4. Core

Core routines of the EQcorrscan project. These routines concern the large-scale matched-filter detection of near-repeating earthquakes. For this to take place a series of templates must be generated: these templates can be created using the functions in template_gen, or you can generate them externally. Templates in EQcorrscan are simply cut waveforms. Because EQcorrscan relies on Obspy, these waveforms can be any type readable by Obspy. For simplicity, portability and data compression reasons, EQcorrscan currently stores templates on-disk as multiplexed miniseed files, where each file is a single template.

4.1. Brightness

bright_lights contains a series of functions to detect events using the brightness-based beamforming method of Frank et. al (2014). This has been tested significantly, but has failed to detect events unambiguously in the central Southern Alps. As such development of these functions has ceased.

4.2. Template generation

template_gen contains routines for cutting waveforms around picks for use as templates in match_filter. Included in this are wrappers to directly read in Seisan formattaed pick files and waveforms associated with the picks, and generate templates from these. There are also wrappers for quakeML events and catalogs, and seishub databases.

4.3. Matched-Filter

match_filter contains the core routines for earthquake detection by cross-correlation. This is optimized for large-scale, multi-paralleled detection, with large numbers of templates. Because we are unsure of your architecture we have not written functions for the top level of possible parallel computing, which would be to compute detections for multiple days in parallel in a High-Performance Computing, cluster environment. If you want to know more about doing this please contact the authors. We use a cluster running SLURM for job scheduling and handle multiple days using the batch job submission capability which distributes daily detections across multiple nodes. This allows us to detect earthquakes through > 6 years of multi-channel data using > 600 templates in less than 36 hours.

Of note: EQcorrscan does not enforce a length of data to process, it is up to the user to exercise caution when thresholding cross-correlation sums. As the figure below shows, if using the median absolute deviation (MAD) thresholding metric the user should be aware that this changes with time, and those variations can be significant when using short windows of data.

plots/range_of_threshold_windows_Parkfield1.png

Plot of a few hours of cross-correlation sum (black) for a five channel template associated with the Parkfield 2004 earthquake with various windowed thresholds, set to 8 x Median Absolute Deviation. x-axis is time in hours, y-axis is cross-correlation sum.

4.4. Lag-Calc

lag_calc contains functions for generating pick-corrections from cross-correlations with a defined template. Originally this was designed for events detected by match_filter, however you can use any well correlated events. Based on the method of Shelly and Hardebeck (2010).

4.5. Subspace

subspace contains a subspace detector for either single-channel cases, or network cases. This is modelled on that described by Harris. This method allows for slightly more variation in detected waveforms than the traditional matched-filter method. In this method, templates are constructed either by using the empirical subspace method, or by computing the basis vectors by singular-value decomposition. Both methods are provided as part of EQcorrscan in the clustering module.