2 Data
We select our study area as an 80 km section of the San Andreas Fault centered on the shallow borehole High Resolution Seismic Network (HRSN, Karageorgi et al., 1992; Malin et al., 1989). This includes parts of the ‘creeping zone’ to the northwest and the ‘locked zone’ to the southeast of Parkfield (Harris & Segall, 1987; Murray & Langbein, 2006). We download the triggered vertical-component waveforms at the 13 HRSN stations for all the earthquakes in the Northern California double-difference catalog (WNC catalog, Waldhauser, 2009; Waldhauser & Schaff, 2008), from Northern California Earthquake Data Center (NCEDC), see Figure 1). We include only earthquakes between 2001 and 2016 (the time of our download); the recording system was less consistent before 2001, with more frequent recording system changes and lower dynamic range (which caused clipping of large earthquake records). The waveforms have sampling rate of 250 Hz and we apply the corrections for gain changes included in the instrument response table at the NCEDC. We use P wave phase picks in the NCEDC database, and use an STA/LTA based auto-picker (Li & Peng, 2016) to determine automatic picks for any waveforms without catalog arrival times.
We calculate the P-wave displacement spectra required for our analysis, using 1-second time windows, starting 0.1 s before the P wave arrival pick (see example in Figure S1). We compute the multi-taper spectral density and convert the recorded velocity spectra to displacement spectra. There is no need to correct for the instrument responses as these are included in the site terms in the spectral decomposition analysis. We also calculate the noise spectra using 1-second windows immediately preceding the signal windows, following the same approach. We select the earthquakes with sufficiently high-quality recording for the spectral analysis. We require an earthquake to be recorded by at least 5 stations, each with signal-to-noise ratio (SNR) higher than 10 at each frequency point between 2 and 60 Hz. The 4537 earthquakes between Mw0 and Mw4 that meet our data selection criteria are shown in Figure 1.