Skip to content
Daniel Wamriew
Projects
Research software

ae-picker

An open-source Python package for automatic first-arrival picking in acoustic emission, microseismic and seismic waveforms, developed from the HydraFrac project and published on PyPI.

Illustrative
Problem
Picking the first arrival on every channel of every event is the bottleneck between raw acoustic emission recordings and located sources, and doing it by hand does not scale.
Approach
Interpretable, physically parameterized pickers (AIC and a refined STA/LTA with envelope-based onset refinement), frequency-adaptive presets, batch processing, quality control and plotting, packaged as a library and a command-line tool.
My contribution
Designed, implemented and released the package, from the picking algorithms to the readers, presets, QC, plotting and command-line interface.

Overview

ae-picker is an open-source Python package for automatic first-arrival time picking in acoustic emission (AE), microseismic and seismic waveform data. It grew out of the analysis needs of the HydraFrac laboratory experiments and is general enough for field data: the same pickers run on laboratory AE recordings, sonic logs and refraction surveys.

pip install ae-picker
ae-picker path/to/your/data_dir --save-plots

Problem addressed

Before an event can be located, the onset of the first wave on each sensor has to be picked precisely. A single experiment can produce thousands of events across many channels. Manual picking is slow and inconsistent, and black-box automatic pickers are hard to trust or tune when they fail.

Technical approach

Two complementary pickers, both with optional moveout-guided search windows that limit picking to physically expected arrival times:

AIC picker. Splits the waveform at sample kk into a noise segment and a signal segment, and picks the split that minimizes the Akaike Information Criterion (after Maeda, 1985):

AIC(k)=klog⁡ ⁣(var⁡(x0:k))+(N−k−1)log⁡ ⁣(var⁡(xk+1:N))\mathrm{AIC}(k) = k \log\!\big(\operatorname{var}(x_{0:k})\big) + (N - k - 1)\log\!\big(\operatorname{var}(x_{k+1:N})\big)

Refined STA/LTA picker. A recursive STA/LTA trigger finds a coarse onset window. A Hilbert-envelope stage then walks back to the earliest persistent rise above a threshold set from robust noise statistics (median absolute deviation), which improves onset timing for emergent arrivals.

Architecture and methodology

  • Frequency-adaptive presets. suggest_parameters returns STA/LTA windows and thresholds calibrated to the sampling rate and dominant frequency, with profiles for AE (hundreds of kHz to MHz), PS logs and refraction surveys, and auto-detection from the sampling rate.
  • Readers. AE text formats, plus SEG-Y and miniSEED through optional ObsPy support.
  • Batch processing. Processes whole directories with custom readers and file globs, and writes a picks_summary.csv.
  • Quality control. Fits a linear or robust (Theil–Sen) moveout model across traces and flags picks whose residuals exceed a MAD-based threshold.
  • Plotting. Waveform review, picker comparison, wiggle sections, STA/LTA diagnostics and spectrograms.
  • Interfaces. A Python API and an ae-picker command-line tool.

My contributions

I designed and implemented the package, released it on PyPI (versions 0.2.0 and 0.2.1 in May 2026, 0.3.0 in November 2026) and archived it on Zenodo with a citation file.

Results

ae-picker is freely available under the MIT license for Python 3.10 and later. If you use it, please cite: Wamriew, D. (2026). ae-picker. Zenodo. https://doi.org/10.5281/zenodo.20126965