery and application of knowledge. This project will develop a robust, theory-guided, statistical inference framework for learning from post-windstorm data that will transform the scale to understand and predict windstorm damage, specifically for low-rise buildings. These advancements will spur the development and implementation of more effective windstorm risk mitigation and more robust education strategies, and further inform more efficient and intelligent post-disaster reconnaissance methodologies. An interactive outreach platform will be developed to translate the research findings to the general public and increase public awareness of the critical factors affecting windstorm performance. A new graduate and undergraduate student organization will be developed to foster inter-disciplinary collaboration within the disaster research community that will produce a new generation of engineers, social scientists, and policy makers that have a more holistic understanding of disasters and disaster risk mitigation. Data from this project will be archived and made publicly available in the Natural Hazards Engineering Research Infrastructure (NHERI) Data Depot (https://www.DesignSafe-ci.org). This grant supports the National Science Foundation (NSF) role in the National Windstorm Impact Reduction Program (NWIRP). Windstorm performance of buildings is a function of a complex set of interacting factors that span meteorology, engineering, public policy, and socioeconomics that are not holistically understood. The specific goal of this research is to combine traditional data science with fundamental theory and expert knowledge to create a theory-guided, statistical inference framework that will enable efficient knowledge discovery from high-dimensional post-windstorm reconnaissance data. The project will utilize high quality post-windstorm datasets from recent windstorms collected by the NSF-supported Structural Extreme Events Reconnaissance network, enriched using additional data layers and human-machine techniques, to form robust testbeds for developing and piloting the new framework. The framework will build upon probabilistic graphical models, which allow established theory and expert knowledge to define known windstorm performance factors and their fundamental interrelationships, while focusing on causal inference as the goal rather than black box predictions. Ultimately, the research will enable a holistic understanding of the relative contributions of known windstorm performance factors, identify previously unknown or underestimated factors, and target new research areas supported by field observations.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.