报告题目：Assessing the reliability of probabilistic event attribution
报告专家：Francis Zwiers 院士
报告时间：2023年9月15日（周五）10:00-11:00
报告地点：气象楼423会议室
主 持 人：周波涛教授
专家简介：
Francis Zwiers，加拿大皇家科学院院士，美国地球物理联合会（AGU）和美国气象学会（AMS）会士。现任加拿大维多利亚大学太平洋气候影响联合中心（PCIC）主任，曾任加拿大气候变化与环境部气候建模与分析中心（CCCma）高级研究科学家兼负责人以及气候研究部主任。发表了200余篇论文和书籍章节，其中12篇论文发表在《Nature》杂志上，是汤普森-路透社高被引科学家。曾担任IPCC第四次评估报告的协调主要作者和IPCC第五次评估报告主席团成员及第一工作组副主席。曾任Journal of Climate的主编和International Journal of Climatology副主编。目前是Journal of Climate 和 Journal of Geophysical Research-Atmosphere的副主编，以及跨学科杂志Advances in Statistical Climatology, Meteorology and Oceanography的联合主编。
报告摘要：
High-impact extreme events bring questions about the role of human-driven climate change in their occurrence. The field of extreme event attribution (EEA) aims to answer these questions, as documented in a 2016 National Academies report. However, some fundamental aspects of this rapidly developing field remain open, including whether attribution metrics and model-estimated event probabilities are reliable. We start by illustrating event attribution concepts by describing one or two recent examples of EEA studies. We then study EEA metric properties by using large ensembles of climate simulations in a perfect (in-model) and imperfect (out-of-model) framework to define extreme events, calculate their probabilities, and evaluate the reliability of the event probability or change in probability. We show that estimates of an event’s return period vary enough between models that the specific probability values cannot often be considered reliable. Model-estimated relative changes in event probability are generally more reliable than the event probabilities themselves. Conclusions about whether there is an attributable increase (or decrease) in event probability are more often reliable for variables with large signal-to-noise ratios such as hot extremes. Overall, we show that it cannot be assumed that model-estimated event attribution metrics and their components are reliable and recommend that event attribution results be presented in a generally qualitative format.
欢迎广大师生踊跃参加！
气象灾害教育部重点实验室
气象灾害预报预警与评估协同创新中心
大气科学学院
2023年08月31日