HiSelect

Python package for automated seismic station selection based on SNR, waveform cross-correlation, epicentral distance, and azimuthal coverage.

Given a directory of ZRT SAC files for a single event, HiSelect scores every candidate station and iteratively picks the subset that jointly maximises data quality and azimuthal coverage — ready for moment-tensor inversion or other waveform-based analyses.

Selection criteria

Metric	Description
SNR	Signal-to-noise ratio over the surface wave window vs. pre-P noise
CC	Peak cross-correlation with the azimuth-sector waveform stack
Distance score	Magnitude-dependent epicentral distance weight (Scognamiglio et al. 2009)
Azimuthal coverage	Effective station number (Ekström 2006, Eq. 10)

Station 1 is chosen by the highest combined score. Each subsequent station minimises the combined rank of score and azimuthal coverage, so the selected set fills azimuthal gaps while maintaining high data quality.

Requirements

• Python ≥ 3.8
• ObsPy ≥ 1.4
• NumPy
• scikit-learn
• Matplotlib

Optional:

• Cartopy — for geographic station maps (falls back to plain Matplotlib if absent)
• pysep — for writing CAP-format weights*.dat files (falls back to a built-in writer if absent)

Installation

git clone git@github.com:Jinyin-Hu/HiSelect.git
cd HiSelect
pip install -e .

Input data

HiSelect reads three-component (Z, R, T) SAC files from a single event directory. Files must follow the naming convention:

{eventid}.{net}.{sta}.{loc}.{band}{comp}.sac

For stations with multiple channel types, the priority order BH > HH > HN is applied automatically.

Example

See the script for an example.

Output files

File	Description
station_file.txt	One line per selected station: STA NET LAT LON DIST BAZ
weights.dat	CAP-format weight file
weights_body.dat	CAP-format body-wave weights
weights_surf.dat	CAP-format surface-wave weights
selected/	Copy of selected SAC files and all output files
hiselect_map.png	Station map (selected in blue, unselected in grey)
hiselect_scores.png	Per-station SNR, distance, CC, and combined score bar charts
hiselect_waveforms.png	Record-section plot of selected stations' signal windows

References

• Hu, J., Tkalčić, H., Phạm, T.-S. & Hejrani, B., 2026. Improved joint constraints on moment tensors, source time functions and uncertainty quantification for moderately large earthquakes. Geophys J Int, 246, ggag154. doi:10.1093/gji/ggag154