CoPe EAGER: Addressing Human-Centric Decision-Making Challenges from Coastal Hazards via Integrated Geosciences Modeling and Stochastic Optimization

沿海社区容易受到热带风暴、飓风和强降雨事件的影响。这些事件最近在频率和强度上都有所增加。因此，开发基于科学的智能系统、工具和模型至关重要，这些系统、工具和模型可以捕捉沿海灾害的潜在行为，并在自然灾害发生之前、期间和之后协调和优化决策，以增强沿海社区的复原力和应对能力。本研究致力于将基于地球科学的沿海洪水建模与基于场景的随机优化相结合，以解决沿海社区在飓风和其他洪水诱发事件后面临的以人为中心的决策问题，这是一项探索性和统一的研究议程。利用这些事件作为典型的沿海灾害，该项目解决了一个以人为中心的具体问题：在这些灾害发生之前从医院和疗养院疏散病人。病人疏散计划尤其重要，因为管理不善多次导致医院、疗养院或疏散期间不必要的死亡。开发一种有效的决策支持工具，供区域疏散协调机构使用，可能会在未来的灾难中对美国各地产生广泛的影响。事实上，这项研究的一个主要目标是创造一种可以传播供国家使用的工具。在大规模多医院病人疏散方面开发的知识和工具，将导致在面对不确定但可预测的事件（如飓风）时，以最佳方式协调有限资源的新方法。此外，这种综合办法可推广到其他沿海后勤问题（例如，预先安置应急用品、安置庇护所、预先安置修复资源和关键基础设施恢复所需的备件），从而启动新的研究议程。这项研究还与包括天气、飓风和洪水预测以及应急管理和疏散在内的各种组织进行了强有力的合作，以确保所生产工具的可行性和可用性。在教育方面，pi将创建疏散建模教学模块，并开发一门关于人道主义行动研究的新课程。该项目侧重于一个对沿海社区有重大影响的具体问题，以突出将基于地球科学的沿海洪水建模与基于场景的随机优化相结合的价值：优化应对洪水诱发事件的大规模多医院和养老院疏散。这个高风险的问题需要精确的洪水预测。特别地，本研究将耦合天气预报、径流生成、河流路线、淹没测绘模型（一般来说，地球科学模型）与决策问题的底层随机优化模型集成在一起。地球科学模型的主要用途将是严格生成洪水情景，这些情景将作为随机优化模型的输入。天气研究和预报模型、水文建模系统（WRF-Hydro）的模块化架构与诺亚陆地表面模型（LSM）将耦合到基于矢量的河流路径模型（RAPID）。综合地球科学模型将在飓风或强降雨事件发生之前生成基于统计的洪水情景，以改进对资源分配和物流决策的建议（例如，集结区位置、医务人员的分配、救护车在发送和接收设施之间的分配/路线等）。最后，认识到飓风预报的不确定性，这一努力产生了一系列洪水情景（而不是单一的实现），用于病人疏散问题，这是以前没有做过的。提议的工作的一个重要优点是将两个通常不密切合作的研究团体结合在一起：运筹学和地球科学建模。在建立这座桥梁的过程中，研究将地球科学建模的预测能力与随机优化的规定能力联系起来。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。

Coastal communities are susceptible to flooding due to tropical storms, hurricanes, and heavy rainfall events. These events have increased recently in frequency and intensity. Therefore, it is critical to develop smart, science-based systems, tools and models that capture the underlying behavior of coastal hazards, and to coordinate and optimize decisions before, during and after natural hazards, to enhance the resilience and response of coastal communities. This research undertakes an exploratory and unifying research agenda focused on integrating geosciences-based modeling of coastal floods with scenario-based stochastic optimization for human-centric decision-making problems that coastal communities face in the wake of hurricanes and other flood-inducing events. Using such events as archetypal coastal hazards, the project addresses a specific human-centric problem: evacuating patients from hospitals and nursing homes just before such hazards. Patient evacuation planning is especially important as mismanagement has several times led to unnecessary deaths in hospitals, in nursing homes, or during evacuation. Development of an effective decision support tool, to be used by regional evacuation coordination agencies, could have wide-ranging impact across the United States in future disasters. Indeed, a primary goal of this research is to create a tool that can be disseminated for national use. The knowledge and tools developed on large-scale multi-hospital patient evacuation will lead to new ways to optimally coordinate limited resources when faced with uncertain but predictable events such as hurricanes. Moreover, this integrated approach is extendable to other coastal logistical problems (e.g., prepositioning emergency supplies, siting shelters, prepositioning repair resources and spares for critical infrastructure recovery) thus initiating new research agendas. This research also features robust collaboration with various organizations, including those involved in weather, hurricane, and flood prediction, and emergency management and evacuation, in order to ensure feasibility and usability of the tools produced. On the educational front, the PIs will create teaching modules on evacuation modeling and develop a new course on humanitarian operations research. This project focuses on a specific problem that significantly affects coastal communities in order to highlight the value of integrating geosciences-based modeling of coastal floods with scenario-based stochastic optimization: optimizing large-scale multi-hospital and nursing home evacuation in response to flood-inducing events. This high-stakes problem needs accurate flood predictions. In particular, this research integrates coupled weather forecast, runoff production, river routing, inundation mapping models (in general, geoscience models) with an underlying stochastic optimization model of the decision-making problem. The main use of the geoscience models will be the rigorous generation of flooding scenarios that will serve as input to the stochastic optimization models. The modular architecture of the Weather Research and Forecasting Model, hydrological modeling system (WRF-Hydro), with the Noah Land Surface model (LSM), will be coupled to a vector-based river routing model (RAPID). The integrated geoscience model will generate statistically-grounded flooding scenarios before a hurricane or heavy rainfall event in order to improve recommendations for resource allocation and logistics decisions (e.g., staging area locations, allocation of medical personnel, allocation/routing of ambulances between sending and receiving facilities, etc.). Finally, recognizing the uncertainty in the hurricane forecasts, this effort generates a series of flood scenarios (instead of a single realization) to be used in the patient evacuation problem, which was not done before. A significant merit of the proposed work is to bring together two research communities that do not usually work closely together: operations research and geosciences modeling. In creating this bridge, the research links the predictive power of geosciences modeling with the prescriptive power of stochastic optimization.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.