www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Seismic wave Propagation and Imaging in

Complex media: a European network

EBRU BOZDAĞEarly Stage Researcher

Host Institution: Utrecht UniversityPlace of Origin: Istanbul, TurkeyAppointment Time: February 2005

Project: Testing and Improving Tomographic Models Using Numerical 3D Wave Propagation

Task Groups: TG Planetary Scale

Cooperation: Oxford University

EBRU BOZDAĞEarly Stage Researcher

Host Institution: Utrecht UniversityPlace of Origin: Istanbul, TurkeyAppointment Time: February 2005

Project: Testing and Improving Tomographic Models Using Numerical 3D Wave Propagation

Task Groups: TG Planetary Scale

Cooperation: Oxford University

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Crustal corrections predicted by ray theory and finite frequency

theory compared to measured time shifts

from SEM seismograms using Crust2.0

Ebru Bozdağ

Jeannot Trampert

SPICE

Utrecht University, Utrecht, the Netherlands

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Crustal corrections are important in surface wave tomography

Crustal structures have a strong effect on surface waves.

Inverting for crust and mantle is difficult. Therefore crustal corrections are preferred.

Phase delays from the crust are removed in surface wave tomography to identify mantle structure.

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

The objective

Investigate how far

great circle approximation exact ray theory and finite frequency theory

predict crustal corrections using the SEM (Komatitsch and Tromp, 2002) seismograms computed in 1D (PREM) and 3D (PREM+Crust2.0) earth models.

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Data generationEarthquakes & Stations

VIC

CMA

SEP

CRT XZNG

WIASOA

NCN

IJR

SIRIJR

11 earthquakes253 stations

Synthetic seismograms using SEM Synthetic seismograms from 1D earth model PREM (Dziewonski and Anderson, 1981) 3D crustal model Crust2.0 (Bassin et al., 2000) is superimposed on top of PREM model and synthetic seismograms are computed for PREM+Crust2.0 model

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Measuring time shifts as a function of frequencyPREM

PREM+Crust2.0

Cross-correlation of the seismograms

Time-variable filter to extract the

fundamental mode Rayleigh wave

Phase correction to PREM seismograms

Unwrap the phase

ω

ddt

Measure the phase of the cross-correlation

d PREM+Crust2.0 - PREM PREM+Crust2.0 – PREM+Corr.

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Great Circle Approximation (GCA)

Exact Ray Theory (ERT)

Finite Frequency Theory (FFT)

Methods used for crustal corrections

dc

dc

cd

0 00

0 0 0

),(),(

ddc

dcKd

dsc

dc

cd

ray

00

1

2

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Calculation of local phase velocity perturbations

At each grid point of Crust2.0, we superimpose the crustal model (plus topography) onto PREM and solve for the exact eigenfrequencies in that 1D earth model.

We thus generate exact local phase velocities at each grid point which are used to calculate crustal corrections along rays.

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

A spherical harmonic expansion of the local phase velocities is used to simulate the

smoothing of Crust2.0 in SEM

Rayleigh, 40s

Without smoothingSmoothing with

spherical harmonic expansion

dc/c0

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Ray paths showing the time shifts computed for 150 s using GCA for one earthquake

Time shifts from PREM+Crust2.0 – PREM

seismograms

Time shiftsafter correction

(PREM+Crust2.0 - PREM+GCA)

time shift (s)

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Time shifts as a function of distance calculated for 150 s using GCA (l=40)

Black lines: average uncertainties of the measurements of Trampert & Woodhouse

Blue bars: before correction(PREM+Crust2.0 - PREM)

Red bars: after correction(PREM+Crust2.0 – PREM+GCA)

dt=±5.9 s

dt=±7.2 s

dt=±10.6 s

dt=±19.4 s

dt=±22.7 s

dt=±27.8 s

90%

76%

81%

89%

92%

96%

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Black lines: average uncertainties of the measurements of Trampert & Woodhouse

Blue bars: before correction(PREM+Crust2.0 - PREM)

Red bars: after correction(PREM+Crust2.0 – PREM+GCA)

Time shifts as a function of distance calculated for 40 s using GCA (l=40)

dt=±4.8 s

dt=±7.2 s

dt=±11.2 s

dt=±17.8 s

dt=±20.8 s

dt=±24.8 s

59%

55%

69%

50%

62%

78%

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Oceans-continents 150 s, GCA, l=40

Blue bars: before correction(PREM+Crust2.0 - PREM)

Red bars: after correction(PREM+Crust2.0 - PREM+GCA)

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Oceans-continents 40 s, GCA, l=40

Blue bars: before correction(PREM+Crust2.0 - PREM)

Red bars: after correction(PREM+Crust2.0 - PREM+GCA)

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Comparison of methods (150 s, l=40)

GCA vs. FFT

GCA vs. ERT FFT vs. ERT

GCA vs. FFT (major arc)

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Comparison of methods (40 s, l=40)

GCA vs. FFT (major arc)

GCA vs. ERT FFT vs. ERT

GCA vs. FFT

www.spice-rtn.orgSPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland

Conclusions

No pronounced difference between GCA, ERT or FFT corrections

Corrections at 150 sec are better than at 40 sec

Corrections get worse as distance increases

We will now check if the imperfect corrections will lead to a detectable mantle signal