Statistical tools for climate science

气候科学统计工具

• 批准号: 1401793
• 负责人: Peter Guttorp
• 金额: $ 12万
• 依托单位: University of Washington
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2014
• 资助国家: 美国
• 起止时间: 2014-07-01 至 2018-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1401793&HistoricalAwards=false
• 关键词: Statistical; tools; climate; science

In order to plan for adaptation to climate change, it is important to develop quantitative tools for the uncertainty of climate projections at global, region, and local levels. In this research we develop statistical tools that will let decision-makers get a clearer perspective on how to interpret the projections. For example, we predict (with attendant uncertainty) temperatures at unobserved sites, given that an observed site achieved the network maximum or minimum. The technical tools used in this research include empirical process models for nonstationary and dependent observations. Data will include global temperature reconstructions, networks of temperature and precipitation stations from Sweden, Norway and the United States, and CMIP5 global model runs as well as regional climate model runs from CORDEX and other repositories. We will also study nonstationary tools in extreme value theory, including spatial and spatio-temporal situations, and implement visualization methods for comparing output from ensembles of climate models to data.

为了规划适应气候变化，开发针对全球、区域和地方层面气候预测不确定性的定量工具非常重要。在这项研究中，我们开发了统计工具，使决策者能够更清晰地了解如何解释预测。例如，假设观察到的站点达到了网络最大值或最小值，我们就预测（伴随着不确定性）未观察到的站点的温度。本研究中使用的技术工具包括非平稳和相关观测的经验过程模型。数据将包括全球温度重建、瑞典、挪威和美国的温度和降水站网络、CMIP5 全球模型运行以及来自 CORDEX 和其他存储库的区域气候模型运行。 我们还将研究极值理论中的非平稳工具，包括空间和时空情况，并实施可视化方法来比较气候模型集合的输出与数据。