A major objective of expanding continuous GPS networks is to enable observation of time-varying deformation. Transient displacements have been observed in data from several existing networks, largely in subduction zone settings, and have been used to infer the spatio-temporal slip-rate history on faults.
Implementation of the Plate Boundary Observatory will soon provide a much larger dataset, motivating the development of increasingly sophisticated analyses both in terms of the models employed and the interpretation of results. Questions remain, however, regarding 1) proper characterization of noise, 2) the variety of transient deformation events that may be imaged (for instance in terms of tectonic setting, magnitude, and duration), and 3) how inferences of slip-rate history may be used to better understand mechanisms of fault slip in different settings. Here we present a study that begins to address these issues. Parkfield, CA, a locale well-known for its history of M6 earthquakes, lies at the transition between creeping and locked segments of the San Andreas fault. In the early 1990s, rate changes were observed in data from several monitoring networks at Parkfield and were accompanied by three moderate earthquakes (approximately M4.5) which nucleated near the inferred hypocenter of previous Parkfield M6 earthquakes.
We have analyzed data from the permanent two-color EDM network using a nonlinear Kalman filtering technique. This method is well-suited to inferring slip-rate changes from data contaminated by random walk and white noise. Furthermore we have implemented a means for simultaneous estimation of the seasonal noise that is common in geodetic time-series. This type of analysis provides a fault slip estimate at every epoch for which there are data, resulting in a time-history of slip rather than a single rate change.
Our results show that fault slip-rate on the San Andreas between Middle Mountain and Parkfield increased following the three moderate earthquakes, perhaps as a response to these events. Similarly, a slow earthquake has been inferred from strainmeter data following a Mw 5.1 earthquake in 1998 at San Juan Bautista, site of another transition between creeping and locked behavior on the San Andreas. The setting of the inferred time-varying slip at Parkfield is clearly different from that of many recently observed transients in subduction zones, although it is as yet unknown whether the underlying mechanisms are similar. To better understand the relation between seismic and aseismic slip we consider the effect of stress changes imposed by the three earthquakes and explore the evolution of shear stress on the fault during the transient event.