11:15 〜 11:30
[SSS04-09] High Seismic Attenuation in the Lower Crust and Subducting Oceanic Crust (with Emphasize to the Problems of the Spectral Inversion Method)
キーワード：GIT, spectral inversion, inversion trade-off, attenuation tomography, subduction zone
Introduction. This presentation reviews experience of the spectral inversion with a specific target to the study of seismic attenuation. Knowing of seismic attenuation are important for the strong ground motion simulation and helpful to constrain tectonic models. Studied region is the Philippine Sea subduction zone. Low-frequency earthquakes, which indicate presence of fluids, are observed within the Subducting Oceanic Crust (SOC) and Lower crust (LC, non-volcanic) here.
Attenuation tomography. Original method of inversion for attenuation structure is presented. Geometrical spreading factor, which is necessary to exclude before inversion for the Q-value, is calculated numerically using 3-D velocity structure model and ray theory. Another feature is adoption of the double-spectral ratio to reduce trade-offs between source and/or site effects and Q-value.
Problems of the Spectral Inversion Method. Spectral inversion (also known as the Generalized Inversion Technique, GIT) is considered as a simple method that allow estimating of a set of parameters simultaneously. However, as any inversion technique it has trade-offs between parameters that require special treatment to solve them. In this study we rise problems and propose solutions for the next list of problems of the method:
1.Trade-off between path effect (Q-value) and source/site effects. Reason: large data residuals in comparison with a small path effect (Petukhin et al., 2003).
2.Trade-off between Q-value and source effect is especially large.
3.Solution the above trade-off: double spectral ratio method (adopted to the attenuation tomography; Petukhin and Kagawa, 2007).
4.Another solution: smoothing of the observed amplitude spectra.
5.Trade-off between source and site effects is well known. Solution: independent estimation of velocity structure, and site effect respectively, for the reference site(s), e.g., by the receiver function method (Petukhin and Irikura, 2000).
6.Large effect of the geometrical spreading factor at the local distances (< 80km).
7.Estimation of the geometrical spreading factor: large effect of the shallow structure.
8.Estimation of the shallow structure: receiver function method.
9.Tectonically consistent block model for the attenuation tomography.
10.On the selection of the S-wave segment of record.
11.Trade-off due to non-uniform data distribution over distance.
Results. Generally, estimated Q-values have good consensus with results of other studies and with physical expectations based on tectonic structure, except for one remarkable result that Q-values in LC and SOC are extremely low (attenuation is high): Q ~ 20-30f (see Fig.1). Comparison with other phenomena related to attenuation that were observed in studied region (see Fig.2: low-frequency earthquakes, Obara, 2002, and reflective layers, Ito, 1999) let us understand that: (1) most probable candidate to explain low Q-value is the loss of energy due to reflections within reflective layers; (2) seismic tomography results indicate high Vp/Vs ratios within SOC and LC, which indicate presence of fluids and high attenuation respectively.
Acknowledgement: Waveform data of seismic networks K-NET, Kik-net, Hi-net, and CEORKA, as well as hypocenter data of JMA are used. Author is thankful to Takao Kagawa, Ken Miyakoshi and Kojiro Irikura for years of fruitful discussions and support of this work.
References.
Ito K. (1999): Tectonophysics, 306, 423-433.
Obara K. (2002): Science, 296, 1679-1681.
Petukhin A, K. Irikura (2000): Geophys.Res.Lett., 27, 3429-3432.
Petukhin A., K. Irikura, S. Ohmi, T. Kagawa (2003): Bull.Seismol.Soc.Am., 93, 1498-1515.
Petukhin A. and T. Kagawa (2007): Geophysical Monograph Series 172, 117-127.
Attenuation tomography. Original method of inversion for attenuation structure is presented. Geometrical spreading factor, which is necessary to exclude before inversion for the Q-value, is calculated numerically using 3-D velocity structure model and ray theory. Another feature is adoption of the double-spectral ratio to reduce trade-offs between source and/or site effects and Q-value.
Problems of the Spectral Inversion Method. Spectral inversion (also known as the Generalized Inversion Technique, GIT) is considered as a simple method that allow estimating of a set of parameters simultaneously. However, as any inversion technique it has trade-offs between parameters that require special treatment to solve them. In this study we rise problems and propose solutions for the next list of problems of the method:
1.Trade-off between path effect (Q-value) and source/site effects. Reason: large data residuals in comparison with a small path effect (Petukhin et al., 2003).
2.Trade-off between Q-value and source effect is especially large.
3.Solution the above trade-off: double spectral ratio method (adopted to the attenuation tomography; Petukhin and Kagawa, 2007).
4.Another solution: smoothing of the observed amplitude spectra.
5.Trade-off between source and site effects is well known. Solution: independent estimation of velocity structure, and site effect respectively, for the reference site(s), e.g., by the receiver function method (Petukhin and Irikura, 2000).
6.Large effect of the geometrical spreading factor at the local distances (< 80km).
7.Estimation of the geometrical spreading factor: large effect of the shallow structure.
8.Estimation of the shallow structure: receiver function method.
9.Tectonically consistent block model for the attenuation tomography.
10.On the selection of the S-wave segment of record.
11.Trade-off due to non-uniform data distribution over distance.
Results. Generally, estimated Q-values have good consensus with results of other studies and with physical expectations based on tectonic structure, except for one remarkable result that Q-values in LC and SOC are extremely low (attenuation is high): Q ~ 20-30f (see Fig.1). Comparison with other phenomena related to attenuation that were observed in studied region (see Fig.2: low-frequency earthquakes, Obara, 2002, and reflective layers, Ito, 1999) let us understand that: (1) most probable candidate to explain low Q-value is the loss of energy due to reflections within reflective layers; (2) seismic tomography results indicate high Vp/Vs ratios within SOC and LC, which indicate presence of fluids and high attenuation respectively.
Acknowledgement: Waveform data of seismic networks K-NET, Kik-net, Hi-net, and CEORKA, as well as hypocenter data of JMA are used. Author is thankful to Takao Kagawa, Ken Miyakoshi and Kojiro Irikura for years of fruitful discussions and support of this work.
References.
Ito K. (1999): Tectonophysics, 306, 423-433.
Obara K. (2002): Science, 296, 1679-1681.
Petukhin A, K. Irikura (2000): Geophys.Res.Lett., 27, 3429-3432.
Petukhin A., K. Irikura, S. Ohmi, T. Kagawa (2003): Bull.Seismol.Soc.Am., 93, 1498-1515.
Petukhin A. and T. Kagawa (2007): Geophysical Monograph Series 172, 117-127.