Collaborative Research: CAS - Climate: Improving Nonstationary Intensity-Duration-Frequency Analysis of Extreme Precipitation by Advancing Knowledge on the Generating Mechanisms

2221803（马斯卡罗）和2221808（昆克尔）。该项目将解决气候动力学和土木工程交叉的基本问题，旨在改善气候变化下基础设施系统的设计。目前针对极端降水和相关洪水设计基础设施的程序依赖于强度-持续时间-频率（IDF）曲线。这些曲线是在平稳性（“未来与过去相同”）和同质统计总体存在的假设下，通过对历史雨量计记录的统计分析而产生的（“由多种产生机制引起的降水的统计变异可以用单一分布来解释”）。这些假设最近受到理论论证和气候模拟的挑战，这些理论论证和气候模拟表明，极端降水（EP）统计数据预计将在未来变暖的气候中发生变化。该项目的主要目标是推进对亚日和日EP产生机制变化的了解，并利用这些新知识开发一种新的物理驱动的统计框架，为改进的非平稳IDF曲线的开发提供信息。本文将探讨两种研究假设：(1)EP产生机制的发生和/或热力学和动力学成分随时间变化，导致IDF设计值发生变化；(2)结合气候模式模拟的EP产生机制变化信息的统计模型可以得到改进的非平稳IDF曲线。研究假设将通过每小时和每日降雨量记录、大气再分析和美国多个地区的气候模拟进行测试，这些地区涵盖了广泛的EP主要机制，包括锋面、温带气旋、热带气旋、中尺度对流系统、气团对流和北美季风。通过应用深度学习技术，将获得导致美国次日和日持续时间EP统计数据非平稳性的物理机制的新知识。通过结合统计测试和耦合模式比对项目第6阶段的历史气候模拟，将设计新的方法来检测、物理验证和不确定性量化EP的区域趋势以及发生机制的特征。这些知识将为考虑混合种群的EP的IDF分析提供非平稳统计模型。区域化技术将旨在减少参数估计的不确定性，并将影响EP的物理机制的区域变化纳入其中。将采用贝叶斯框架来解释不确定性。与目前的平稳方法相比，非平稳统计模型的附加价值将通过交叉验证自举实验进行评估。然后，统计模型将与不同未来情景下选定的环流模型模拟的EP机制变化一起应用，并用于计算考虑非平稳极端情况的新设计指标。方法上的进步旨在解决当前固定IDF设计方法的关键限制。最重要的是，EP有多种产生机制，其频率和强度可能以不同的方式变化，以响应人为强迫的气候变化。这项研究将建立一个框架，将这些知识客观地纳入IDF曲线，为未来的变化提供更可靠的估计。这个项目的目标是产生多重更广泛的影响。它旨在改善基础设施设计和管理，以应对气候变化下的环境污染，这将(1)减少与风暴有关的损失（2001年至2020年美国损失2611亿美元），从而造福社会；（ii）提高关键基础设施的韧性，从而改善国家安全。该项目将制定一个扩展计划，通过与工程和建筑公司、公共机构和美国土木工程师协会的持续合作，包括从业者教育。此外，该项目还将为研究机构在工程、水文学和大气科学领域开发新的课程。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。

2221803 (Mascaro) and 2221808 (Kunkel). This project will address fundamental questions at the intersection of climate dynamics and civil engineering aimed at improving the design of infrastructure systems under climate change. Current procedures to design infrastructure against extreme precipitation and associated flooding rely on intensity-duration-frequency (IDF) curves. These curves are generated through statistical analyses of historical rain gage records under the assumptions of stationarity ("the future is the same as the past") and existence of a homogeneous statistical population ("the statistical variability of precipitation caused by multiple generating mechanisms is explained by a single distribution"). These assumptions have been recently challenged by theoretical arguments and climate simulations, which suggest that extreme precipitation (EP) statistics are expected to change in a future warmer climate. The main goal of this project is to advance knowledge on changes in the generating mechanisms of sub-daily and daily EP and use this new knowledge to develop a novel physics-driven statistical framework to inform the development of improved nonstationary IDF curves. Two research hypotheses will be investigated: (1) The occurrence and/or thermodynamic and dynamic components of the generating mechanisms of EP are changing in time, leading to changes in IDF design values; and (2) improved nonstationary IDF curves can be developed through statistical models that incorporate information on changes in the generating mechanisms of EP simulated by climate models. The research hypotheses will be tested using hourly and daily rainfall records, atmospheric re-analyses, and climate simulations in multiple regions of the U.S. spanning a wide range of dominant mechanisms of EP, including fronts, extratropical cyclones, tropical cyclones, mesoscale convective systems, air mass convection, and North American monsoon. New knowledge on the physical mechanisms leading to the non-stationarity of EP statistics at sub-daily and daily durations in the U.S. will be acquired by applying deep learning techniques. Novel methods for detection, physical validation, and uncertainty quantification of regional trends in EP and in occurrence and characteristics of the generating mechanisms will be designed by combining statistical tests and historical climate simulations from the Coupled Model Intercomparison Project Phase 6. This knowledge will inform a nonstationary statistical model for IDF analysis of EP that accounts for mixed populations. Regionalization techniques will be designed to reduce uncertainty in parameter estimation and incorporate regional changes in the physical mechanisms affecting EP. A Bayesian framework will be adopted to account for uncertainty. The added value of the nonstationary statistical model compared to current stationary approaches will be assessed through cross-validation bootstrapping experiments. The statistical model will be then applied with changes in EP mechanisms simulated by selected general circulation models under different future scenarios and used to compute new design metrics accounting for nonstationary extreme regimes. The methodological advancements are intended to address critical limitations of current stationary IDF design methods. Most importantly, there are multiple generating mechanisms for EP whose frequency and intensity may change in different ways in response to anthropogenically-forced climate change. The research will develop a framework for the objective incorporation of this knowledge into IDF curves, providing more robust estimates of future changes. This project is targeted to have multiple broader impacts. It seeks to improve infrastructure design and management against EP under climate change, which would (i) reduce storm-related damages ($261.1 billion from 2001 to 2020 in the U.S.), thus benefitting society; and (ii) increase the resilience of critical infrastructure, thus improving national security. The project will develop an outreach program that involves practitioner education through ongoing collaborations of the investigators with engineering and construction firms, public agencies, and the American Society of Civil Engineers. Additionally, the project will develop new aspects of curricula at the investigators’ institutions in the areas of engineering, hydrology, and atmospheric sciences.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.