scripts

Generate seismic waveforms

In order to generate synthetic seismic waveforms, call the following on the command line and follow the instructions:

python generate_waveforms.py

Example calls

Create 32 samples using command line arguments:

python generate_waveforms.py \
  --hypocentral_distance 531 \
  --is_shallow_crustal 1 \
  --magnitude 6 \
  --vs30 154 \
  --num_samples 32 \
  --output waveforms.h5 \
  --edm_checkpoint ../weights/edm.ckpt \
  --autoencoder_checkpoint ../weights/autoencoder.ckpt

Create samples using a CSV file:

python generate_waveforms.py \
  --csv [japan.csv | little_japan.csv] \
  --output waveforms.h5 \
  --edm_checkpoint ../weights/edm.ckpt \
  --autoencoder_checkpoint ../weights/autoencoder.ckpt

Convert to SeisBench framework

To convert the generated waveforms.h5 file into SeisBench format, first prepare the station metadata consisting of a list of columns in .csv format, similar to the following figure:

Then, run the write_to_seisbench.py script by:

1. Inspecting the command:

python write_to_seisbench.py -h

2. Running the command:

python write_to_seisbench.py \
  <path_to_station_metadata>/station_metadata.csv \
  <path_to_gwm>/waveforms.h5 \
  --origin_time "2024-02-01T00:00:00.0" \
  --hypocenter 12 10 4 \
  --magnitude 6 \
  --num_realizations 10 \
  --trace_sampling_rate 100 \
  <path_to_seisbench>/events