PALM

Phase Picking, Association, Location, and Matched Filter workflow for building high-resolution earthquake catalogs. PAL or AI-PAL detections can provide the event templates used by the matched-filter (MFT) stage.

PALM v5.0 reorganizes the package into numbered executable workflows and shared PAL and MFT source packages. It adds compact dual-rate NPY template storage, buffered daily MFT scanning, high-resolution phase refinement, and final MFT association products ready for hypoDD relocation.

PALM workflow

PALM/
|-- 1_run_pal/ Numbered current PAL workflows for local or AWS execution
|-- 2_run_mft/ Numbered template-selection and matched-filter workflows
|-- 3_location/ Relocation tools for matched-filter detections
|-- PAL_src/ Shared current PAL implementation
|-- MFT_src/ Shared CPU/GPU MFT matched-filter implementation
`-- References/ Documentation files

Executable directories contain case settings, input metadata, and numbered launchers. Shared source directories contain implementation code and should not be edited or overwritten when starting a new experiment.

The PAL stage is synchronized with the current AI-PAL implementation. It includes daily parallel picking, optional separate association, pre-QC STA/LTA trigger inventories, current S picking, AWS waveform access, and resumable AWS jobs. See 1_run_pal/README.md.

PALM/1_run_pal and AI-PAL/1_run_pal use the same PAL implementation and produce the same scientific output contract for identical inputs and settings: daily accepted-pick files, STA/LTA trigger-count sidecars, subnet and merged catalog/phase files, association-rate tables, and completion metadata. Their installation roots, logs, job metadata, and configured output locations may differ, so whole output trees are not expected to be byte-for-byte identical.

For a local example:

cd 1_run_pal/run_pal_local
python 1_run_pal_pick_assoc_eg.py

The split workflow runs the same scientific implementation:

python 2.1_run_pal_pick_eg.py
python 2.2_run_pal_assoc_eg.py

A common self-supervised workflow is:

1. Run PAL to obtain conservative phase detections and association rates.
2. Use those PAL products to cut training samples and train AI-PAL pickers.
3. Run the trained AI-PAL models on the continuous archive and associate, optionally repick/reassociate, and locate the enhanced detections.
4. Use the located AI-PAL events as the MFT templates.

This lets PAL provide the initial training supervision while the broader AI-PAL catalog supplies a larger template bank for MFT. A direct PAL-to-MFT path remains useful as a conservative baseline or when no trained AI-PAL model is available. In 2_run_mft/1_select_templates_<case>.py, set TEMPLATE_SOURCE = "ai-pal" or "pal" and provide the matching detection and located phase files in TEMPLATE_INPUTS.

The detection and located files must describe the same source catalog. The selector uses the located event IDs to recover names from the corresponding detection file; mixing a PAL detection file with an AI-PAL location file (or the reverse) is invalid.

The MFT stage performs conventional multi-station matched-filter detection, followed by cross-correlation P- and S-pick refinement. Its input is a located PAL phase file and the corresponding continuous waveforms.

cd 2_run_mft
python 1_select_templates_eg.py
python 2_cut_templates_eg.py
python 3.1_run_mft_gpu_eg.py
# or: python 3.2_run_mft_cpu_eg.py

Edit the user-settings block in each numbered launcher and the model parameters in config_<CASE_CODE>.py. Launchers select that config without copying it into MFT_src/, so multiple case workdirs can safely share one installed source. See 2_run_mft/README.md.

The MFT stage associates duplicate template detections and writes catalog.csv, phase.csv, event.dat, and dt.cc. 3_location/hypodd/ consumes the latter two files and runs hypoDD relocation. Its external executable path and case-specific location settings remain user configuration because hypoDD is installed separately.

The local PAL and MFT workflows require Python, NumPy, SciPy, ObsPy, and PyTorch. GPU MFT additionally requires a CUDA-compatible PyTorch installation. External location workflows require their corresponding hypoInverse or hypoDD binaries.

• Zhou, Y., H. Ding, A. Ghosh, and Z. Ge (2025). AI-PAL: Self-Supervised AI Phase Picking via Rule-Based Algorithm for Generalized Earthquake Detection. Journal of Geophysical Research: Solid Earth. doi:10.1029/2025JB031294
• Zhou, Y., A. Ghosh, L. Fang, H. Yue, S. Zhou, and Y. Su (2021). A High-Resolution Seismic Catalog for the 2021 MS 6.4/Mw 6.1 Yangbi Earthquake Sequence, Yunnan, China. Earthquake Science, 34(5), 390-398. doi:10.29382/eqs-2021-0031
• Zhou, Y., H. Yue, L. Fang, S. Zhou, L. Zhao, and A. Ghosh (2021). An Earthquake Detection and Location Architecture for Continuous Seismograms: Phase Picking, Association, Location, and Matched Filter (PALM). Seismological Research Letters, 93(1), 413-425. doi:10.1785/0220210111