automatic moment tensor solution for seiscomp3

Automatic Moment Tensor Automatic Moment Tensor solution for SeisComP3solution for SeisComP3

Second European Conference on Earthquake Engineering and Second European Conference on Earthquake Engineering and Seismology, Istanbul Aug. 25-29, 2014Seismology, Istanbul Aug. 25-29, 2014

Nikolaos TriantafyllisNikolaos Triantafyllis11, Efthimios Sokos, Efthimios Sokos22 and Aristidis Ilias and Aristidis Ilias11

1 University of Patras, Computer Engineering and Informatics Department, Rio 26504, Patras, Greece, 1 University of Patras, Computer Engineering and Informatics Department, Rio 26504, Patras, Greece,

[email protected]@ceid.upatras.gr, , [email protected]@ceid.upatras.gr

2 University of Patras, Geology Department, Rio 26504, Patras, Greece, 2 University of Patras, Geology Department, Rio 26504, Patras, Greece, [email protected]@upatras.gr

mailto:[email protected]



IntroductionIntroduction

Moment Tensors are important for studies like Moment Tensors are important for studies like shakemap generation, tsunami warnings, shakemap generation, tsunami warnings, ground motion evaluation and moreground motion evaluation and more

Importance of automatic, quick and reliable Importance of automatic, quick and reliable Moment Tensor solutionMoment Tensor solution

ScisolaScisola

Open-source python based softwareOpen-source python based software User friendly GUIUser friendly GUISupports:Supports: Automatic Moment Tensor calculation of events provided by Automatic Moment Tensor calculation of events provided by SeisComP3 in real-time SeisComP3 in real-time

Easy solution overviewEasy solution overview Quick solution revisionQuick solution revision Extensive configurationExtensive configuration

SeisComP3SeisComP3

Likely the most widely distributed software package Likely the most widely distributed software package for seismological purposesfor seismological purposes

Evolved within approximately the last 10 yearsEvolved within approximately the last 10 years Its use ranges from pure acquisition or real-time data Its use ranges from pure acquisition or real-time data exchange over Internet to a fully featured real-time exchange over Internet to a fully featured real-time earthquake monitoringearthquake monitoring

Provides real-time waveform data through Seedlink Provides real-time waveform data through Seedlink protocolprotocol

ISOLAISOLAISOLA moment tensor algorithm:ISOLA moment tensor algorithm: Point source iterative deconvolution methodPoint source iterative deconvolution method Full wavefield is taken into accountFull wavefield is taken into account Discrete wavenumber method for local or regional distancesDiscrete wavenumber method for local or regional distances Moment tensor is found by least-square minimization of misfit between Moment tensor is found by least-square minimization of misfit between

observed and synthetic waveformsobserved and synthetic waveforms Position and time of centroid is optimized through grid searchPosition and time of centroid is optimized through grid search

• Used in a manual way in GI-NOA, UPSL and other institutes worldwideUsed in a manual way in GI-NOA, UPSL and other institutes worldwide• More info in Sokos & Zahradnik, 2008 and Sokos & Zahradnik, 2013More info in Sokos & Zahradnik, 2008 and Sokos & Zahradnik, 2013

ArchitectureArchitecture

SeisComP3 InterconnectionSeisComP3 Interconnection

Automatic ProcedureAutomatic Procedure

Origin TriggeringOrigin Triggering Automatic ModeAutomatic Mode

Watch SeisComP3 through scevtlsWatch SeisComP3 through scevtls Retrieve origin's info through scxmldumpRetrieve origin's info through scxmldump

Manual ModeManual Mode Execution through python script (mostly for Execution through python script (mostly for

testing purposes)testing purposes)

Station Selection based on DistanceStation Selection based on Distance Retrieves Stations/Streams Info from scisola databaseRetrieves Stations/Streams Info from scisola database Filters Stations/Streams by certain type Filters Stations/Streams by certain type

(e.g. HHN, HNE)(e.g. HHN, HNE) Removes Blacklisted Stations/Streams defined by the userRemoves Blacklisted Stations/Streams defined by the user Calculates distance and azimuth of stations according to Calculates distance and azimuth of stations according to

epicenterepicenter Selects Stations/Streams according to “distance rule”, Selects Stations/Streams according to “distance rule”,

defined by the user defined by the user (e.g. 3.5 (e.g. 3.5 ≤ mw ≤ 4.5 20 ≤ distance ≤ 100 km)→≤ mw ≤ 4.5 20 ≤ distance ≤ 100 km)→

Bad Station Data FilteringBad Station Data Filtering

Filters unavailable stations/streams according Filters unavailable stations/streams according to seedlink to seedlink

Retrieves records from seedlink in mseed Retrieves records from seedlink in mseed format format

Filters stations/streams with gapsFilters stations/streams with gaps Filters stations/streams with clipping Filters stations/streams with clipping

Station Data CorrectionStation Data Correction

Rotates stations/streams automaticallyRotates stations/streams automatically Corrects stations/streams dataCorrects stations/streams data

Removes instrumental effectRemoves instrumental effect Aligns according to origin timeAligns according to origin time Cuts to predefined durationCuts to predefined duration ResamplesResamples

Station Selection based on AzimuthStation Selection based on Azimuth

Station distribution Station distribution according to epicenter according to epicenter location location

8 sectors, each of 458 sectors, each of 45°° Minimum number of Minimum number of

sectors defined by the user sectors defined by the user Maximum stations per Maximum stations per

sector defined by the user sector defined by the user

Green's Functions ComputationGreen's Functions Computation

1D crustal model defined by the user1D crustal model defined by the user Computation of elementary seismograms (delta time Computation of elementary seismograms (delta time

function)function) Time window length of inversion procedure (tl) is defined by the Time window length of inversion procedure (tl) is defined by the

user according to “tl rule”user according to “tl rule”(e.g. 3.5 (e.g. 3.5 ≤ mw ≤ 4.5 tl = 327.68 sec)→≤ mw ≤ 4.5 tl = 327.68 sec)→

Centroid horizontal position fixed at epicenter location while Centroid horizontal position fixed at epicenter location while centroid depth is grid searchedcentroid depth is grid searched

The number of The number of trial sources above and below automatic trial sources above and below automatic depth estimation is defined by the userdepth estimation is defined by the user

Inversion ComputationInversion Computation

Source inversion predefined to deviatoric typeSource inversion predefined to deviatoric type Inversion frequency band is defined by the Inversion frequency band is defined by the

user according to “frequency rule” user according to “frequency rule” (e.g. 3.5 (e.g. 3.5 ≤ mw ≤ 4.5 frequencies = [0.04, →≤ mw ≤ 4.5 frequencies = [0.04, →0.05, 0.08, 0.09] Hz)0.05, 0.08, 0.09] Hz)

Time grid search for centroid time is defined Time grid search for centroid time is defined by the userby the user

Result PlottingResult Plotting

Generates text files with results and the final focal mechanismGenerates text files with results and the final focal mechanism Generates a map containing the epicenter location and Generates a map containing the epicenter location and

contributing stations' location contributing stations' location Generates a misfit of observed and synthetic waveforms plotGenerates a misfit of observed and synthetic waveforms plot Generates correlation plots containing focal mechanisms on Generates correlation plots containing focal mechanisms on

different depth and time and the best solutions for the different depth and time and the best solutions for the different depths different depths

Generates contributing streams plotGenerates contributing streams plot Saves results to scisola databaseSaves results to scisola database

Revise ProcedureRevise Procedure

User can manually remove streams that have User can manually remove streams that have been selected by the automatic procedure from been selected by the automatic procedure from the inversionthe inversion

User can change inversion frequenciesUser can change inversion frequencies

WatcherWatcher

Automatic Vs ManualAutomatic Vs Manual


Comparison of 46 manual MT Comparison of 46 manual MT solutions by GI-NOA, with 46 solutions by GI-NOA, with 46 automatic by scisolaautomatic by scisola

Comparison using the Kagan Comparison using the Kagan angle metricangle metric

Locations Locations red dots→ red dots→


34/46 = similar solutions≈ 74% success


39/46 = similar mw≈ 85% success

origins' difference not more than 0.3 of units


35/46 = similar c.depth≈ 76% success


Results: Results: Automatic Vs Manual 74% similar solutionsAutomatic Vs Manual 74% similar solutions Automatic MTs recognize the size and depth of the seismic Automatic MTs recognize the size and depth of the seismic

source with adequate accuracy a few minutes after the event. source with adequate accuracy a few minutes after the event. Important for quick estimation of ground motions or tsunami Important for quick estimation of ground motions or tsunami hazardhazard

Average time of the automatic procedure is Average time of the automatic procedure is ≈ 5 minutes≈ 5 minutes

Quick revision, in just a few minutes, provides highly accurate Quick revision, in just a few minutes, provides highly accurate solutionssolutions

Future ImprovementsFuture Improvements Advanced methods and artificial intelligence techniques for Advanced methods and artificial intelligence techniques for

stations/streams selectionstations/streams selection Implementation of multiple crustal models based on Implementation of multiple crustal models based on

earthquake’s locationearthquake’s location Advanced signal processing methods to be applied to seismic Advanced signal processing methods to be applied to seismic

waveforms (to avoid various problems e.g. disturbances, noise waveforms (to avoid various problems e.g. disturbances, noise and data transmission problems)and data transmission problems)

Search of centroid in 3D grid surrounding hypocenterSearch of centroid in 3D grid surrounding hypocenter Pre-calculated Green’s functions (for faster performance)Pre-calculated Green’s functions (for faster performance) Estimation of solution’s qualityEstimation of solution’s quality

GUIGUI

Log-in screenLog-in screen scisola DB (MySQL)scisola DB (MySQL)

SeisComP3 DB SeisComP3 DB (MySQL/PostgreSQL) (MySQL/PostgreSQL)

Log-in screenLog-in screen scisola DB (MySQL)scisola DB (MySQL)

SeisComP3 DB SeisComP3 DB (MySQL/PostgreSQL) (MySQL/PostgreSQL)

databases

Main ScreenMain Screen


main tabs


main buttons


latest 20 origins

Main ButtonsMain Buttons

start/stop watcherstart/stop watcher settings screensettings screen search screensearch screen update latest 20 update latest 20

originsorigins about screenabout screen

Settings ScreenSettings Screen


settings tabs


edit scisola stations

Settings ScreenSettings Screenclick to edit


double click to review

Review ScreenReview Screen


review tabs

Review ScreenReview Screenrevision

Review ScreenReview Screenrevision

double click to remove stream


delete origin


scroll to zoom in/out

Main ButtonsMain Buttons

start/stop watcherstart/stop watcher settings screensettings screen search screensearch screen update latest 20 update latest 20

originsorigins about screenabout screen

Search ScreenSearch Screen

shows origins in user shows origins in user defined datetime defined datetime (GMT) range(GMT) range


searched origins


clicking on minimizeclicking on minimize

For More...For More...

source code and manual can be found here:source code and manual can be found here:https://github.com/nikosT/scisolahttps://github.com/nikosT/scisola

https://github.com/nikosT/scisola

automatic moment tensor solution for seiscomp3

Documents