Earthquake and Tsunami early Warning
We have developed the G-FAST early warning module that is currently being incorporated into the USGS ShakeAlert early warning system and NOAA's Tsunami Warning Centers. Additional code changes are being made to ensure a stable and robust earthquake characterization within a few minutes of rupture initiation.
G-FAST papers: Crowell et al. (2016), Crowell et al. (2018a), Crowell et al. (2018b)
Seismogeodetic analyses of large earthquakes
We have studied large events using joint datasets of seismic and geodetic data. Sometimes this involves combining data streams directly (1 Hz GNSS with 100 Hz strong-motion accelerations), and other times we perform joint inversions with GNSS, InSAR, and seismic data. All these datasets are sensitive to different parts of the rupture process, so the joint analyses reveal with higher fidelity what exactly is going on.
Examples can be found in: Goldberg et al. (2020), Melgar et al. (2015), Melgar et al. (2013), Crowell et al. (2013)
automated transient detection
One of the main challenges in understanding long-term tectonic deformations in GNSS time series is removing all earthquake and slow-slip signals. Earthquakes are fairly simple as only large events (M>5.5) cause offsets in the time series and can be added manually to a jump table, however, with slow-slip and other transient features, there is no appreciable seismic signal or it cannot be used as an automated diagnostic. We are working on methodologies, both single station and network-based, to coax out these small signals automatically so we can differentiate between different tectonic processes.
An overview of the financial technical indicator, Relative Strength Index (RSI), applied to GNSS time series can be found in Crowell et al. (2016)
Peak ground displacements
We developed a scaling law that relates the peak ground displacements (PGD) recorded on high-rate GNSS with the distance and the moment magnitude of the event. We are further refining this to include more complex terms commonly used in ground motion prediction equations (GMPEs) and also adding in seismic recordings. We also found a similar relationship between peak ground strains recorded on Plate Boundary Observatory borehole strainmeters.
The PGD scaling approach was first proposed in Crowell et al. (2013), and refined in Melgar et al. (2015) and Ruhl et al. (2018). The strain scaling method is outlined in Barbour and Crowell (2017).
Global advocacy of GNSS observations
We participate in several working groups and organizations to ensure wider adoption of GNSS data in hazards monitoring. We work with UNESCO-IOC to improve regional tsunami resilience in New Zealand, the Tonga-Kermadec trench, Ecuador, Colombia and Central America, and have worked with the World Bank to advise on early warning in Indonesia. We also worked with the Chilean Seismic Network on geodetic early warning.
Machine learning in geodetic data analysis
We are working towards improving standard GNSS positioning and deformation characterization through machine learning approaches.
Global Monitoring with GNSS
Through a USGS NEHRP grant, we built a real-time database with server-side tools for the National Earthquake Information Center. This keeps 5 days of high-rate data for over 1400 stations globally in MongoDB. If a big earthquake strikes near a station, an analyst can quickly obtain the waveforms and perform some basic modeling. The end goal is to incorporate this information into the standard event overview pages.
Real-time positioning of GNSS data runs into many issues with solution stability due to path delays from satellite to receiver. By variometrically processing the data, that is taking a single difference in time between orbital positions and phase observables, we obtain a highly precise broadband velocity record. We are working on ways to incorporate this data into standard seismic network analyses and early warning