Building physical constraints into weather-related risk estimates

当需要推断非常罕见的回报期时，标准极值模型及其典型的统计推断技术可能导致具有大量不确定性的估计。在许多情况下，这些估计开始超过合理的水平，如此大的值会降低最终用户对极值模型的最终结果的信心。在最近极端天气事件的背景下，自然灾害特征仍然是法国电力公司研究的一个重要领域。这个博士项目被视为一种改进EDF能源公司使用的自然灾害表征方法的方法。总体目标是从历史记录中获得更可靠的极端降雨风险约束，利用物理原理从某些极端事件的可观察原因中得出推论。该项目将把极值理论的新发展与理解极端事件物理驱动因素的最新进展相结合。

When required to extrapolate to very rare return periods, standard extreme value models and their typified statistical inference techniques can lead to estimates with a large amount of uncertainty. In many situations, these estimates start to exceed plausible levels and such large values can reduce the confidence that end-users have in the final results from extreme value models.In the context of recent extreme weather events, natural hazard characterisation continues to be an important area of research for EDF Energy. This PhD project is seen as a way to improve the natural hazard characterisation approaches used within EDF Energy. The overall goal is to obtain more reliable constraints on extreme rainfall risk from historical records, using physical principles to draw inference from observable causes to certain extreme events. This project will combine new developments in extreme value theory with recent advances in approaches for understanding the physical drivers of extreme events.