Estimating time-dependent magma source properties at the Long Valley caldera using InSAR and GPS time series

Thursday, 18 December 2014: 11:50 AM
Bryan V Riel1, Mark Simons1 and Piyush S. Agram2, (1)California Institute of Technology, Pasadena, CA, United States, (2)NASA Jet Propulsion Laboratory, Pasadena, CA, United States
For the past few decades, the Long Valley caldera in California has experienced several inflation and deflation episodes leading to uplift and subsidence, respectively, of the resurgent dome. The abundance of geodetic measurements from leveling, microgravity, GPS, two-color electronic distance meter (EDM), and interferometric Synthetic Aperture Radar (InSAR) time series has allowed the geodetic community to closely monitor surface deformation of the caldera. However, estimation of the source properties of the underlying magma chamber has led to a number of different models that often disagree over important features, such as chamber depth, chamber shape, influence of the south moat fault on the southern edge of the caldera, etc. These discrepancies can often be attributed to strong non-linearities of the different source models and trade-offs between various parameters. Furthermore, underestimation of the observation errors and uncertainties regarding the elastic structure can lead to biases and overconfidence of the inversion results. We address these issues by studying the 2002-2003 inflation episode due to the availability of GPS, EDM, and InSAR time series for that time period. We first use a transient detection approach that incorporates compressed sensing principles to compactly reconstruct temporally and spatially coherent transient signals for the GPS and EDM data. A similar approach is used to construct a continuous InSAR time series formed from over 200 ERS and Envisat interferograms over a 20-year time period. We compute static offsets for the 2002-2003 inflation event and perform a fully Bayesian analysis to estimate parameter values and uncertainties for various non-linear source models. Additionally, we also test a linear source model consisting of a grid of point volume-sources. By again using compressed sensing principles, we show that the linear model is largely consistent with the more compact non-linear analytical models. We use the linear model in a time-dependent fashion to characterize subtle magma migration for the 2002-2003 inflation episode using the reconstructed continuous time series for all data sets. We also extend the inversion to simultaneously estimate time-dependent aseismic slip on the south moat fault.