Skip to content

Computer vision geodesy and simulations of caldera collapse cycles at Kīlauea

Josh Crozier 1, Paul Segall1, Kyle Anderson2, Anna Melega3, Matthew Patrick4

  • Affiliations: 1Department of Geophysics, Stanford University, Stanford, USA; 2California Volcano Observatory, U.S. Geological Survey, Moffett Field, USA; 3Department of Geophysics, Colorado School of Mines, Golden, USA; 4Hawaiian Volcano Observatory, U.S. Geological Survey, Hilo, USA

  • Presentation type: Poster

  • Presentation time: Thursday 16:30 - 18:30, Room Poster Hall

  • Poster Board Number: 55

  • Programme No: 3.11.15

  • Theme 3 > Session 11

Abstract

A growing number of basaltic caldera collapse observations have demonstrated the fundamental couplings between collapse and eruption dynamics. For example, every recent collapse has occurred via a sequence of large earthquakes that repressurize magma reservoirs, suggesting that a delicate balance between fault friction and effusion rate controls eruption progression. However, limited data have been available to resolve important factors such as fault slip distributions and inelastic rock deformation. A three-month-long eruption at Kīlauea in 2018 provided the best monitored collapse, with caldera subsidence captured by two GNSS stations and several digital elevation models from overflights. We obtain further resolution by creating a 'video geodesy' dataset with continuous imagery collected by the Hawaiian Volcano Observatory from multiple cameras on the caldera rim. We compute ground deformation by using optical flow methods to track feature movement over time and then projecting pixel offsets into 3D models. This gives both broad spatial coverage and high sample rates. We resolve vertical and horizontal displacements across the caldera on scales from 0.1-10 m. The imagery helps constrain ring fault slip distributions during and between collapse earthquakes, and provides preliminary evidence for significant off-fault inelastic deformation between collapse events. We compare these data with caldera collapse earthquake cycle simulations to infer rock rheology and ring fault properties, with the aim of revealing how large amounts of caldera subsidence between earthquakes relates to fault friction. This work thus shows the utility of webcam imagery for deformation monitoring and provides insights into caldera collapse eruption mechanics.