HR: 0800h
AN: G21A-0136    [Abstracts]
TI: The Network Strain Filter - A new tool for monitoring and detecting transient deformation signals in GPS arrays
AU: * Ohtani, R
EM: ryohta@pangea.stanford.edu
AF: Stanford University, Department of Geophysics, Stanford, CA 94305 United States
AU: Segall, P
EM: segall@pangea.stanford.edu
AF: Stanford University, Department of Geophysics, Stanford, CA 94305 United States
AU: McGuire, J
EM: jmcguire@whoi.edu
AF: Woods Hole Oceanographic Institution, Department of Geology and Geophysics, Woods Hole, MA 02543 United States
AB: Data from large-scale continuous GPS arrays have revealed transient signals caused by aseismic fault slip and magmatic intrusion. It is almost certain that more subtle signals due to smaller magnitude events exist, but have gone undetected. The huge amount of data from large GPS arrays makes it difficult to search for time varying processes by eye, and automated methods to detect transients are urgently required. We have developed a time-domain filtering method to detect spatially and temporally coherent signals in data from large GPS arrays, which we refer to as a Network Strain Filter (NSF). The position time series are expressed as a sum of tectonic motions, local benchmark motions, reference frame errors, and unmodeled errors. The tectonic displacements are expanded in a wavelet basis with time varying coefficients. The coefficients are modeled as integrated random walks, such that the velocities (and strain rates) follow a random walk. The wavelet coefficients, as well as the benchmark motion and reference frame errors, are estimated with Kalman filtering techniques. The variance of the integrated random walk, which controls temporal smoothing of the displacements, is also estimated simultaneously using an "extended" Kalman filtering algorithm. Orthogonal wavelets are used to form the basis, which enables us to carry out a multiresolution analysis (MRA). MRA expands a signal into a sum of translations and dilations of a unit "mother" wavelet and a sum of translations of one "father" wavelet whose scale corresponds to the spatial extent of the network. The smallest scale wavelet is chosen so that the wavelet covers some minimum number of stations, which ensures that time-varying deformation is separable from local benchmark motions. We determine the minimum scale such that the residual variance is consistent with the a priori errors of the data. We carried out simulations to test the performance of the filter using a variety of wavelets. Hypothetical time-varying deformation fields were constructed for GPS arrays in Japan and California. Preliminary results show that with appropriate initialization of the temporal smoothness hyperparameter, the filtered displacement field recovers the pattern of the input transient. This demonstrates the capability of the NSF to separate transient from steady state deformation. Analysis of GPS array data from Japan will be discussed.
DE: 8110 Continental tectonics--general (0905)
DE: 1206 Crustal movements--interplate (8155)
DE: 1242 Seismic deformations (7205)
DE: 1243 Space geodetic surveys
SC: Geodesy [G]
MN: 2004 AGU Fall Meeting