Introduction

G-FAST (Geodetic First Approximation of Size and Timing) was developed by Brendan Crowell as a Python based geodetic earthquake early warning (EEW) module. Funding from the Amazon Catalyst program has allowed the Pacific Northwest Seismic Network (PNSN) and ISTI to developed a compiled-language production-grade variant of G-FAST for integration into the USGS's next generation ShakeAlert EEW system. The software is released under ?? license.

Nominally, G-FAST acquires real-time geodetic data and, when triggered by a ShakeAlert message, activates three geodetic co-seismic inversion suites

1. A peak ground displacement (PGD) inversion
2. A centroid moment tensor (CMT) inversion
3. A CMT driven slip or finite fault inversion.

The PGD module provides a quick and robust estimate of magnitude. The CMT inversion provides a robust estimate of magnitude, depth, and moment tensor components. The moment-tensor derived fault planes and depth are then passed onto the slip inversion which resolves the CMT fault plane ambiguity while providing an additional magnitude estimate. The PGD and slip model are then passed onto the ShakeAlert decision module as the event unfolds.

As in the Python implementation, G-FAST still acquires real-time data from the Pacific Northwest Geodetic Array (PANGA) and, in the context of EEW, is triggered on ShakeAlert messages. The key differences are now that an Earthworm server is responsible for the import of PANGA data and the ShakeAlert communication is directly implemented via ActiveMQ. Additionally, G-FAST has been separated into a library and application to encourage reuse of routines by users with varying data sources and objectives.