Real-time Performance of the PLUM Earthquake Early Warning Algorithm for the West Coast of the US
We summarize the real-time performance (July 2019-Present) of the Propagation of Local Undamped Motion (PLUM) earthquake early warning algorithm in California, Oregon, and Washington. PLUM is being tested for potential inclusion in the ShakeAlert system. The ShakeAlert system is currently based on two methods that estimate earthquake source parameters. PLUM is different, predicting expected ground motions from observed ground motions directly. The original algorithm was developed in Japan to complement source-based algorithms to address their deficiencies, such as missed or false alerts during complex sequences and limited to no warning times in the near source region. PLUM detects an event when two neighboring stations observe ground motions above defined trigger thresholds, then forward predicts those and subsequent ground motions to nearby regions. Retrospective testing of PLUM was used to optimize the configuration for the station distribution and levels of shaking targeted for alerts in the ShakeAlert system. Real-time testing of PLUM since July 2019 has allowed for further refinement of the algorithm. We used a primary station trigger threshold of Modified Mercalli Intensity (MMI) 4.0; the secondary station trigger threshold changed from MMI 2.5 to MMI 3.0 in Sept. 2020. We also increased the PLUM prediction radius from 30 km to 60 km to optimize performance when using the 0.2° by 0.2° ShakeAlert grid as alert regions. Soon after initializing, PLUM issued 11 false alerts (alerts not associated with an earthquake) caused by station noise and duplicate station data, but minor code changes and data quality checks eliminated all false alerts since July 2020. During the real-time tests PLUM alert grids accurately covered areas of shaking for all earthquakes of concern, including M6.4 & M7.1 Ridgecrest, M6.5 Monte Cristo, M5.8 Lone Pine, M5.1 & M6.2 Petrolia, M6.0 & M5.0 Antelope Valley. The median time from the origin to detection is around 6 seconds.
Session: Advances in Earthquake Early Warning: Research, Development, Current State of Practice and Social Science II
Type: Oral
Room: Cedar
Date: 4/21/2022
Presentation Time: 10:45 AM Pacific
Presenting Author: Elizabeth S. Cochran
Student Presenter: No
Additional Authors
Elizabeth Cochran Presenting Author Corresponding Author ecochran@usgs.gov U.S. Geological Survey |
Debi Kilb dkilb@ucsd.edu University of California, San Diego |
Jessie Saunders jksaunders@usgs.gov U.S. Geological Survey |
Julian Bunn julian.bunn@caltech.edu California Institute of Technology |
Colin O'Rourke corourke@usgs.gov U.S. Geological Survey |
Sarah Minson sminson@usgs.gov U.S. Geological Survey |
Annemarie Baltay abaltay@usgs.gov U.S. Geological Survey |
|
|
Real-time Performance of the PLUM Earthquake Early Warning Algorithm for the West Coast of the US
Category
Drop Cover and Hold On! ShakeAlert: Past, Present and Future
Description