Station Quality Monitoring for the Pacific Northwest Seismic Network and Shakealert

Date: 4/24/2019

Time: 06:00 PM

Room: Fifth Avenue

The Pacific Northwest Seismic Network is responsible for monitoring seismic activity in Cascadia and is also part of the USGS ShakeAlert earthquake early warning system. Fulfilling our role effectively requires monitoring station state of health metrics, those related to waveform quality such as power spectral density estimates, and those specific to ShakeAlert. We currently monitor station quality in near-real time at all PNSN stations as well as all stations contributing to ShakeAlert. Using our large database of metrics spanning many months and in collaboration with the ShakeAlert regional coordinators, we have established acceptance criteria for adding new stations to the ShakeAlert system. These include measures of latency, data completeness and number of potential earthquake triggers among others. We are also mining our database as well as logfiles from Earthquake Point-source Integrated Code (EPIC), one of two algorithms used by ShakeAlert, to identify parameters that can be modified and additional waveform checks that can be added to reduce the number of false triggers while not negatively impacting EPIC's performance. To test out ideas, we run test instances of EPIC as well as replay past earthquakes to assess any suggested changes. Finally we are exploring ways to establish thresholds for various metrics to be used to grey/black-list problematic stations in ShakeAlert. These are assessed using a map-based tool of theoretical alert time. To better visualize and comprehend the mountain of information that we collect and generate, we have begun work on an API which will be able to digest any properly formatted JSON message sent to the API and thus can ingest metrics from databases, AQMS messages, other APIs, upload via wget, etc. The API will be web-serviceable that utilizes Tableau visualization tools.

Presenting Author: Alex Hutko

Authors