We describe in this paper a semi-parametric bivariate extreme value approach for studying rare extreme precipitation events considered as events that result from a combination of extreme precipitable water (PW) in the atmospheric column above the location where the event occurred and extreme precipitation efficiency, described as the ratio between precipitation and PW. An application of this framework to historical 6-h precipitation accumulations simulated by the Canadian Regional Climate Model CanRCM4 shows that uncertainties and biases of very long-period return level estimates can be substantially reduced relative to the standard univariate approach that fits Generalized Extreme Value distributions to samples of annual maxima of extreme precipitation even when using modest amounts of data.

我们在本文中描述了一种半参数双变量极值方法，用于研究稀有的极端降水事件，这些事件被认为是事件发生地点上方大气柱中的极端可降水量（PW）与极端降水发生率的组合所导致的事件，描述为降水量与PW之比。该框架在加拿大区域气候模型CanRCM4模拟的历史6小时降水累积中的应用表明，相对于适合广义极值分布的标准单变量方法，可以大大减少很长时期收益水平估计的不确定性和偏差。即使使用少量数据，极端降水的年度最大值的样本。