Skip to content

How do we move closer to near real-time volcanic hazard and risk estimation?

Julie De Groote , Benoit Taisne, Susanna Jenkins

Abstract

The impacts of future volcanic eruptions are expected to amplify due to global population growth, driving more people to settle on hazardous lands, and modern society's rising vulnerability to volcanic hazards. However, those future eruption impacts (e.g., material, non-material, human) are challenging to forecast robustly. One of the reasons is that, when forecasting impacts, we often consider a future volcanic eruption as an instantaneous event at a specific point in time, rather than considering the dynamic, coincident, and cascading impacts associated with a potentially multi-phase and multi-hazard volcanic eruption. To more accurately constrain forecasts of impact and loss resulting from eruptions, we ideally need to incorporate these dynamic aspects into our impact forecasts. One approach is to move closer to near real-time volcanic hazard and risk estimations, which use real-time data to rapidly and robustly forecast imminent impacts. In this project, we will extend our current machine-learning model based framework, which uses monitoring data (i.e., progressively added seismic, deformation, degassing data with time; before and during an eruption) to optimise the forecasting of volcanic eruption onsets. We will investigate how well monitoring data can forecast eruption duration and the changing characteristics of activity within a multi-phase eruption to inform dynamic and updateable forecasts of potential hazard. By combining this with developments on dynamic and interacting volcanic eruption impacts (i.e., studying the relationships between impacts from different/coinciding/cascading hazards and the forecasted eruption parameters), we aim to develop a dynamically updating framework for optimised impact estimation.