CMG: Non-stationary Spherical Processes to Synthesize Multimodel Climate Change Simulations

该项目的目标是评估全球和区域气候变化预测的内在不确定性，其依据是使用先进和专门定制的统计方法从大约20个不同的最先进大气-海洋环流模型获得的模型输出。 所提出的方法涉及的贝叶斯统计框架的发展，量化的不确定性和归因于不同的因素。 将使用在地球仪相邻空间区域借势的空间统计模型，以便在统计上准确评估气候模型偏差和模型间变异性。 该方法的另一个特点是能够综合不同模型的气候变化预测，然后缩小到几乎任意的区域，提供一致的不确定性估计。 需要解决的问题包括：未来气候变化在不同时间和空间尺度上的不确定性有多大？ 均值的变化与变异性和极端事件的变化有何关系？ 当使用联合气候变量时，预测技能可以提高多少？ 在多大程度上，与观察结果的一致意味着对未来的预测能力？ 如何有效地模拟地球仪上气候变化的明显非平稳性？拟议的项目结合了最近在统计和气候科学方面的研究工作，并促进在气候科学中使用先进的统计方法。 它也适用于许多其他科学领域，如宇宙背景辐射建模，污染物测绘和疾病传播。 拟议的工作也将有助于培养研究生。 该项目的跨学科性质以及统计学家和气候科学家之间的密切合作将是一次非凡的教育经历，也是学生和所有参与者知识网络的来源。 该项目的成果将为科学家、决策者和公众提供评估气候变化的重要工具。

The objective of this project is the assessment of the inherent uncertainty of global and regional climate change projections based on model output obtained from about twenty different state-of-the-art atmosphere-ocean general circulation models using advanced and specifically tailored statistical methodologies. The proposed approach concerns the development of a Bayesian statistical framework for quantifying uncertainty and attributing it to different factors. Spatial statistical models that borrow strength across adjacent spatial regions of the globe will be used in order to provide a statistically accurate assessment of climate model bias and inter-model variability. An additional feature of the methodology is the ability to synthesize climate change projections across the different models and then to down-scale to almost arbitrary regions providing a coherent uncertainty estimate. Questions to be addressed include: How large are the uncertainties in future climate change on different time and spatial scales? How are changes in the mean related to changes in variability and extreme events? By how much can predictive skill be improved when using joint climate variables? To what degree does agreement with observations imply predictive skill for the future? How can one efficiently model the evident non-stationarity of climate variability on the globe?The proposed project combines recent research efforts in statistics and climate science and promotes the use of advanced statistical methodology in climate science. It is also applicable to many other scientific areas, such as cosmic background radiation modeling, pollutant mapping, and disease spreading. The proposed work will also contribute to the training of graduate students. The interdisciplinary nature of the project and the close collaboration between statisticians and climate scientists will be an extraordinary educational experience and a source of intellectual networks for students and all participants. The results of this project will provide important tools for the assessment of climate change to scientists, policy makers and the general public.