Near-real-time matched-filtering for the development of dense earthquake
catalogs during sequences of seismicity
Abstract
Conventional earthquake detection methods suffer significant
degradations in completeness during high-rate sequences such as
aftershock sequences or volcanic swarms. Missed earthquakes during the
early periods of aftershock sequences can affect aftershock forecasts
and hazard estimates. Missed events during volcanic unrest sequences can
impact rate estimations, leading to the sequence being
mis-characterized. Much recent work has addressed how matched-filters
can be used to overcome some aspects of catalog incompleteness during
high-rate sequences, by detecting similar events using
cross-correlation. Here we describe the application of open-source (GPL
v3.0) software to the near-real-time implementation of matched-filter
earthquake detection. Our software (RT-EQcorrscan) is written in Python,
and leverages the extensive Obspy package, as well as EQcorrscan and
Obsplus to provide matched-filter methods and database handling
respectively. RT-EQcorrscan is designed to be modular, so that users can
readily utilize only the components they require, or make use of
pre-built command-line utilities controlled by simple that can handle
thousands of templates over tens of channels of seismic data within the
processing capacity (memory and CPU usage) of a standard desktop
personal computer. Detections are made within a few seconds of data
arriving, with latency due to data delivery and a requirement for full
network move-out. At the same time, RT-EQcorrscan has an overarching
“Reactor” module to listen to a web-service and respond to new events.
If an event occurs that meets user-defined criteria, the Reactor will
initiate a near-real-time matched-filter process encompassing the region
surrounding the trigger event. Subsequent trigger events in different
regions can also be handled with threaded operations. This system is
backed-up by a constantly updating template database built on Obsplus,
allowing groups of templates to be rapidly deployed. In this
presentation we will discuss the key implementation details, as well as
showcasing some examples of the system in operation.