TREMOR speeds up finding earthquake patterns in huge seismic data
TREMOR: Template Matching for Large Seismic Data Collections
Databases
Summary
Seismologists collect huge amounts of data to detect earthquakes and other ground movements. To find repeating events quickly, they look for examples called templates in this data. The authors created TREMOR, a system that splits the work across many computers so this search happens much faster. Their tests on real earthquake data show TREMOR can be up to 12 times quicker than other methods while giving the same exact matches.
What this means in practice
- •For seismic monitoring teams: Speed up searching large earthquake sensor data for known events using scalable distributed processing.
- •For geophysical data centers: Handle and analyze growing seismic waveform collections efficiently without single-machine memory limits.
Authors
Manos Chatzakis, Rodrigo Flores-Allende, Mikael Freire, Yoann Cano, Léonard Seydoux, Themis Palpanas
Abstract
Seismic station networks continuously record the ground velocity at several locations on earth in the form of waveform data series, which seismologists analyze to detect various kinds of geophysical events, including earthquakes. Some of these events are of particular interest: they are called templates and are used to search the seismic data collections for matching, similar events. This is known as template matching, and is a fundamental task in seismology, serving as the backbone for various seismic analyses. However, template matching requires extensive processing times, especially for seismic collections that exceed the memory capacity of a single machine. This poses a significant challenge to seismologists and is becoming worse as the seismological datasets continue to grow in size. In this paper, we introduce TREMOR, a distributed data series processing framework for template matching, designed to efficiently handle large waveform collections. We apply TREMOR to two representative real-world seismic use cases for template matching and, through an extensive experimental evaluation, we demonstrate its efficiency, with TREMOR being up to 12x faster than the best competing method, while returning the same, exact results.