Monitoring of the Long Valley Caldera Inflation Episode from a Principal Component Analysis-based Inversion Method applied to InSAR and EDM Data

Interferometric Synthetic Aperture Radar (InSAR) is used to explore and monitor surface deformation associated with a variety of, including tectonic, anthropogenic and glacier phenomena. Several studies have proposed methods of establising InSAR time series, for example the small baseline subset algorithm (SBAS, Berardino2002). These methods, however, either require tropospheric corrections on the InSAR images ahead of time or assume that tropospheric signals are negligible during the periods of acquisitions. In this study we apply the principal component analysis-based inversion method (PCAIM, Kositsky2009) on InSAR time series. PCAIM can be used to extract from the original data the signals that are coherent both in time and in space. Because tectonic and non-tectonic signals have different properties of coherency, they contribute different amounts to each component. By inverting and summing the components with the largest portion of tectonic signal separately with known Green’s functions and discarding those that appear to primarily be noise, we can rebuild the InSAR time series with primarily tectonic signals. We test this method in the Long Valley Caldera area which experienced multiple inflation episodes since the 80's. We focus on the 1997-98 episode, using 24 ERS scenes and 65 interferograms. To give the decomposition more temporal resolution and continuity, we add 8 two-color electronic distance meter (EDM) time series which have dense temporal sampling rates to carry out joint decomposition and inversion. The result shows that the first principal component contains most of the inflation signals with a clear pulse in 1997-98. A direct inversion using Mogi's model shows the inflation of magma near 10 km in depth. This study proves the capability of PCAIM to model and interpret InSAR time series and perform true joint inversions between multiple data source.