Earthquake Early Warning Research and Development in California, USA

Earthquake Early Warning Research and Development in California, USA

Hauksson E., Boese M., Heaton T.,Seismological Laboratory, California Institute of Technology,

Pasadena, CA, Given D., USGS, Pasadena, CA,

Oppenheimer D., USGS, Menlo Park, CA, Allen R. , Hellweg P. ,

Seismological Laboratory, UC Berkeley, Berkeley, CA, Cua G. , Fischer M. , Caprio M.

Swiss Seismological Service, ETH Zurich

California

ANSS/CISN Early Warning R&D Project• Collaboration:

• USGS • Caltech • UC-Berkeley• ETH, Zurich• USC/SCEC

• Develop EEW algorithms to detect and analyze earthquakes within seconds

• Identify needed improvements to the existing monitoring networks

• Implement an end-to-end prototype test system

trigger time

+ 1 sec

+ 3 sec

• EEW requirements:-- Rapid earthquake detection-- Early Mag. estimation-- Ground shaking prediction-- Robust seismic networks-- Well trained uses

3

• CISN real-time testing of 3 algorithmsτc-Pd On-site algorithm, VS, & ElarmS

• State-wide implementation382 stations with 585 broadband & strong motion instruments

• Many small to moderate earthquakes 2007 Mw5.4 Alum Rock & 2008 Mw5.4 Chino Hills2010 Mw7.2 Baja California

• CISN EEW Testing Center established at University of Southern California (USC)/SCEC

τc-Pd

On-site Algorithm

Single sensor

Virtual Seismologist

(VS)ElarmS

Sensor network Sensor network

Caltech ETH Zurich/Caltech UC Berkeley

CISN EEW Algorithm Testing (2007-2009)

Progress:

4

CISN ShakeAlert (2009-2012)Project Goals

Year 1 (2009/10):Implementation

Year 2 (2010/11):Testing/Optimization

Year 3 (2011/12):Evaluation

System specificationsCode design specificationsCode developmentDefine formats and protocols

Implement end-to-end processingTesting with archived dataTesting with real-time dataImprove performanceTesting at the SCEC Testing CenterTesting with selected users

Prototype system in operationAdd features to Decision ModuleResearch adding GPS RT positionsResearch on finite sourcesPlans for future systems

5

Results

τc-Pd

On-site Algorithm

Virtual Seismologist

(VS)ElarmS

Speed:What causes delays?

CISN EEW Algorithm Testing (2007-2009)

R. Allen

Median: ~ 5.2 sec California

Data latency (datalogger/telemetry

delays)

Station density

0 10 20 sec

Single sensor Sensor network Sensor network

6

Results

τc-Pd

On-site Algorithm

Virtual Seismologist

(VS)ElarmS

Speed:

CISN EEW Algorithm Testing (2007-2009)

R. Allen

California

Data latency (datalogger/telemetry

delays)

Station density

0 10 20 sec

Single sensor Sensor network Sensor network

How can these delays be reduced in the future ?1. reduce data latency up-grade of ~220 CISN stations with new Q330s dataloggers (~1-2 sec delay) before Sept-2011 (ARRA stimulus funding)

2. increase processing speed current delays: ~5 sec

3. Increase station density

4. Decreas number of stations required for trigger

7

Results

τc-Pd

On-site Algorithm

Virtual Seismologist

(VS)ElarmS

Examples:Mw5.4 Alum Rock: 5 sec before peak shaking in San Francisco.Mw5.4 Chino Hills: 6 sec warning at Los Angeles City Hall. Mw7.2 Baja Calif. 70?? sec warning at Los Angeles City Hall.

Speed:after O.T. > 5 sec ~20 sec

~30 sec

CISN EEW Algorithm Testing (2007-2009)

Single sensor Sensor network Sensor network

± 0.5 ± 0.2 ± 0.4* *includes M>7 data from JapanMMI: ±0.7

false alerts: (M>6.5) 1* 0

0* three month period

Mag.: Mw:Reliability:

8

(2009-2012)

- most probable… Mw

… location… origin time… ground motion

and uncertainties

- probability of false trigger, i.e. no earthquake

- CANCEL message if needed

Bayesian approachup-dated with time

Task 1: increase reliability

Decision Module(Bayesian)

CISN ShakeAlertτc-Pd

On-site Algorithm

Virtual Seismologist

(VS)ElarmS

Single sensor Sensor network Sensor network

9

USER Module- Single site warning- Map view

SCEC/ EEW Testing Center

Decision Module(Bayesian)

Test users

CISN ShakeAlertτc-Pd

On-site Algorithm

Virtual Seismologist

(VS)ElarmS

(2009-2012)

Task 1: • increase reliability

Task 2: demonstrate

• predicted and observed ground motions• available warning time• probability of false alarm•…

feed

-bac

k

Single sensor Sensor network Sensor network

10

CISN EEW Testing Center

CISN ShakeAlertτc-Pd

On-site Algorithm

Virtual Seismologist

(VS)ElarmS

Decision Module(Integration Module) feed-back

by test users

User Display

M. Boese

CISN ShakeAlert

• platform independent (Java)

• ability to add multiple map layers & navigational features (OpenMap application programming interface)

11

User Display

M. Boese

CISN ShakeAlert

12

• remaining time until S-wave arrival

User Display

M. Boese

CISN ShakeAlert

13

• remaining time until S-wave arrival• expected intensity at user site

User Display

M. Boese

CISN ShakeAlert

14

• remaining time until S-wave arrival• expected intensity at user site• (moment) magnitude

User Display

M. Boese

CISN ShakeAlert

15

• locations of epicenter & user

User Display

user

epicenter

M. Boese

CISN ShakeAlert

16

• locations of epicenter & user• locations of P- /S-wavefronts

User Display

P-waveS-wave

M. Boese

CISN ShakeAlert

17

• locations of epicenter & user• locations of P- /S-wavefronts• intensity map (ShakeMaps color-code)

User Display

M. Boese

CISN ShakeAlert

18

• siren• voice announcement:

• count-down• “weak shaking”, “strong shaking”…

User Display

future: different announcements depending on distance

M. Boese

19

CISN ShakeAlert

See also Doug Given’s Webpage: http://pasadena.wr.usgs.gov/office/given/eew/

2008 M5.4 Chino Hills

1994 M6.7 Northridge

1989 M6.9 Loma Prieta (UCB)

1989 M6.9 Loma Prieta (San Jose)

M7.8 ShakeOut Scenario

User Display - Demos

M. Boese

20

CISN ShakeAlertProblem:

Point source approximationExpected intensity in LA:

point source: IV light shaking

M. Boese

21

CISN ShakeAlertProblem:

Point source approximationExpected intensity in LA:

point source: IV light shakingfinite fault: VIII severe shaking

M. Boese

Finite Fault Detector

22

Near/far-source Classification

e.g, 7.233*log10(Za) + 6.813*log10(Hv)-15.903 0 . (Yamada et al., 2007)

Za: vertical acceleration cm/s2

Hv: horzontal velocity cm/s

near-source

M. Boese

Finite Fault Detector

23

1. Estimated Magnitude: 6.6 2. Estimated Magnitude: 6.9

3. Estimated Magnitude: 7.1 4. Estimated Magnitude: 7.5

Real-time near/far-source classification

M. Boese

Basic Research Projects

• Development of algorithms to analyze long ruptures (Heaton, Böse, and Karakus; Allen and Brown)

• Development of User Decision module based on cost/benefit (Beck and Wu)

• Development of slip detectors based on real-time GPS (Hudnut and Herring)

Conclusions

• Finally have put the elements together to produce real-time alerts

• much work remains to produce a reliable system for general use