Collaborative Research: Advancing the Data-to-Distribution Pipeline for Scalable Data-Consistent Inversion to Quantify Uncertainties in Coastal Hazards

沿海灾害是对世界各地公民、工业和政府的持续威胁。美国利益尤其关注的是从墨西哥湾到北大西洋西部社区的风暴潮和飓风洪水，北极风暴和不断变化的海冰覆盖之间的相互作用，影响北美沿海社区，以及从油轮和深水钻井平台等来源扩散的石油泄漏。因此，对这些沿海灾害的建模和模拟中的不确定性进行量化的能力，对于就如何最好地准备、减轻和应对这类灾害作出基于数据的决策至关重要。研究小组的目标是提高最先进的数学、统计和计算能力，以解决这些具有社会重要性的应用程序。此外，数学、统计和计算研究广泛适用于科学界和工程界都感兴趣的广泛应用。教育影响包括这一领域的本科生和研究生的培训。该项目需要建立在严格的测量理论基础上的多方面研究方法，以扩大数据一致性反演(DCI)的应用范围，DCI是一种识别、量化和减少基于物理的计算模型的输入(参数)的不确定性源的方法，适用于广泛的复杂物理系统。一个方面是开发和分析基于深度学习的数据到分布管道，以将时空数据云转换为DCI的非参数分布，从而可以将最佳实验设计标准纳入管道中。另一个方面是开发了一种可扩展的DCI方法，该方法同时解决了高维特征空间引起的计算问题，以及由于计算昂贵的模型而导致的模拟数据有限的问题。第三个方面是开发一种迭代的DCI方法，该方法可以部署在操作环境中，以在数据变得可用时识别最可能的关键模型参数。公共信息系统将执行数据通信公共领域软件和数据到分发管道的算法开发。PIS将主要使用最先进的高级环流(ADCIRC)模型及其变体来模拟沿海危险。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Coastal hazards are a persistent threat to citizenry, industry, and governments worldwide. Of particular concern to US interests are storm surge and flooding from hurricanes in communities stretching from the Gulf of Mexico to the western North Atlantic, interactions between Arctic storms and evolving sea ice coverage impacting North American coastal communities, and oil spill spread from sources such as tankers and deep-water drilling rigs. The ability to quantify uncertainties in the modeling and simulation of these coastal hazards is therefore critical to making data-informed decisions about how to best prepare, mitigate, and respond to such hazards. The research team aims to advance state-of-the-art mathematical, statistical, and computational capabilities to address these applications of societal importance. Moreover, the mathematical, statistical, and computational research are broadly applicable to a wide range of applications of interest to both the scientific and engineering communities. Educational impacts include the training of undergraduate and graduate students in this field. This project requires a multi-faceted research approach built upon a rigorous measure-theoretic foundation to expand the application of Data-Consistent Inversion (DCI), a methodology to identify, quantify, and reduce sources of uncertainty for inputs (parameters) of physics-based computational models, to a wide range of complex physical systems. One facet is the development and analysis of a deep learning based data-to-distribution pipeline to transform spatial-temporal data clouds into non-parametric distributions for DCI that can incorporate optimal experimental design criteria within the pipeline. Another facet is the development of a scalable approach to DCI that simultaneously addresses computational issues arising from high-dimensional feature-spaces as well as limited availability of simulated data due to computationally expensive models. A third facet is the development of an iterative approach to DCI that can be deployed in an operational setting to identify the most likely critical model parameters as data become available. The PIs will implement the algorithmic developments in public domain software for DCI and the data-to-distribution pipeline. The PIs will primarily utilize the state-of-the-art Advanced Circulation (ADCIRC) model and its variants for modeling coastal hazards.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.