Introduction

3 5

1 Introduction

Since 2004, the EU funded Marie Curie Training Network "Seismic wave propagation and imaging in complex media: a European network" joins 14 institutions and several associated partners in a project that aims at carrying out research in the field of computational seismology (Igel et al., 2004). One of its key deliverables is a www-based digital library with wave propagation codes, training material in numerical methods applied to the wave propagation problem and eventually simulation data. In 2005 the code library was initiated and several algorithms are now available to the scientific community. The goal is to provide any codes, tools, etc. that may be useful for researchers getting started in the field or observational seismologists interested in using the simulation techniques. In addition to sophisticated, parallelized 3D wave propagation algorithms based on finite differences, finite (spectral) elements or the pseudospectral methods for local, regional, and global models there are also simple training codes that help getting started with a particular method or can be used in tutorials. The library also contains "classical" approaches like ray-theoretical approaches, the reflectivity and the normal mode methods. One of the most important features are three benchmarks related to global tomography, wave-propagation code validation, and source imaging.

The SPICE Library: Codes, Training Material and Benchmarking in Computational Seismology

F. Gallovic1, R. Barsch1, H. Igel1, P. Moczo2, P. Pazak2, M. Mai3, Y. Qin4, and the SPICE team1 Department of Earth and Environmental Sciences, Section Geophysics, Ludwig-Maximilians-Universitaet, Muenchen, Germany2 Faculty of Mathematics, Physics and Informatics, Comenius University, Bratislava, Slovak Republic3 Institute of Geophysics, ETH Hoenggerberg, Zurich, Switzerland4 Departement de Sismologie, Institut de Physique du Globe de Paris, Paris, France

4

2 Software Library - Contents

In addition to numerical methods a strong focus is the provision of analytical solutions (e.g., Lamb’s problem, source at bimaterial interface) that can be used to test numerical solutions and are often hard to get hold of. Anybody interested can provide code through the SPICE interface (contact administrator).

Analytical solution type Numerical solution type

www.spice-rtn.orgwww.spice-rtn.org

Digital Library - ScopeTraining Material: From the SPICE workshop numerous presentations on the fundamentals of numerical methods applied to wave propagation problems are available (-> events ->workshops). Topics include: finite differences, spectral elements, pseudospectral methods, ray methods, normal modes, boundary conditions, dynamic rupture, etc.

Books: Among the training materials, two books are available: (1) Moczo, P., Kristek, J., Halada, L. (2004). The Finite-Difference Method for Seismologists. An Introduction. (2) Brokesova, J (2006), Asymptotic ray method in seismology: A tutorial.

Quasi-analytical solution type

Software: The software library is an open platform, in which anybody can participate by offering his/her code. Each code is classified in categories, which can be used to filter the entries: For example, the category “Code Level” offers: Production Code, Research and Training. The category “Solution Type” can be found below.

References:References:Brokesova, J (2006). Asymptotic ray method in seismology: A tutorial. Matfyzpress, Brokesova, J (2006). Asymptotic ray method in seismology: A tutorial. Matfyzpress, Praha, ISBN 80-86732-74-6 155 pp.Praha, ISBN 80-86732-74-6 155 pp.Igel, H., and the SPICE Team (2004). New European Training Network to Improve Young Igel, H., and the SPICE Team (2004). New European Training Network to Improve Young Scientists' Capabilities in Computational Wave Propagation. EOS, Transactions of the Scientists' Capabilities in Computational Wave Propagation. EOS, Transactions of the American Geophysical Union, Volume 85, Number 28.American Geophysical Union, Volume 85, Number 28.

Kristekova, M., Kristek, J., Moczo, P., Day, S. M. (2006). Misfit Criteria for Quantitative Kristekova, M., Kristek, J., Moczo, P., Day, S. M. (2006). Misfit Criteria for Quantitative Comparison of Seismograms, Bull. Seism. Soc. Am. 96 (5), 1836-1850, doi:10.1785/0120060012.Comparison of Seismograms, Bull. Seism. Soc. Am. 96 (5), 1836-1850, doi:10.1785/0120060012.Moczo, P., Kristek, J., Halada, L. (2004). The Finite-Difference Method for Seismologists. An Moczo, P., Kristek, J., Halada, L. (2004). The Finite-Difference Method for Seismologists. An Introduction. Comenius University, Bratislava, ISBN: 80-223-2000-5, 158 pp.Introduction. Comenius University, Bratislava, ISBN: 80-223-2000-5, 158 pp.Qin, Y., Capdeville, Y. (2006). Synthetic Dataset To Benchmark Global Tomographic Methods. Qin, Y., Capdeville, Y. (2006). Synthetic Dataset To Benchmark Global Tomographic Methods. EOS, Transactions of the American Geophysical Union, Volume 87, November 14.EOS, Transactions of the American Geophysical Union, Volume 87, November 14.

Global tomography benchmark

A synthetic dataset for testing global tomographic methods is provided (Qin et al., 2006). This global-scale benchmark dataset comprises seismograms synthesized for a realistic three-dimensional mantle model, containing complexities on various spatial scales and different types of heterogeneities in velocity, anisotropy, attenuation, and density. In addition, topography, ellipticity, Earth’s rotation, self-gravity and ocean thickness are taken into account. Each participant can download the benchmarking synthetic seismograms and test performance of his/her tomography code.

Wave-propagation validationWave-propagation code validation: an interactive web interface has been developed, offering a simple way to compare numerical and analytical (semi analytical) solutions. A set of testing examples are included, together with their thorough description and reference solutions, allowing the participant to fully replicate the model with his/her own code. Reference solutions can be downloaded for the comparison purposes. Alternatively, the simulation result can be then uploaded to the server, which will calculate and display the time-frequency misfits between selected solutions (Kristekova et al., 2006).

Source-imaging benchmark(a blind-test on kinematic source inversion)

A synthetic data set for a hypothetical source model (based on the 2000 Tottori earthquake) with heterogeneous slip distribution is provided for a set of receivers. Participants can download this data set together with full description of the problem, try to invert for the slip distribution and compare the result with the correct one that was used to obtain the benchmarking database.

Models available so far

You can also only download the model description and the solution and compare it yourself

0 9 0 1 8 0 2 7 0 3 6 0- 9 0

0

9 0

0 9 0 1 8 0 2 7 0 3 6 0- 9 0

0

9 0

P A Y GO T A VT A R AB T D F

_ H D C_ K O D_ A G D

G U M O_ T G Y _ M B O K O W A_ F D F_ T P X S A C VP N IGJ O H N_ H Y B _ S J G_ S R RC H T O _ Q I Z C O L MW A K E P O H A

X M A S_ P S I

T E IGL Y U BA B U S_ T A M W S A RR AY N M C S J_ K M I N A G O _ O G S _ H 2 O_ E VN D W P F_ M D Y_ N D I _ T B TS IW A _ K B D _ L T X_ E IL _ H K T_ Q U E _ E N H _ S S E _ TK A

G H A R _ TU C _ B E CM A R J _ B A R _ M D TN H S C_ H C H_ N I L _ O X FW M O K_ A L Q_ C S SK R IS _ K S H _ T T T_ W D D_ T H R _ P K D_ L Z H_ M E B _ TJ N _ K N B_ S S Y E R I PC M L A_ TA B _ A O BM A L T G W D E_ G A R_ V S L

_ B N G B O C O _ C AY

_ B J T A C S O_ B M N_ A R C_ W U S_ T I R_ C A D G I G S _ H R VA H ID J F W S_ N M R_ W M Q_ K IV R S S D_ M D J _ H A LS A D O_ B N I _ R B N_ M L R R A ION E 5 6 M A K Z _ C B M_ U L N_ P S Z E Y M N_ F U R _ N E W D R L N_ H I A K A P OA K T K _ S B D_ P H CK U R K_ G S H_ B H M _ T L Y _ A D KN E 5 4 _ S M Y_ V O S _ P ET M O BC _ E D M_ U V AB S E G_ F F C S C H Q_ N V S_ M H V _ E K BP R YA B M B C_ A R U _ S IT_ K D CN E 5 2 D L B C _ F C C

_ M A 2K O N O _ U P P

D B IC

_ N R S_ P M R_ Y A K Y K W 2_ K M S _ FR BD A W Y B O R G_ S F J_ L V Z B I LL _ I N K

N R ILA R A 0

PA T S_ D A V P A LU

_ S C O _ J M IT I X I

_ R E S_ M B C _ D A G

_ K B S

F U N AK W A J _ S D V

_ A L E

M B A R P T G AK M B OM S K U

R I O BM N H K_ N N AC O C O

_ A FI_ L S Z B D F B S H ELZ O B OL B S A

_ P P T_ P V CF I T Z VA V AA B P OT S U M_ W R A

K N T N_ J A Y

C T A OM B W A_ R E R _ R A R

B R A S_ L V CLB T B P T C N

W A R B C P U P_ R P N_ L C ON R F K _ R A O T R I NA R M AW O O L S T K A_ S U R B B O O

_ C A NT R I S_ A I S T R Q A

S E Y 1

P L C AS N Z O_ TA U

C O Y CC R Z F

_ PA F

K A PI R C B R

_ E F IH O P E_ M C Q

O R C D

P M S AC A S Y _ D R V_ M AW

_ S Y O

D G A R _ U G M A S C N

S N A A_ T N V_ D C C

Q S P A

M A R Q_ P M G _ H N R

_ S P A

0 9 0 1 8 0 2 7 0 3 6 0-9 0

0

9 0

Distribution of realistic acquisition geometry. (Top) Distribution of 29 events located along plate boundaries, representing the real seismicity of world, with divergent, convergent and strike-slip plate margins represented in ‘beach ball’ diagrams. The size of beach balls is proportional to the magnitude of the events. (Bottom) Distribution of 256 stations extracted from FDSN (Federation of Digital broad-band Seismograph Network) station book. Stations were selected for good global spatial coverage.

The waveforms has been computed for 1D crustal model using COMPSYN code, i.e., discrete frequency-wavenumber integration without attenuation. The frequency range of the calculations spans 0.01 - 3.0 Hz.

And others…