Insights from Mapping Distributed and Focused Volcanism

Loading...
Thumbnail Image

Date

2024-08-07

Authors

Bussard, Rebecca

Journal Title

Journal ISSN

Volume Title

Publisher

University of Oregon

Abstract

This dissertation utilizes a variety of mapping techniques to explore surface and subsurface processes at different volcanic systems. As each volcanic system is unique, it is important to understand which mapping methods can be applied to vents contained within the system and which will face more challenges. This dissertation details the use of statistical analysis and remote sensing for addressing questions of magma transport/storage as well as volcanic surface change at distributed and focused systems.To begin, Voronoi tessellations are used to map the area between vents in a variety of distributed volcanic fields. The distributions of vent areas are then compared to distributions of areas between randomly simulated vents. If the distribution of vent areas across a volcanic field differs from a random distribution, then clustering is occurring in this field; from this, causes of clustering such as regional tectonic forces, magma supply rates, and magma storage can be explored. Five of the six fields analyzed experience clustering and visualize the length scale at which clustering occurs through Kernel Density Estimation. The work then shifts from distributed to focused volcanism, specifically Mt. St. Helens, a stratovolcano in southwest Washington. A computationally inexpensive neural network developed with open access code classifies snow cover in optical imagery covering the Mt. St. Helens region through time. The snow cover estimates produced through classification are then compared with coherence maps from Sentinel-1 Interferometric Synthetic Aperture Radar (InSAR) data to quantify how snow cover effects coherence (signal strength). Snow cover reduces coherence up to 70% and widespread snow cover can almost entirely mask out uplift greater than one centimeter from an inflating magma source. Finally, InSAR timeseries and velocity data over Medicine Lake volcano measures ground deformation from 2017-2021. The vertical velocity data shows subsidence across the broad edifice that increases in magnitude within the volcano’s central caldera to ~ 1 cm/yr. Markov Chain Monte Carlo (MCMC) modeling constrains several parameters of a potential volume loss source at depth beneath the volcano, including depth and volume change for a point source and depth, length, wide, opening, and strike for a rectangular sill. The highest likelihood point source sits at 7.7 km depth with a volume decrease of 0.0013 km3/yr, and for the sill source sits at 10.1 km depth with a volume decrease of 0.0016 km3/yr. Subsidence due to edifice loading is also analytically modeled and is assumed to occur (subtracted from the InSAR vertical deformation signal). When the MCMC is rerun taking loading into account, source depths become shallower and volume changes decrease. This dissertation includes previously submitted and unpublished co-authored material.

Description

Keywords

deformation, neural network, satellite, snow-cover, volcano

Citation