Collaborative Research: CAS-Climate: Risk Analysis for Extreme Climate Events by Combining Numerical and Statistical Extreme Value Models

气候深刻地影响着诸如大型野火、热浪、洪水和干旱等极端现象的严重程度和频率。应对气候变化最剧烈影响的能力需要了解未来气候条件下的极端事件。气候模式是研究地球系统动力学的宝贵工具，但它们在极端事件分析方面存在不足。首先，与现实观测相比，许多气象变量的极端行为表现出严重的不匹配。其次，他们生活在网格空间中，必须与连续的世界协调一致，在这个世界中进行观察并进行风险分析。为了解决这两个困难，我们将在天气现象的真实世界领域进行建模，利用来自气候模型的大尺度模式表示的信息，以一种保留极端事件的现实属性的方式。该项目还为研究生提供研究培训机会。PI将专注于两个主要研究目标，即开发和应用分析工具，将气候模型的输出转化为对极端事件的现实分析。首先，该项目将开发模型和相关的模型拟合软件，利用气候模型输出的动态衍生大尺度特征，为连续空间中局部极端气象现象的随机过程描述提供信息。这将需要两个相互关联的建模组件：1)一个随机模拟模型，将网格空间中的气候模型输出与连续空间中的极端空间事件联系起来；2)一个随机过程模型，以模拟模型为条件，真实地表示空间尾部依赖性。其次，该项目将生成基于模型的极端事件预测，用于影响分析。从第一个研究目标中开发的模式中随机抽取，以气候模式大规模特征的预估为条件，在功能上构成极端天气发生器。PI将使用这些随机结果作为两种影响模型之一的输入：降水将数据输入到水文模型中，以预测暴雨洪水风险，而风、温度和降水将数据输入到火灾蔓延模型中，以预测野火风险。他们的模型的软件实现将产生具有实际尾部行为的潜在未来气候变量的随机图，这些图可以在下游用作直接帮助风险评估的其他数值模型的输入。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。

Climate profoundly influences the severity and frequency of extreme phenomena like large wildfires, heatwaves, floods, and droughts. Resilience to the most dramatic effects of climate change requires an understanding of extreme events under future climate conditions. Climate models are invaluable tools for interrogating the dynamics of the Earth system, but they have shortcomings with respect to extreme event analysis. First, the behavior of extremes of many meteorological variables shows a profound mismatch compared to real-life observations. Second, they live in gridded spaces that must be reconciled with the continuous world in which observations are made and for which risk analysis is performed. To resolve these two difficulties, we will model weather phenomena in their native real-world domain, leveraging information from representations of large-scale patterns from climate models, in a way that preserves realistic properties of extreme events. The project also provides research training opportunities for graduate students. PI will focus on two main research aims, which develop and apply analytical tools that turn climate model output into a realistic analysis of extreme events. First, this project will develop models, and associated model-fitting software, that leverage dynamically-derived large-scale features from climate model output to inform stochastic process descriptions of local extreme meteorological phenomena in continuous space. This will require two interconnected modeling components: 1) a stochastic analogue model to link climate model output in gridded space to extreme spatial events in continuous space, and 2) a stochastic process model, conditional on the analogue model, that realistically represents spatial tail dependence. Second, this project will generate model-based projections of extreme events for use in impact analysis. Random draws from the model developed in the first research aim, conditional on climate model projections of large-scale features, functionally constitute an extreme weather generator. PI will use these random draws as inputs to one of the two impact models: precipitation draws feed into a hydrology model to project pluvial flood risks, and wind, temperature, and precipitation draw feed into a fire spread model to project wildfire risk. The software implementation of their model will produce stochastic draws of potential future climate variables that have realistic tail behavior, which can be used downstream as inputs to other numerical models that directly aid in risk assessment.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.