Smartphone-Based Earthquake Early Warning in Chile Abstract

abstract

  • Chile faces a high seismic hazard with magnitude 7.5+ earthquakes occurring roughly every 1.5 years. In addition to posing a trans-Pacific hazard due to tsunamis, the effects of large Chilean subduction zone earthquakes are especially severe on local populations and infrastructure. Accordingly, Chile has a need for both earthquake early warning and local tsunami early warning. Early warning systems require dense networks of seismic and geodetic sensors and the Chilean Centro Sismologico Nacional is in the process of establishing this infrastructure with scientific grade instrumentation. In parallel, we are developing and deploying an early warning system in Chile using only smartphones and an inexpensive GNSS add-on to incorporate SBAS corrections. Our design uses inexpensive sensor boxes, each containing a smartphone and an external consumer-quality GPS chip and antenna: the total cost of each box is on the order of a few hundred dollars, nearly two orders of magnitude lower than scientific-grade installations. The smartphone uses an Android application to collect data from the smartphone's onboard accelerometer and the GPS chip. The application analyzes and transmits relevant data to a central server where we use the FinDer-BEFORES algorithm to detect earthquakes and produce a real-time joint seismic-geodetic finite-fault distributed slip model (for sufficiently large magnitude earthquakes) or a near-field acceleration-based line source model (for smaller magnitude earthquakes). Accurate ground shaking forecasts could be provided by either earthquake source model, and distributed slip models for larger offshore earthquakes could be used to infer seafloor deformation and thus provide local tsunami warning. Our goal is to build and deploy over 200 smartphone-based monitoring stations this year. Although this project utilizes the smartphone-based sensor as part of a fixed network, the approach, building on our earlier work, could also be implemented in a crowd-sourced manner. In November, 2015, the first 8 sensor units were installed in the region of the 2015 Mw 8.3 Illapel, Chile earthquake. Retrospective batch processing all of the data collected from these few sensors shows that our proposed analysis method successfully detects, locates, and estimates the magnitude for both Mw>5 earthquakes that have occurred since the sensors were deployed while producing zero false alarms. We expect the system's performance to improve significantly once the sensor network is expanded beyond this nominal initial deployment.

authors

publication date

  • 2016

presented at event