← Back to Sessions

Smartphone-Based Earthquake Early Warning in Costa Rica

Session: Earthquake Early Warning System in the Americas: The On-Going Effort and the State of the Art II

Type: Oral

Date: 4/20/2021

Presentation Time: 10:00 AM Pacific

Description: 

We find that off-the-shelf smartphones deployed in a fixed network can provide earthquake early warning performance comparable to scientific-grade systems. In 2020 we constructed the ASTUTI (Alerta Sismica Temprana Utilizando Teléfonos Inteligentes) smartphone network of more than 80 stations affixed to walls in baseboards of buildings distributed throughout Costa Rica. We evaluate an accelerometer-based non-parametric ground-motion detection and alerting strategy that considers the dominant source of shaking hazard to be Middle America Trench subduction zone earthquakes. Our strategy focuses on Costa Rica’s densely populated interior region near San Jose city where most of the population experiences shaking during earthquakes >M6. We evaluate issuing alerts when an acceleration threshold is exceeded at four stations. For 6 months of operation and a small group (15) of participants, data latency over cell-phone networks (phones to processing machine) is acceptably low, 0.35-0.45 secs. From on-phone vibration simulation of the 2012 M7.6 Nicoya earthquake we find median first-alert latency of ~9-13 secs and alert-receipt latency of ~4 secs. Greater latency may occur when alerting larger numbers of recipients. 13 earthquakes caused felt shaking “did-you-feel-it” (DYFI) reports in Costa Rica in our evaluation period. We detected and alerted on 5 of these events (offline re-runs), each of which produced shaking in San Jose. Our system issued no false alerts and undetected events were out-of-network and did not produce wide-spread felt shaking near San Jose. We further evaluate alerting the entire country after detecting an earthquake and find that 15-70% of Costa Rica’s population could receive alerts preceding shaking with enough time to carry out drop-cover-hold-on (DCHO). As suggested by current social science research, if populations tend to be false-alarm tolerant, this more inclusive alerting approach may be effective for populations with geographic distribution and hazard exposure similar to Costa Rica.

Presenting Author: Benjamin A. Brooks

Student Presenter: No

Authors