6 Conclusions
We use shallow borehole recordings of 4537 earthquakes at Parkfield, California, to investigate scaling, and spatio-temporal variation of stress drop. We find:
  1. Correction for spatio-temporal variation in material properties is needed to improve the accuracy of stress drop estimations. In the Parkfield region, corrections for variations in attenuation and rupture velocity with depth provide the most significant improvements to the stress drop estimates.
  2. The frequency bandwidth of the signal is important. Insufficient bandwidth can lead to systematic underestimation, and increased scatter of corner frequency and hence stress drop estimates. Interpretation of stress drop distributions should only include events within the resolution limit of the available dataset.
  3. Apparent depth and magnitude dependence can be caused by poor correction for material properties and the limited frequency range. We see neither dependence at Parkfield.
  4. At Parkfield we observe strong, stable spatial variability at small scale, with no significant dependence on the nature of fault slip (creeping versus locked) or the slip distribution of the 2004 M6 earthquake.
  5. Temporal variations of stress drop following the 2004 M6 earthquake are smaller than the spatial variations, but can be observed in regions with sufficient number of events and well-resolved coseismic/post-seismic slip.