Overview of ShakeAlert EEW Algorithms and EEW Operations at Caltech

• ShakeAlert EEW Operations at Caltech
• EEW Algorithms in Current ShakeAlert Operations
  • EPIC (Earthquake Point-source Integration Code)
  • FinDer (Finite-fault Rupture Detector)
  • G-FAST (Geodetic First Approximation of Size and Time)
  • Solution Aggregator
  • eqInfo2GM (Earthquake Information to Ground Motion)
  • Decision Module
• EEW Algorithms in Development for Potential Future ShakeAlert Operations
  • PLUM (Propagation of Local Undamped Motion)
  • Ground Motion Aggregator (GMA)

ShakeAlert EEW Operations at Caltech

The ShakeAlert^® earthquake early warning (EEW) system currently takes a source-characterization-based approach to EEW. In this approach, the EEW algorithms that process the incoming station data use the ground-motion observations from an earthquake to estimate diagnostic characteristics of the earthquake (like its magnitude and source location). These earthquake source characteristics are then input into ground-motion models which calculate the expected distribution of ground shaking intensities, which form the basis of the alert information (called the ShakeAlert Message). When a ShakeAlert Message is published, external alert delivery partners issue alerts to recipients according to specific criteria related to the earthquake’s magnitude, location, and ground shaking distributions.

ShakeAlert-powered EEW alerts are determined by the information contained in the ShakeAlert Message, which is composed of three main message products:

• The Event Product contains information about the earthquake’s estimated magnitude, origin time, epicenter, and rupture information (if a large event).
• The Contour Product provides polygons of the estimated extents of specific Modified Mercalli Intensity (MMI), peak ground acceleration (PGA), and peak ground velocity (PGV) values.
• The Grid Product provides more detailed estimates of MMI, PGA, and PGV distributions for a 0.2°x0.2° grid of location coordinates.

In order to create the ShakeAlert Messages, the ShakeAlert EEW system is made up of several individual algorithms that perform different tasks in the alert generation process. Many of these algorithms were developed by researchers at the universities that operate the seismic networks (including Caltech and the SCSN). As part of the ShakeAlert project cooperative agreement with the USGS, these institutions are responsible for helping provide the real-time station data as well as for helping operate and improve the EEW algorithms that make up the ShakeAlert system.

Alongside these efforts, Caltech is also involved in the development and testing of other EEW methods that may be considered for integration into the ShakeAlert system in the future.

References

• McGuire, J.J., C.W. Ulberg, A.I. Lux, M. Böse, J.R. Andrews, D. Smith, B. Crowell, J.R. Murray, I. Henson, R. Hartog, et al. (2025). ShakeAlert version 3: Expected performance in large earthquakes. Bulletin of the Seismological Society of America, XX, 1-29. https://doi.org/10.1785/0120240189
• Lux, A.I., D. Smith, M. Böse, J.J. McGuire, J.K. Saunders, M. Huynh, I. Stubailo, J. Andrews, G. Lotto, B. Crowell, et al. (2024). Status and performance of the ShakeAlert earthquake early warning system: 2019-2023. Bulletin of the Seismological Society of America, 114(6), 3041-3062. https://doi.org/10.1785/0120230259
• Kohler, M.D., E.S. Cochran, D. Given, S. Guiwits, D. Neuhauser, I. Henson, R. Hartog, P. Bodin, V. Kress, S. Thompson, C. Felizardo (2018). Earthquake early warning ShakeAlert system: West Coast wide production prototype. Seismological Research Letters, 89(1), 99-107. https://doi.org/10.1785/0220170140
• Given, D.D., R.M. Allen, A.S. Baltay, P. Bodin, E.S. Cochran, K. Creager, R.M. de Groot, L.S. Gee, E. Hauksson, T.H. Heaton, M. Hellweg (2018). Revised technical implementation plan for the ShakeAlert system—An earthquake early warning system for the West Coast of the United States (No. 2018-1155). US Geological Survey. https://doi.org/10.3133/ofr20181155

EEW Algorithms in Current ShakeAlert Operations

EPIC (Earthquake Point-source Integration Code)

The EPIC algorithm is a source-characterization-based EEW algorithm that uses the first few seconds of P-wave observations to estimate an earthquake’s magnitude, epicenter, origin time, and corresponding uncertainties for these values. Magnitude is estimated in EPIC using an empirical scaling relationship to peak P-wave displacement observations and considers point-source assumptions. Because EPIC uses P-wave observations as its primary input data, EPIC is very sensitive and usually issues event detections a bit faster than other EEW algorithms.

EPIC is primarily based on the ElarmS EEW algorithm developed at UC Berkeley, and incorporates elements of the Onsite EEW algorithm developed at Caltech. The two algorithms were combined into a single EEW algorithm for ShakeAlert operations in 2018. Current operations and development of the EPIC algorithm is led by researchers at UC Berkeley.

References

• McGuire, J.J., C.W. Ulberg, A.I. Lux, M. Böse, J.R. Andrews, D. Smith, B. Crowell, J.R. Murray, I. Henson, R. Hartog, et al. (2025). ShakeAlert version 3: Expected performance in large earthquakes. Bulletin of the Seismological Society of America, XX, 1-29. https://doi.org/10.1785/0120240189
• Lux, A.I., D. Smith, M. Böse, J.J. McGuire, J.K. Saunders, M. Huynh, I. Stubailo, J. Andrews, G. Lotto, B. Crowell, et al. (2024). Status and performance of the ShakeAlert earthquake early warning system: 2019-2023. Bulletin of the Seismological Society of America, 114(6), 3041-3062. https://doi.org/10.1785/0120230259
• Kohler, M.D., D.E. Smith, J. Andrews, A.I. Chung, R. Hartog, I. Henson, D.D. Given, R. de Groot, S. Guiwits (2020). Earthquake early warning ShakeAlert 2.0: Public rollout. Seismological Research Letters, 91(3), 1763-1775. https://doi.org/10.1785/0220190245
• Chung, A.I., M.A. Meier, J. Andrews, M. Böse, B.W. Crowell, J.J. McGuire, D.E. Smith (2020). ShakeAlert earthquake early warning system performance during the 2019 Ridgecrest earthquake sequence. Bulletin of the Seismological Society of America, 110(4), 1904-1923. https://doi.org/10.1785/0120200032
• Chung, A.I., I. Henson, R.M. Allen (2019). Optimizing earthquake early warning performance: ElarmS‐3. Seismological Research Letters, 90(2A), 727-743. https://doi.org/10.1785/0220180192

FinDer (Finite-fault Rupture Detector)

The FinDer algorithm is a source-characterization-based EEW algorithm that tracks an earthquake’s growth in real time and produces estimates of the earthquake’s rupture (fault model) alongside traditional source estimates like magnitude, epicenter, and origin time. Rupture models are important when calculating shaking distributions, particularly for larger earthquakes (M>~6). This is because the severity of shaking in a location depends on the location’s proximity to the earthquake rupture. Larger earthquakes can rupture for many kilometers along a fault, so using just the epicenter location (where the earthquake began to rupture) as input to the ground-motion calculations may underpredict the shaking intensity distribution for these earthquakes.

FinDer estimates earthquake source parameters by matching the observed distribution of peak ground acceleration (PGA) at the seismic stations with pre-computed templates of PGA distributions from empirical ground-motion models. The best-matching template determines the location, length, and strike (orientation) of the earthquake rupture, parameters which define a fault model that can be used in downstream ShakeAlert algorithms for the alert region computation. FinDer estimates magnitude in two ways: using an empirical relationship to fault length (typically preferred for moderate and larger earthquakes, M>~5.5), or using an empirical regression of the PGA observations (typically preferred for smaller earthquakes). The current FinDer algorithm allows detecting and reporting events with magnitudes M≥2.5.

The FinDer algorithm has been part of the USGS ShakeAlert public alerting system since 2018. FinDer development is a collaborative effort, and ShakeAlert works with SED (Switzerland) on the current innovations. Many people at Caltech, SED, University of Washington, and USGS have contributed.

Ongoing FinDer Research at Caltech

While the FinDer EEW algorithm has been integrated into ShakeAlert operations since 2018, FinDer is continually undergoing assessments and improvements (as are the other aspects of ShakeAlert). Recent improvements to FinDer include adjustments that allow FinDer to handle multiple sets of ground-motion templates as well as the development of fault-specific templates. The primary template set in FinDer was created using simple (linear) fault geometry. Simple fault geometry makes these templates extremely versatile, as they can be applied to most earthquakes that occur within the ShakeAlert reporting region. However, if we know what faults are likely to host large earthquakes, such as the San Andreas fault and the Cascadia subduction zone, we can also create additional FinDer template sets that incorporate the location-specific geometry of these faults, which could make FinDer’s source estimates more accurate should a large earthquake occur on one of these faults. Fault-specific templates for the San Andreas fault and the Cascadia subduction zone are currently undergoing refinement and are slated for formal testing in the near future. Other FinDer developments include improving FinDer’s performance during complex earthquake sequences, testing potential updates to the ground-motion models used for FinDer’s templates, and investigating how site-specific ground-motion corrections can be incorporated into FinDer’s source estimation procedure.

References

• McGuire, J.J., C.W. Ulberg, A.I. Lux, M. Böse, J.R. Andrews, D. Smith, B. Crowell, J.R. Murray, I. Henson, R. Hartog, et al. (2025). ShakeAlert version 3: Expected performance in large earthquakes. Bulletin of the Seismological Society of America, XX, 1-29. https://doi.org/10.1785/0120240189
• Böse, M., S. Ceylan, J. Andrews, F. Massin, J. Clinton, J.K. Saunders, O. Tatar, M. Türkoglu (2024). Rapid Finite-Fault Models for the 2023 Mw 7.8 Kahramanmaras, Türkiye, Earthquake Sequence. Seismological Research Letters, 95(5), 2761-2778. https://doi.org/10.1785/0220230426
• Lux, A.I., D. Smith, M. Böse, J.J. McGuire, J.K. Saunders, M. Huynh, I. Stubailo, J. Andrews, G. Lotto, B. Crowell, et al. (2024). Status and performance of the ShakeAlert earthquake early warning system: 2019-2023. Bulletin of the Seismological Society of America, 114(6), 3041-3062. https://doi.org/10.1785/0120230259
• Böse, M., J. Andrews, R. Hartog, C. Felizardo (2023). Performance and Next‐Generation Development of the Finite‐Fault Rupture Detector (FinDer) within the United States West Coast ShakeAlert Warning System. Bulletin of the Seismological Society of America. https://doi.org/10.1785/0120220183
• Böse, M., J. Andrews, C. O’Rourke, D. Kilb, A. Lux, J. Bunn, J. McGuire (2023). Testing the ShakeAlert earthquake early warning system using synthesized earthquake sequences. Seismological Research Letters, 94(1), 243-259. https://doi.org/10.1785/0220220088
• Böse, M., D.E. Smith, C. Felizardo, M.-A. Meier, T.H. Heaton, J.F. Clinton (2018). FinDer v.2: Improved real-time ground-motion predictions for M2-M9 with seismic finite-source characterization. Geophysical Journal International, 212(1), 725-742. https://doi.org/10.1093/gji/ggx430
• Böse, M., C. Felizardo, T.H. Heaton (2015). Finite-Fault Rupture Detector (FinDer): Going Real-Time in Californian ShakeAlert Warning System. Seismological Research Letters, 86(6), 1692-1704. https://doi.org/10.1785/0220150154
• Böse, M., T.H. Heaton, E. Hauksson (2012). Real-time Finite Fault Rupture Detector (FinDer) for large earthquakes. Geophysical Journal International, 191(2), 803-812. https://doi.org/10.1111/j.1365-246X.2012.05657.x

G-FAST (Geodetic First Approximation of Size and Time)

G-FAST is a source-parameter-based EEW algorithm that uses geodetic observations from Global Navigation Satellite Systems (GNSS) instruments to estimate earthquake magnitude and other source parameters. Earthquake magnitude is estimated using scaling relationships to peak ground displacement (PGD) observations. These PGD-magnitude estimates can be very useful for large earthquakes (M>~7) where P-wave-based magnitude estimates may saturate or underestimate the earthquake’s magnitude. G-FAST can also use observations of permanent displacements caused by the earthquake (called coseismic offsets) in rapid finite-fault inversions to obtain an estimate of the fault orientation as well as how much slip was on the fault during the earthquake. These finite-fault slip models can be useful for refining ground-motion distribution estimates as well as for tsunami early warning applications.

The development of G-FAST has been led by researchers at the University of Washington (UW). A version of G-FAST that just produces the PGD-based magnitude estimates was developed specifically for ShakeAlert by researchers at UW and the USGS, and was officially incorporated into ShakeAlert operations in March 2024. In this implementation, G-FAST uses the combined epicenter and origin time estimates from the Solution Aggregator as the input source information for its magnitude estimation.

References

• McGuire, J.J., C.W. Ulberg, A.I. Lux, M. Böse, J.R. Andrews, D. Smith, B. Crowell, J.R. Murray, I. Henson, R. Hartog, et al. (2025). ShakeAlert version 3: Expected performance in large earthquakes. Bulletin of the Seismological Society of America, XX, 1-29. https://doi.org/10.1785/0120240189
• Murray, J.R., B.W. Crowell, M.H. Murray, C.W. Ulberg, J.J. McGuire, M.A. Aranha, M.T. Hagerty (2023). Incorporation of Real‐Time Earthquake Magnitudes Estimated via Peak Ground Displacement Scaling in the ShakeAlert Earthquake Early Warning System. Bulletin of the Seismological Society of America. https://doi.org/10.1785/0120220181
• Crowell, B.W., D.A. Schmidt, P. Bodin, J.E. Vidale, J. Gomberg, J. Renate Hartog, V.C. Kress, T.I. Melbourne, M. Santillan, S.E. Minson, D.G. Jamison (2016). Demonstration of the Cascadia G‐FAST geodetic earthquake early warning system for the Nisqually, Washington, earthquake. Seismological Research Letters, 87(4), 930-943. https://doi.org/10.1785/0220150255

Solution Aggregator

The Solution Aggregator algorithm takes the earthquake source parameter estimates (magnitude, epicenter, and origin time) produced by the individual EEW algorithms (EPIC, FinDer, and G-FAST) and combines them into a unified set of source parameter estimates. The Solution Aggregator optimally combines the point-source parameters in terms of the uncertainties from the individual EEW algorithms. The combined source parameters form the basis for the ShakeAlert Message event product, and are used as the input to the ground-motion models to calculate the ground-motion distributions used for the other ShakeAlert Message alert products. If the combined magnitude is above a given threshold, the FinDer fault line-source estimate will be included as part of the unified source estimate.

References

• McGuire, J.J., C.W. Ulberg, A.I. Lux, M. Böse, J.R. Andrews, D. Smith, B. Crowell, J.R. Murray, I. Henson, R. Hartog, et al. (2025). ShakeAlert version 3: Expected performance in large earthquakes. Bulletin of the Seismological Society of America, XX, 1-29. https://doi.org/10.1785/0120240189
• Lux, A.I., D. Smith, M. Böse, J.J. McGuire, J.K. Saunders, M. Huynh, I. Stubailo, J. Andrews, G. Lotto, B. Crowell, et al. (2024). Status and performance of the ShakeAlert earthquake early warning system: 2019-2023. Bulletin of the Seismological Society of America, 114(6), 3041-3062. https://doi.org/10.1785/0120230259
• Kohler, M.D., D.E. Smith, J. Andrews, A.I. Chung, R. Hartog, I. Henson, D.D. Given, R. de Groot, S. Guiwits (2020). Earthquake early warning ShakeAlert 2.0: Public rollout. Seismological Research Letters, 91(3), 1763-1775. https://doi.org/10.1785/0220190245

eqInfo2GM (Earthquake Information to Ground Motion)

The eqInfo2GM algorithm is designed to take in earthquake source parameters (magnitude, location, rupture information) from ShakeAlert algorithms and compute predicted ground motions (shaking intensity, peak ground acceleration, and peak ground velocity). eqInfo2GM uses published ground-motion prediction equations (GMPEs) and ground-motion-to-intensity conversion equations (GMICEs) to create maps of predicted ground motions as well as polygons enclosing regions expected to experience ground motions above different thresholds, which are used for the ShakeAlert Message grid and contour products, respectively.

Ongoing eqInfo2GM Research at Caltech

Researchers at Caltech are using shaking intensity observations from the USGS “Did You Feel It?” community science product to help re-evaluate the combination of GMPEs and GMICEs used in the ShakeAlert EEW alert region creation in the eqInfo2GM algorithm. Ground-motion model development, especially GMICEs, has traditionally focused on accurately modeling damaging ground-motion levels, as these ground motions are of greatest concern for seismic hazard analysis applications. Non-damaging shaking intensity levels are more difficult to model, as these are defined primarily by peoples’ experiences of the earthquake and do not easily relate to specific peak ground acceleration and velocity observations. Accurately modeling the extents of non-damaging shaking is important for ShakeAlert, as the EEW system uses alert thresholds that correspond to these lower intensity levels.

References

• McGuire, J.J., C.W. Ulberg, A.I. Lux, M. Böse, J.R. Andrews, D. Smith, B. Crowell, J.R. Murray, I. Henson, R. Hartog, et al. (2025). ShakeAlert version 3: Expected performance in large earthquakes. Bulletin of the Seismological Society of America, XX, 1-29. https://doi.org/10.1785/0120240189
• Lux, A.I., D. Smith, M. Böse, J.J. McGuire, J.K. Saunders, M. Huynh, I. Stubailo, J. Andrews, G. Lotto, B. Crowell, et al. (2024). Status and performance of the ShakeAlert earthquake early warning system: 2019-2023. Bulletin of the Seismological Society of America, 114(6), 3041-3062. https://doi.org/10.1785/0120230259
• Saunders, J.K., A.S. Baltay, S.E. Minson, and M. Böse (2024). Uncertainty in Ground-Motion-to-Intensity Conversions Significantly Affects Earthquake Early Warning Alert Regions. The Seismic Record, 4(2), 121-130. https://doi.org/10.1785/0320240004
• Thakoor, K., J. Andrews, E. Hauksson, T. Heaton (2019). From Earthquake Source Parameters to Ground‐Motion Warnings near You: The ShakeAlert Earthquake Information to Ground‐Motion eqInfo2GM Method. Seismological Research Letters, 90(3), 1243-1257. https://doi.org/10.1785/0220180245

Decision Module

The Decision Module algorithm receives the messages from the Solution Aggregator and eqInfo2GM algorithms and uses these to create the alert information as the ShakeAlert Message. The Decision Module then publishes the ShakeAlert Message if specific criteria are met, primarily if the magnitude estimate is above a set threshold and if the location estimate is within the ShakeAlert reporting region. The Decision Module also issues updates to the ShakeAlert Message for an earthquake if there are significant updates to the earthquake location and magnitude estimates. ShakeAlert alert delivery partners use the ShakeAlert Message information published by the Decision Module to issue alerts to EEW system users.

References

• McGuire, J.J., C.W. Ulberg, A.I. Lux, M. Böse, J.R. Andrews, D. Smith, B. Crowell, J.R. Murray, I. Henson, R. Hartog, et al. (2025). ShakeAlert version 3: Expected performance in large earthquakes. Bulletin of the Seismological Society of America, XX, 1-29. https://doi.org/10.1785/0120240189
• Kohler, M.D., D.E. Smith, J. Andrews, A.I. Chung, R. Hartog, I. Henson, D.D. Given, R. de Groot, S. Guiwits (2020). Earthquake early warning ShakeAlert 2.0: Public rollout. Seismological Research Letters, 91(3), 1763-1775. https://doi.org/10.1785/0220190245

EEW Algorithms in Development for Potential Future ShakeAlert Operations

PLUM (Propagation of Local Undamped Motion)

The PLUM algorithm takes a ground-motion-based approach to EEW. In this type of EEW approach, ground-motion distributions are determined by the station observations directly, without the need to estimate the earthquake’s magnitude and location as an intermediate step like for source-based methods. PLUM was originally developed in Japan and was incorporated into their EEW operations in 2018, where it works in tandem with source-characterization-based EEW algorithms.

A version of the PLUM algorithm is being adapted for the U.S. to operate under the desired alerting parameters for the ShakeAlert EEW system. Modifications include the adjustment of PLUM’s earthquake detection strategies to be sensitive to ground motions produced by smaller magnitude earthquakes of interest to the ShakeAlert EEW system while minimizing false detections on noise signals. The U.S. version of PLUM also projects its ground-motion estimates onto the same grid used for the ShakeAlert Message grid product.

Internal real-time testing of PLUM using the ShakeAlert data streams began in 2019. Like the other EEW algorithms already included in ShakeAlert operations, these internal tests have allowed for further refinement of the algorithm. The development of PLUM in the U.S. is led by researchers at the USGS and at Caltech.

Ongoing PLUM Research at Caltech (the APPLES configuration)

Ongoing research for the U.S. version of the PLUM algorithm includes investigating potential adjustments to PLUM’s ground-motion estimation methodology. In the original PLUM approach, the ground-motion observations at a given seismic station are forward-predicted to an area within a set radius of that station, and the maximum-predicted value at a location is used to determine the alert for that location. PLUM operations in Japan consider a relatively small prediction radius of 30 km, but because EEW alert regions in Japan (sub-prefectures) are larger than the alert grid used in the ShakeAlert grid product, a larger prediction radius of 60 km is needed to maintain the same alert performance for PLUM in the U.S. While this larger prediction radius was found to be optimal for most alerting needs, it can overwarn for smaller earthquakes where the spatial footprint of significant ground motions is smaller than the PLUM prediction radius.

Current research involves adding attenuation into PLUM’s ground-motion estimation procedure, where the estimated ground motion at a location varies according to the observed ground motion at the seismic station and the distance from the station to the location of interest; we call this new configuration Attenuated ProPagation of Local Earthquake Shaking (APPLES). This adjustment helps reduce overwarning during smaller magnitude earthquakes as well as potentially increase warning times in some areas during larger magnitude earthquakes, as available warning times would no longer be restricted by a fixed forward-prediction distance. Adding attenuation in PLUM’s ground-motion forward-prediction procedure will also produce ground-motion distributions that are more similar in style to ShakeAlert’s ground-motion distributions, and may make them easier to incorporate into ShakeAlert operations.

References

• Saunders, J.K., E.S. Cochran, J.J. Bunn, A.S. Baltay, S.E. Minson, C.T. O’Rourke (2024). Incorporating intensity distance attenuation into PLUM ground-motion-based earthquake early warning in the United States: the APPLES configuration. Earth’s Future, 12(2), e2023EF004126. https://doi.org/10.1029/2023EF004126
• Saunders, J.K., S.E. Minson, A.S. Baltay, J.J. Bunn, E.S. Cochran, D.L. Kilb, C.T. O’Rourke, M. Hoshiba, Y. Kodera (2022). Real‐time earthquake detection and alerting behavior of PLUM ground‐motion‐based early warning in the United States. Bulletin of the Seismological Society of America, 112(5), 2668-2688. https://doi.org/10.1785/0120220022
• Cochran, E.S., J.K. Saunders, S.E. Minson, J. Bunn, A. Baltay, D. Kilb, C. O’Rourke, M. Hoshiba, Y. Kodera (2022). Alert optimization of the PLUM earthquake early warning algorithm for the western United States. Bulletin of the Seismological Society of America, 112(2), 803-819. https://doi.org/10.1785/0120210259
• Kilb, D., J.J. Bunn, J.K. Saunders, E.S. Cochran, S.E. Minson, A. Baltay, C.T. O’Rourke, M. Hoshiba, Y. Kodera (2021). The PLUM earthquake early warning algorithm: A retrospective case study of West Coast, USA, data. Journal of Geophysical Research: Solid Earth, 126(7), e2020JB021053. https://doi.org/10.1029/2020JB021053
• Minson, S.E., J.K. Saunders, J.J. Bunn, E.S. Cochran, A.S. Baltay, D.L. Kilb, M. Hoshiba, Y. Kodera (2020). Real‐time performance of the PLUM earthquake early warning method during the 2019 M 6.4 and 7.1 Ridgecrest, California, earthquakes. Bulletin of the Seismological Society of America, 110(4), 1887-1903. https://doi.org/10.1785/0120200021
• Cochran, E.S., J. Bunn, S.E. Minson, A.S. Baltay, D.L. Kilb, Y. Kodera, M. Hoshiba (2019). Event detection performance of the PLUM earthquake early warning algorithm in southern California. Bulletin of the Seismological Society of America, 109(4), 1524-1541. https://doi.org/10.1785/0120180326
• Kodera, Y., Y. Yamada, K. Hirano, K. Tamaribuchi, S. Adachi, N. Hayashimoto, M. Morimoto, M. Nakamura, M. Hoshiba (2018). The Propagation of Local Undamped Motion (PLUM) Method: A Simple and Robust Seismic Wavefield Estimation Approach for Earthquake Early Warning. Bulletin of the Seismological Society of America, 108(2), 983-1003. https://doi.org/10.1785/0120170085

Ground Motion Aggregator (GMA)

Researchers within the ShakeAlert Project (including those at Caltech) are prototyping a Ground Motion Aggregator (GMA) algorithm which will combine ground-motion distributions estimated by the individual EEW algorithms into a unified ground-motion distribution to use for the alert regions. The ShakeAlert system currently combines the outputs of the individual EEW algorithms in terms of the algorithms’ source parameter estimates (magnitude, location, and origin time) using the Solution Aggregator algorithm, and the combined earthquake magnitude and location estimates are then passed to the eqInfo2GM algorithm to compute a single ground-motion distribution model used to create the ShakeAlert Message alert products. In a GMA procedure, the earthquake magnitude and location estimates from the individual EEW algorithms (EPIC, FinDer, and G-FAST) could be used to create separate ground-motion distributions, where the GMA will combine these three distributions with those from PLUM to create a unified ground-motion distribution for the alert products.

Combining the EEW algorithm estimates using the GMA has the potential to produce more accurate alert regions compared to the current procedure in some situations. These situations are rare and tend to happen in locations with sparse seismic station coverage as well as when the earthquake occurs outside of the normal seismic network boundaries. The GMA algorithm is also a key component for the future integration of the PLUM algorithm into the ShakeAlert system, as PLUM does not produce earthquake source estimates like the other EEW algorithms in ShakeAlert.

References

• Minson, S.E., S. Wu, J.L. Beck, T.H. Heaton (2017). Combining multiple earthquake models in real time for earthquake early warning. Bulletin of the Seismological Society of America, 107(4), 1868-1882. https://doi.org/10.1785/0120160331