Optimizing Plum Alerting Regions for Shakealert: A Test Case Tuning the Plum Prediction Radius
Session: Earthquake Early Warning Live in California! Current Status and Challenges [Poster]
Type: Poster
Date: 4/23/2021
Presentation Time: 11:30 AM Pacific
Description:
We optimize performance of the Propagation of Locally Undamped Motion (PLUM) Earthquake Early Warning (EEW) algorithm in the ShakeAlert system by testing a range of distances used to forecast ground motions. ShakeAlert aims to provide advance warning of potentially damaging shaking to the residents of California, Oregon, and Washington. EEW systems must balance the primary goal of providing timely alerts for moderate to strong ground shaking while limiting unnecessary alerts (alerts for ground motions too low to be of concern). PLUM forecasts ground motion directly from real-time observations of ground motion without estimating the source parameters of the earthquake. PLUM is already used by the Japan Meteorological Agency’s EEW system, and currently is under consideration for inclusion in ShakeAlert. PLUM forward-predicts observed ground motions to nearby prediction sites within a defined radius without any attenuation. PLUM was originally developed in Japan where alerts are issued to districts with sizes ranging from approximately 500 to 5,000 km2, and uses a prediction radius of 30 km. In contrast, ShakeAlert uses a generally smaller alerting grid of 0.2o by 0.2o cells (~500 km2). Given this contrast in alert region size, we test a range of prediction radii (20-100 km) to examine whether PLUM’s alert performance for ShakeAlert can be optimized. We calculate PLUM’s optimum prediction radius by applying ground motion-based performance assessments to the ShakeAlert historical event test suite. We compare predicted ground motions from PLUM alerts to ShakeMap-reported ground motions. We find the prediction radius that maximizes the correct (algorithm correctly predicts that ground motion will be above a given threshold at a site) and timely (alert is issued prior to ground motions exceeding a given threshold at a site) alerts while minimizing unnecessary alerts for a range of different ground motion levels of interest.
Presenting Author: Elizabeth S. Cochran
Student Presenter: No
Authors
Elizabeth Cochran Presenting Author Corresponding Author ecochran@usgs.gov U.S. Geological Survey |
Jessie Saunders jksaunders@usgs.gov U.S. Geological Survey |
Sarah Minson sminson@usgs.gov U.S. Geological Survey |
Julian Bunn julian.bunn@caltech.edu Caltech |
Deborah Kilb dkilb@ucsd.edu University of California, San Diego |
Annemarie Baltay abaltay@usgs.gov U.S. Geological Survey |
Colin O'Rourke corourke@usgs.gov U.S. Geological Survey |
Mitsuyuki Hoshiba mhoshiba@mri-jma.go.jp JMA |
Yuki Kodera y_kodera@mri-jma.go.jp JMA |
Optimizing Plum Alerting Regions for Shakealert: A Test Case Tuning the Plum Prediction Radius
Category
Earthquake Early Warning Live in California! Current Status and Challenges