A dynamic extreme value model with applications to volcanic eruption forecasting

TL;DR

This paper adapts a dynamic extreme value model, originally used for financial risk, to forecast volcanic eruptions, demonstrating its effectiveness in identifying and predicting extreme natural hazard events.

Contribution

It introduces a novel application of a dynamic EVT-based model for volcanic eruption forecasting, expanding its use beyond financial risk analysis.

Findings

The EVT-informed thresholds improve extreme event detection.

The model demonstrates accurate eruption forecasts.

Application to Piton de la Fournaise validates its effectiveness.

Abstract

Extreme events such as natural and economic disasters leave lasting impacts on society and motivate the analysis of extremes from data. While classical statistical tools based on Gaussian distributions focus on average behaviour and can lead to persistent biases when estimating extremes, extreme value theory (EVT) provides the mathematical foundations to accurately characterise extremes. In this paper, we adapt a dynamic extreme value model recently introduced to forecast financial risk from high frequency data to the context of natural hazard forecasting. We demonstrate its wide applicability and flexibility using a case study of the Piton de la Fournaise volcano. The value of using EVT-informed thresholds to identify and model extreme events is shown through forecast performance.