lfpspy - A Python package for low-frequency passive seismic (LFPS) method processing

👨‍💻 lfpspy is an open-source Python package I developed in 2024 to process Low-Frequency Passive Seismic (LFPS) ambient-noise data. It was the primary output of my undergraduate thesis project in Geophysics at Gadjah Mada University, born out of both necessity and curiosity.

🔁 Background

This journey began with my desire to apply LFPS methods to map subsurface hydrocarbon potential. However, early in the research, I ran into a major obstacle—there were no open-source tools available to handle LFPS data. Unlike conventional passive seismic processing (e.g., for HVSR or microtremor data), LFPS posed unique challenges in terms of preprocessing, spectral analysis, and window-quality control. I realized I’d either have to abandon the method or build the tools myself—so I chose the latter.

My inspiration came from a paper by Cox et al. (2020), which proposed a frequency-domain window-rejection algorithm for improving H/V spectral ratio estimates by removing noisy or inconsistent windows before averaging, which they implement it on hvsrpy. While their method had been widely used in microtremor and earthquake engineering studies, it hadn’t yet been applied to LFPS data. I saw this as an opportunity to innovate: I designed lfpspy to adapt and extend Cox et al.’s method specifically for the lower frequency bands (0.1–1.0 Hz) typical of LFPS surveys.

The project culminated in a full case study on the Majalengka Sub-basin, Indonesia. Using LFPS recordings acquired from a short-period seismic station, I processed the data through lfpspy, computed stable V/H curves (VHSR), and identified dominant low-frequency peaks associated with soft sedimentary layers. I then validated these results using polarization attribute and cross-compared them with gravity anomaly maps from previous studies. The results were promising—showing that LFPS, when combined with careful spectral QC and directional analysis, could serve as a cost-effective method for preliminary subsurface assessment.

⚙️ Methodology

The core of lfpspy consists of:

• Spectrum Analysis: Computes Power Spectral Density on the vertical component (PSD‑Z) and Horizontal‑to‑Vertical Spectral Ratio (H/V) to identify low‑frequency anomalies associated with reservoir rock.

• Window‑Rejection Algorithm: Implements Cox et al.’s, which uses statistical thresholds in the frequency domain to reject contaminated windows before averaging, boosting signal‑to‑noise robustness and significantly improving the clarity of low-frequency V/H peaks.

Frequency-domain Window-rejection Algorithm Flowchart (modified from Cox et al.)

• Polarization Analysis: Performs Principal Component Analysis (PCA) to evaluate particle motion in sliding time windows. This yields dip, azimuth, rectilinearity, and eigenvalue ratios—important for distinguishing noise from signal and validating spectral peaks.

🧾 Citation

• Cox, B. R., Cheng, T., Vantassel, J. P., & Manuel, L. (2020). “A statistical representation and frequency-domain window-rejection algorithm for single-station HVSR measurements. Geophysical Journal International, 221(3), 2170–2183. https://doi.org/10.1093/gji/ggaa119
• Joseph Vantassel. (2020). jpvantassel/hvsrpy: latest (Concept). Zenodo. http://doi.org/10.5281/zenodo.3666956