Efficient simulation of extreme events (Speaker: Daan Crommelin, CWI & UvA)

Speaker: Daan Crommelin, CWI & UvA

Extreme events are examples of rare events, combining high impact with low probability of occurrence. Simulating such rare events with a numerical model is computationally challenging, requiring very long simulations to observe just a few events. To tackle this challenge, a collection of mathematical techniques and algorithms have been developed, aimed at making simulation of rare events more efficient, and in particular making the statistical characterization of these events (such as probability of occurrence) more accurate. These "rare event simulation" techniques have found their way into various application fields, ranging from queueing systems (e.g call centers) to computational chemistry. Despite the importance of extreme events in weather and climate, rare event simulation has hardly been explored in this context. I will discuss some of the concepts and techniques that are used in rare event simulation, and their potential use for weather and climate extremes.