Seismic_BPMF

Complete framework for earthquake detection and location with GPU-accelerated processing.

Backprojection and matched-filtering (BPMF) is a two-step earthquake detection workflow with 1) backprojection for template finding and 2) template matching for lowering the magnitude of completeness of the catalog. BPMF offers a number of routine for the automatic location of the detected events with the deep neural network phase picker PhaseNet and the earthquake locator NLLoc. BPMF leverages the low-level C and CUDA-C programming languages for the efficient processing of large data volumes. The core routines for backprojection and template matching are provided in our two packages beampower and fast_matched_filter, respectively.

BPMF v2.0.0-alpha is now out! Checkout the online tutorial at https://ebeauce.github.io/Seismic_BPMF/tutorial to learn how to use our fully automated workflow and build your own earthquake catalog.

Documentation

Check out the online documentation at https://ebeauce.github.io/Seismic_BPMF/index.html.

Installation

Download or clone the repository. Go to the root folder, activate your virtual environment, and execute the following command lines:

    python setup.py build_ext
    pip install .

The first line, python setup.py build_ext, executes the Makefile and compiles the C and CUDA-C librairies. If you don't have an Nvidia GPU and/or the nvcc compiler, you will see a warning message (and every time you will load the BPMF librairy). You can still use BPMF on CPUs.

Details on how to set up a working environment at https://ebeauce.github.io/Seismic_BPMF/tutorial/general.html.

Reference

Please, cite:

Beaucé, E., Frank, W. B., Paul, A., Campillo, M., & van der Hilst, R. D. (2019). Systematic detection of clustered seismicity beneath the Southwestern Alps. Journal of Geophysical Research: Solid Earth, 124(11), 11531-11548.

and/or

Beaucé, E., van der Hilst, R. D., & Campillo M. (2022). Microseismic Constraints on the Mechanical State of the North Anatolian Fault Zone Thirteen Years after the 1999 M7.4 Izmit Earthquake. Journal of Geophysical Research: Solid Earth. DOI: https://doi.org/10.1029/2022JB024416.

If you use this package for your research.

Note: Our paper Beaucé et al., 2022 (see References below) was prepared with BPMF v1.0.1, than you can find at https://github.com/ebeauce/Seismic_BPMF/releases/tag/v1.0.1.

To do:

PhaseNet and NLLoc independent relocation method.
Convert availability and source_receiver_dist to properties.
Robust and fast detection threshold for template matching.
Use Seisbench for the easier interfacing of ML pickers.
Convert moveouts and weights to xarray-like objects with explicit indexing using station names?

Contact

Questions? Contact me at:
ebeauce@ldeo.columbia.edu