Physics‐Based Forecasts of Eruptive Vent Locations at Calderas

Constraining stresses in the Earth's crust in volcanic regions is critical for understanding many mechanical processes related to eruptive activity. Dike pathways, in particular, are shaped by the orientation of principal stress axes. Therefore, accurate models of dike trajectories and future vent locations rely on accurate estimates of stresses in the subsurface. This work presents a framework for probabilistic constraint of the stress state of calderas by combining three‐dimensional physics‐based dike pathway models with observed past vent locations using a Monte Carlo approach. The retrieved stress state is then used to produce probability maps of future vent opening across a caldera. We test our stress inversion and vent forecast approach on synthetic scenarios, and find it successful depending on the distribution of the available vents and the complexity of the volcano's structural history. We explore the potential and limitations of the approach, show how its performance is sensitive to the assumptions in the models and available prior information, and discuss how it may be applied to real calderas. Plain Language Summary Many processes at volcanoes are influenced by the stress state in the subsurface, which results from various forces, such as the gravitational loading due to redistribution of mass following the formation of topographic features. In particular, magma often opens its own pathway from a magma chamber to the surface along trajectories that are sensitive to the distribution of stress within the rocks. Therefore, knowledge of the stress state in volcanoes is critical to understand, and possibly forecast, magma pathways and the locations where they may reach the surface. In this work, we develop a framework to constrain the stress state at calderas. We estimate the stress field so that it is consistent with pathways that link the magma chamber with the eruptive vents of past eruptions. Then, we employ the estimated stress state to simulate future magma pathways and study the expected distribution of the resulting eruptive vents across the caldera, identifying the areas which are more likely to host future eruptions. We test our approach on artificially generated scenarios and explore the potential, limitations, and possible future applications of our work. Key Points We use past vent locations and a Bayesian framework to constrain the stress state at calderas with a physics‐based model of dike pathways We exploit the posterior information