Marco Olivieri 1 & Richard M. Allen 2 1) Istituto Naz. Di Geofisica e Vulcanologia, Roma, Italy

A Simple and Rapid earthquake Detection and Discrimination

System for ELARMSor

A new conceptual schema for earthquake detection

Marco Olivieri1 & Richard M. Allen2

1) Istituto Naz. Di Geofisica e Vulcanologia, Roma, Italy2) SeismoLab, UCBerkeley, Berkeley, CA

Aim of the project

To develop a new earthquake detection algorithm

Objective:• To speed-up the detection process• To reduce false detections

• Peculiar attention to EEW systems

Traditional approach

“Absence of evidence is not evidence of absence”Why?

Traditional approach (2)

Because traditional phase associators do not know: • If the missing station is broken. • If not, when the pick for that station will be made

available.

i.e. Detection bases ONLY on the presence of picks in

a certain time-window.

Modern seismic networks • Status monitor:

– active-down station (real-time flow monitors)– good ground motion signal (real-time PSD)

• Latency monitor: – delay in data trasmission

These can be updated in real-time

A new concept• Assumption: If the network is reliable, all the operating

stations in the surrounding of the epicenter will detect ground motion change and the picker will produce a P-wave detections with a a priori known delay

If the network is reliable, we can look only at close stations

SubnetsWe define the concept of subnet:A subnet is small network made by a master

station + the five closest surrounding stations.

(If the azimuthal coverage is not appropriate (> 180o) we replace some of the stations to enhance it)

An earthquake is detected at a subnet when more than 3 stations produce a pick consistent with an hypocenter beneath the master station

Or: for each earthquake we will have only one subnet centered above its epicenter

• MGR subnet (left) • AOI subnet (right) this is the case of a

“border” subnet (azimuthal gap > 200o)• Each station belong to “its” subnet plus some of the

surrounding ones

Subnet examples

Subnet update schema

MTTG subnetStation lat lon distance azimuth latencyMTTG 38.00 15.70 0. 14 MSCL 38.23 15.79 27.06 21.48 13 MPAZ 37.95 16.01 34.19 99.26 19 CEL 38.26 15.89 35.44 37.03 15 MSRU 38.26 15.50 35.60 323.7 13 SOI 38.07 16.05 39.73 78.80 11

Expected relative arrival time for substation:MSCL 3.7 3MPAZ 5.1 10CEL 5.2 6MSRU 5.2 4SOI 5.8 3

Side effect:The fourth pick is expected 6 +14 seconds= 20. This means that for a EEW we know that an alert can not be earlier than 20 seconds after the eq occurrence, i.e. the area that can not be protected has a radius of about 70km

MTTG subnetTIME 20080606213905 -1TIME 20080606213906 -1TIME 20080606213907 -1TIME 20080606213908 -1MTTG 20080606213909 20080606213855.730TIME 20080606213909 -1MSRU 20080606213910 20080606213900.840TIME 20080606213910 -1CEL 20080606213911 20080606213900.820MSCL 20080606213911 20080606213859.430SOI 20080606213911 20080606213901.300TIME 20080606213911 -1TIME 20080606213912 -1TIME 20080606213913 -1TIME 20080606213914 -1TIME 20080606213915 -1TIME 20080606213916 -1MPAZ 20080606213917 20080606213857.020

Master matching pick

MP

AZ

subn

etD

ata

Flow

Substation matching pick Step 1


2 picks

MP

AZ

subn

etD

ata

Flow

1 pick Step 2, 2 seconds later

TIME 20080606213905 -1TIME 20080606213906 -1TIME 20080606213907 -1TIME 20080606213908 -1MTTG 20080606213909 20080606213855.730TIME 20080606213909 -1MSRU 20080606213910 20080606213900.840TIME 20080606213910 -1CEL 20080606213911 20080606213900.820MSCL 20080606213911 20080606213859.430SOI 20080606213911 20080606213901.300TIME 20080606213911 -1TIME 20080606213912 -1TIME 20080606213913 -1TIME 20080606213914 -1TIME 20080606213915 -1TIME 20080606213916 -1MPAZ 20080606213917 20080606213857.020



5 picks: event detected

MP

AZ

subn

etD

ata

Flow

3 picks not matching:Event killed

Step 3, 4 seconds later


Conclusions

• enhance the rapid detection capability for a dense network of seismic stations in terms of:– Delay for the event declaration– False alert generated by sparse picks

• give a real-time and punctual information for the detection delay of the forthcoming earthquake (a crucial input for EEW systems)

the work is absolutely in progress

Association Diagram

example• TIME 20080606213901 -1• TIME 20080606213902 -1• TIME 20080606213903 -1• TIME 20080606213904 -1• TIME 20080606213905 -1• TIME 20080606213906 -1• TIME 20080606213907 -1• TIME 20080606213908 -1• MTTG 20080606213909 20080606213855.730• TIME 20080606213909 -1• MSRU 20080606213910 20080606213857.840• TIME 20080606213910 -1• CEL 20080606213911 20080606213855.820• MSCL 20080606213911 20080606213855.030• SOI 20080606213911 20080606213856.300• TIME 20080606213911 -1• TIME 20080606213912 -1• TIME 20080606213913 -1• TIME 20080606213914 -1• TIME 20080606213915 -1• TIME 20080606213916 -1• MPAZ 20080606213917 20080606213857.020

Latency issue:If latency is not constant over the different stations (and usually it is not!) the first incoming trigger can belong to a station that is not closest one to the epicenter

Marco Olivieri 1 & Richard M. Allen 2 1) Istituto Naz. Di Geofisica e Vulcanologia, Roma, Italy

Documents

certain timewindow

concept of subnet

expected relative arrival

rapid earthquake detection

aoi subnet right

detection processto

detection bases

border subnet azimuthal