Towards scientific forecasting of magmatic eruptions

Forecasting eruptions is a fundamental goal of volcanology. However, difficulties in identifying eruptive precursors, fragmented approaches and lack of resources make eruption forecasting difficult to achieve. In this Review, we explore the first-order scientific approaches that are essential to progress towards forecasting the time and location of magmatic eruptions. Forecasting in time uses different monitoring techniques, depending on the conduit-opening mode. Ascending magma can create a new conduit (closed-conduit eruptions), use a previously open conduit (open-conduit eruptions) or flow below a solidified magma plug (semi-open-conduit eruptions). Closed-conduit eruptions provide stronger monitoring signals often detected months in advance, but they commonly occur at volcanoes with poorly known pre-eruptive behaviour. Open-conduit eruptions, associated with low-viscosity magmas, provide more subtle signals often detected only minutes in advance, although their higher eruption frequency promotes more testable approaches. Semi-open-conduit eruptions show intermediate behaviours, potentially displaying clear pre-eruptive signals days in advance and often recurring repeatedly. However, any given volcano can experience multiple conduit-opening modes, sometimes simultaneously, requiring combinations of forecasting approaches. Forecasting the location of vent opening relies on determining the stresses controlling magma propagation, deformation and seismic monitoring. The use of physics-based models to assimilate monitoring data and observations will substantially improve forecasting, but requires a deeper understanding of pre-eruptive processes and more extensive monitoring data.Volcanic eruptions are major natural hazards, but forecasting their activity remains challenging. This Review discusses scientific and monitoring approaches used to forecast magmatic eruptions.Magmatic eruptions are usually preceded by some level of volcanic unrest, meaning a deviation of monitoring parameters from the baseline, a condition that can provide eruptive precursors.Different types of precursors can be identified among the monitoring signals, mainly depending on the conduit-opening type feeding the eruption ('conduit-opening mode'). The three main modes are open conduit, semi-open conduit and closed conduit.The conduit-opening mode affects the intensity of the monitoring signals, provided that an adequate monitoring system is available, and hence the warning time to eruption and ultimately the effectiveness of an alert.Understanding of unrest processes and knowledge of specific volcanoes (structure and dynamics of the plumbing system, magma composition and volatiles, pre-eruptive history) also provide appropriate input to operational aspects, improving eruption forecasting.Forecasting the location of opening of eruptive vents is moving from description-based to physics-based models that consider the regional and local stress field, in addition to monitoring data.Better eruption forecasting will rely on merging physics-based models with real-time data assimilation, to provide probabilities and uncertainties.