Project Earthquake Early Warning Networks

  • Earthquake Early Warning

    You don't need a smartphone to tell you when an earthquake strikes, BUT, what if you had 40 secs to prepare for an earthquake, what could you do? In the Bay Area you could shut down BART; In the Bay Area and else where, where earthquakes strike, you could shut down or bring industrial processes into safe mode, notify first responders and hospitals, get children in schools to safety? The benefits in terms of social capital and human safety can be easily imagined.

  • Seismic Science @ UC Berkeley

    DT's SVIC collaborating with the University of California, Berkeley's Seismology Department to develop a community-based early warning system utilizing smart phones as a very large sensor network capable of early detection. Using advanced neural networks to classify a variety of signals generated from a range of activities, the UC Berkeley algorithm achieved a 99.8% classification accuracy in distinguishing earthquake motion from other detected motion.

  • SVIC-UCB Partnership

    DT SVIC has been part of this effort with UC Berkeley since 2012 and focused on both the handset technology as well as network efficiency and communication. Berkeley developed the intelligence that makes the early warning possible. We developed the application that captures and bundles the smartphone's sensor data and optimized the time it takes to communicate the data from the handset to the back-end server. Together, UC Berkeley and we are working to develop the next generation of earthquake early warning networks.

  • When Accuracy Counts

    The graphic above shows the results of the ANN algorithm, developed by UC Berkeley, in classifying smartphone sensor data across a broad set of activities, including driving, walking, running, and riding bicycles with smartphones being carried in pockets and elsewhere, as well as with smartphones sitting on desks and tables, and being subject to testing on high precision earthquake shake tables. These very same signals will be generated by smartphones in everyday life and processed by the ANN. Because the results of the classification will be used to automatically signal industrial, transportation, and emergency responder systems, for example, accuracy is paramount to achieving a very high degree of public safety in an earthquake emergency. With that said, UC Berkeley’s results, having achieved 99.8% classification accuracy, is our starting point.

  • When Speed Counts

    Because time is of the essence in earthquake prediction and response, shaving secs off of the amount of time it takes for signals to reach the back end processing is a primary goal of the end-to-end system. Phase I results, depicted by the graph above, shows the effect of optimizing signal data transit time from the Smartphone to the back end, giving more time, from secs to millisecs, to classify and therefore alert. As an example, in the California Bay Area it takes 7 secs to signal the BART trains to slow down to safe speeds; with an advanced warning of only 40 secs and sub-second communication time, the ability to avoid BART derailment is great. Phase II addresses device management so that critical signal information is communicated, while flooding and non-essential signaling is eliminated, thereby preserving network bandwidth and allowing other communications to the degree possible.

  • When Global Reach Matters

    No longer will early detection be confined to a relatively few, fixed, high performance and expensive earthquake sensors. UC Berkeley has demonstrated that the sensors in a smartphone can be used as part of an Earthquake Early Warning system. Smartphones, are becoming commonplace in North and South America, Europe, Asia, the Middle East, in established and emerging economies: world wide penetration is 30% and rising. Smartphones “come with” the requisite devices and communication interfaces to become an ad-hoc and vast community network of sensors capable of providing timely early warning. Network infrastructure and globally accessible servers provide rapid analysis and alert distribution with the added value that earthquake waveform data from across the globe becomes accessible to researchers, likewise enhancing the opportunities for rapid evolution and enhancement of detection algorithms.