Register    Attendee Sign In Sign in to access papers, presentations, photos and videos
Previous Abstract Return to Session B1b Next Abstract

Session B1b: Precise GNSS Positioning and Applications

GNSS Earthquake Early Warning In the Cloud
David Mencin, Tim Dittmann, Alex Hamilton, Charlie Sievers, Henry Berglund, Rachel Terry, Rowan Gaffney, Mike Gottlieb, Tim Ronan, Chad Trabant, Bill Fasbinder, Matt Collins, Jerry Carter, Rob Casey, EarthScope Consortium
Location: Beacon A
Date/Time: Tuesday, Jan. 23, 4:23 p.m.

Effective earthquake early warning (EEW) systems provide populations valuable seconds that save lives and infrastructure costs by mitigating the surprise of destructive ground shaking from an earthquake. The USGS ShakeAlert is an operational EEW system in the Western United States that includes spatially distributed sensor networks continuously transmitting seismic data to centralized EEW decision modules. Over 15 years of research has led to the inclusion of low-latency, high-rate GNSS position streams as a unique sensor source to improve this operational system. Inclusion of GNSS mitigates rapid magnitude estimate saturation of the most energetic and destructive earthquakes that occurs using only traditional, inertial-based seismic sensors.
The effectiveness of the GNSS component to inform the ShakeAlert decision module is dependent on: 1) ingesting GNSS observables from spatially distributed, heterogenous receivers in independently operated GNSS networks for 2) low latency real-time PPP processing to reliably deliver time-aligned topocentric displacements. The existing cyberinfrastructure (CI) ‘plumbing’ connecting disparate network operators, geodetic processing centers and decision modules has been developed incrementally as a proof-of-concept within existing data streaming architectures, but did not offer sufficient fault tolerance, scalability or efficient maintainability for this integral component of a safety-of-life system.
Fortunately, event-streaming architectures to support such requirements is now relatively commonplace in the Internet of Things explosion of large-scale, high-throughput data pipelines. We present on a commercial cloud event streaming platform EarthScope has developed through support by a USGS interagency agreement. In this development effort, Earthscope has leveraged an open-source distributed event streaming framework coupled with Amazon Web services high-availability, serverless compute for ingesting raw GNSS observables and PPP streams from distributed sources. This platform normalizes raw and processed GNSS streams into a single, consolidated access point to high throughput, low latency streams (currently n=~3500). Furthermore, adoption of this mature event streaming CI ecosystem supports efficient code deployments of value-add real-time stream processing and real-time dashboard visualizations for enhanced EEW network monitoring and EEW CI observability.
We report on the resulting DevOps pipeline architecture’s performance in meeting the demanding requirements for successful hazard mitigation in ShakeAlert’s geodetic component. We also present on next steps, including tighter CI coupling of raw data streams and PPP processing, additional sensor stream input capacity, and data-driven augmented stream processing (MLOps) for faster and more reliable input streams to EEW modules. Finally, we report on the benefits of this generalized CI strategy for future support of additional real-time geophysical applications for hazards monitoring, efficient network operations and next-generation science.

Previous Abstract Return to Session B1b Next Abstract