Information regarding the spatio-temporal progression of afterslip has the potential to shed light on the underlying processes that give rise to postseismic deformation and aftershocks, as well as fault frictional properties. A strong postseismic deformation signal was recorded by creepmeters and continuous Global Positioning System (GPS) stations starting in the minutes to hours following the 2004 M6 Parkfield earthquake. Several studies have shown that the observed displacements were likely due to afterslip, and the estimated moment release of this postseismic slip was anomalously large, exceeding the coseismic moment. The degree to which the spatio-temporal postseismic slip distribution inferred from geodetic data may be used to address questions relating to underlying processes depends upon which features of the imaged slip history are well-resolved.
Through a series of simulations and assessments of model resolution we have explored to what extent it is possible to accurately recover the spatial distribution of slip through inversion of GPS data given the available station coverage at Parkfield. We applied the findings of these assessments to construct a model fault geometry, comprised of variably-sized subfaults, that better reflects the resolving power of the data, particularly the limited resolution below 6 km. This fault geometry has the added benefit that it lessens the computational burden by reducing the number of model parameters. Although features of the inferred slip distribution with a spatial scale less than several kilometers are not resolvable, we find that the spatial resolution using this fault geometry is sufficient to characterize differences in the slip estimated at several points on the fault surface. Using the revised fault geometry we also compare two approaches to the time-dependent inversion of the continuous GPS data recorded at 14 sites in the Parkfield area during the first 60 days following the earthquake. The first approach uses a Kalman filter to recover the time history of afterslip, while the second involves fitting the postseismic time series with a modified Omori’s law and then inverting the displacement predicted by the modified Omori parameters at each epoch sequentially to obtain a slip history. The results of these analyses are consistent with earlier work suggesting that afterslip evolved to largely surround an area of large coseismic slip but also highlight the difficulty in recovering the high slip rate which occurred in the early postseismic period.