Skip to content

Input and Preprocessing

Input: Continuous Seismic Data

  • SAC or MiniSEED formats

    • Run FAST independently for each component at each station
    • Filename should have station, component, date-time

      Warning

      Filename must end with “.sac” or “.mseed” – if not, rename it!

    • FAST uses ObsPy functions to read in data

  • Time gaps ok, but must not be filled with 0’s

    • 0’s are similar to other 0’s --> time gaps overwhelm your earthquake detections, and take forever to run
    • Run scripts to fill 0’s with random uncorrelated noise; outputs files starting with “Filled*”
    • $ python fill_time_gaps_uncorrelated_noise.py calls detect_time_gaps.py
    • Need to modify for your input data
    • Example scripts in: parameters/preprocess_utils/

Preprocessing

Bandpass Filter: important parameter

  • FAST uses short sections of spectrogram to create fingerprints

    • Nyquist upper limit = (Sampling rate)/2
    • Typically sampling rate is 100 Hz, but depends on data
  • Frequencies outside passband are cut, thrown away

  • Demean, detrend data before applying filter

Decimate to lower sampling rate

  • Depends on your selected filter band; usually to 50 Hz, 25 Hz, or 20 Hz

Remove spikes and glitches (write your own script)

Example Python scripts in parameters/preprocess_utils/ (uses ObsPy):

  • bandpass_filter_decimate.py, mseed_bandpass_filter_decimate.py
  • SaudiMonth_bandpass_filter_decimate.sh: calls Python script multiple times for different stations and components, different filter bands