Crowd-Sourced Global Earthquake Early Warning

Monday, 15 December 2014
Sarah E Minson1, Benjamin A Brooks2, Craig L Glennie3, Jessica R Murray4, John O Langbein2, Susan E Owen5, Bob A Iannucci6 and Darren L Hauser3, (1)California Institute of Technology, Pasadena, CA, United States, (2)US Geological Survey, Menlo Park, CA, United States, (3)National Center for Airborne Laser Mapping, Houston, TX, United States, (4)US Geological Survey, Earthquake Science Center, Menlo Park, CA, United States, (5)Jet Propulsion Laboratory, Pasadena, CA, United States, (6)Carnegie Mellon University Silicon Valley, Moffett Field, CA, United States
Although earthquake early warning (EEW) has shown great promise for reducing loss of life and property, it has only been implemented in a few regions due, in part, to the prohibitive cost of building the required dense seismic and geodetic networks. However, many cars and consumer smartphones, tablets, laptops, and similar devices contain low-cost versions of the same sensors used for earthquake monitoring. If a workable EEW system could be implemented based on either crowd-sourced observations from consumer devices or very inexpensive networks of instruments built from consumer-quality sensors, EEW coverage could potentially be expanded worldwide. Controlled tests of several accelerometers and global navigation satellite system (GNSS) receivers typically found in consumer devices show that, while they are significantly noisier than scientific-grade instruments, they are still accurate enough to capture displacements from moderate and large magnitude earthquakes. The accuracy of these sensors varies greatly depending on the type of data collected. Raw coarse acquisition (C/A) code GPS data are relatively noisy. These observations have a surface displacement detection threshold approaching ~1 m and would thus only be useful in large Mw 8+ earthquakes. However, incorporating either satellite-based differential corrections or using a Kalman filter to combine the raw GNSS data with low-cost acceleration data (such as from a smartphone) decreases the noise dramatically. These approaches allow detection thresholds as low as 5 cm, potentially enabling accurate warnings for earthquakes as small as Mw 6.5. Simulated performance tests show that, with data contributed from only a very small fraction of the population, a crowd-sourced EEW system would be capable of warning San Francisco and San Jose of a Mw 7 rupture on California’s Hayward fault and could have accurately issued both earthquake and tsunami warnings for the 2011 Mw 9 Tohoku-oki, Japan earthquake.