Mining Continuous GPS Data for Crustal Transients

William Holt, Patrick Abejar
Department of Geosciences, Stony Brook University

In this paper we discuss a method for evaluating continuous GPS time series for detection of crustal transients. Time series of displacement are first fit with a fourth order polynomial that also contains annual and semiannual terms. Fitted time series of displacements, with annual and semiannual terms removed, yield continuous time series of velocity. We output velocity values every 0.05 years and evaluate the data for southern California from 2000.0 – 2010.0. For each epoch we invert model velocity field estimates for a velocity gradient tensor field estimate within Southern California, where basis functions are defined using bi-cubic splines with spherical treatment. Optimal smoothing of model solutions is obtained through adjustment of the expected strain rate variances until the reduced Chi-squared misfit with the continuous estimates of GPS velocity are 1.0. A reference solution is subtracted from each epoch solution to obtain a residual strain rate solution, where the reference solution was obtained from the SCEC4.0 GPS data set for Southern California. Initial investigation for transients is performed by first determining total accelerations of difference strain rates, the total finite strain tensor field, and total finite displacements, obtained by integrating the difference strain rates. These enable a quantitative measure of spatial and temporal coherence, along with formal uncertainties of potential transient signals. Results to date show strong and clear patterns following the Hector Mine, San Simeon, and Parkfield earthquakes along with transients south of the Salton Sea. Distributions of strain rates following Hector Mine are heterogeneous in the far-field with apparent and surprising large anomalies within the Los Angeles Basin regions.