Outrunning Seismic Waves with the Next Generation Seismic Monitoring Systems
Abstract
Modern seismic monitoring systems empower us to essentially 'outrun' seismic waves: with seismic real-time data we can design Earthquake Early Warning (EEW) algorithms that detect and characterize earthquakes right as they unfold. If a detected earthquake is large enough, alerts can be provided to areas that are at risk, often - but not always - before the damaging seismic waves arrive. Providing useful EEW alerts remains a challenging task, however, because of the high propagation speed of seismic waves, and because of the high variability of seismic signals. In order to be fast enough, the real-time alerts have to be based on very little input data, which brings with it a high risk of data misinterpretations, e.g. in the form of false alerts.
In this talk I will discuss how supervised machine learning techniques allow us to strongly improve the automation of various seismic monitoring procedures, and to provide faster and more reliable EEW alerts. While automated monitoring techniques have been successfully used in earthquake seismology for decades, the resulting data products have always been of lesser quality than those produced by trained human experts. The fundamental change that state-of-the-art machine learning algorithms bring about is that this class of algorithms can now reach, and sometimes even exceed, the performance levels of trained human experts. The machine learning-powered monitoring modules we are developing form the building blocks for a next generation of seismic monitoring and inference systems. These intelligent, fully automated and highly robust systems will not only facilitate actionable real-time alerts but also provide a fundamentally improved observational foundation for a wide range of geophysical and natural hazards questions.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFMNH11A..05M
- Keywords:
-
- 4313 Extreme events;
- NATURAL HAZARDS;
- 4314 Mathematical and computer modeling;
- NATURAL HAZARDS;
- 4318 Statistical analysis;
- NATURAL HAZARDS;
- 4339 Disaster mitigation;
- NATURAL HAZARDS