Rare event simulation of extreme European winter rainfall in an intermediate complexity climate model

Abstract

We test the application of a rare event simulation algorithm to accelerate the sampling of extreme winter rainfall over Europe in a climate model. The genealogical particle analysis algorithm, an ensemble method that interrupts the simulation at intermediate times to clone realizations in which an extreme event is developing, is applied to the intermediate complexity general circulation model PlaSim. We show that the algorithm strongly reduces the numerical effort required to estimate probabilities of extremes, demonstrating the potential of rare event simulation of seasonal precipitation extremes.

Bringing Statistics to Storylines: Rare Event Sampling for Sudden, Transient Extreme Events

“…However, heavy precipitation events (or high wind events) have some dynamical characteristics that distinguish them from the previous applications and pose challenges to existing rare event algorithms. Unlike continental‐scale, seasonally averaged anomalies studied previously (Ragone et al., 2018; Wouters et al., 2023), heavy precipitation events of interest are often sudden, transient, and relatively small‐scale. Their timescale at a particular location is often limited by the propagation of the dynamical feature causing the precipitation such as cyclones and fronts (Dwyer & O’Gorman, 2017).…”

“…Wouters et al. (2023) sampled extreme European seasonal precipitation accumulations, also using GPA. Webber et al.…”

“…(2018) and Wouters et al. (2023) relies on some slow‐moving notion of progress toward the extreme event, naturally given by the integrated temperature anomaly itself when targeting extreme seasonal average temperatures, in order to decide which simulations to clone or kill. In the precipitation study of Wouters et al.…”

“…In the precipitation study of Wouters et al. (2023), the extreme event is again a seasonal total, for which a mid‐seasonal total is a reasonable measure of progress. But for individual precipitation events, if one uses precipitation itself to measure progress toward the event, and applies perturbations to a simulation when precipitation picks up, it is too late for these perturbations to take effect by the time of maximum precipitation.…”

A new alpha logarithmic-generated class to model precipitation data with theory and inference

“…The advantages of employing multiple probability distributions to describe daily rainfall may be better understood by looking at changes in precipitation patterns that occur in different locations of the globe, which can give a helpful insight into those benefits. Extensive research on this subject has been conducted by a community of scientists from all over the world in locations such as Bangladesh (Alam et al [ 1 ]), Ethiopia (Liyew and Melese [ 2 ]), Costa Rica (Altman et al [ 3 ]), Nagpur (India) (Mohanty et al [ 4 ]), China (Chen et al [ 5 ]), Pakistan (Yonus et al [ 6 ]), Europe (Wouters et al [ 7 ]), Colombia (Correa et al [ 8 ]), Saudi Arabia (Hasanean and Almazroui [ 9 ]), Brazil (Beskow et al [ 10 ]), Pakistan (Amin et al [ 11 ]), Egypt (Gado et al [ 12 ]), and many others.…”