10:45 AM - 12:15 PM
[SGD02-P18] Comparison of numerical results of coseismic deformations between spherically symmetric and semi-infinite models
Keywords:elastic deformation model, the 2011 earthquake off the Pacific coast of Tohoku, crustal deformation, gravity change, inversion, forward modeling
Green's functions are used to calculate coseismic deformations through forward modeling and inversion. The characteristics of the Green's functions vary depending on how much each elastic dislocation model simplifies the complexity of the real Earth. For example, Okada (1985, 1992) and Okubo (1991, 1992) assumed the Earth's model of a uniform semi-infinite medium without gravity (hereafter referred to as the O model). In contrast, Sun et al. (2009) considered the curvature of the Earth's surface, self-gravitation and vertical heterogeneity in their elastic model (hereafter referred to as the S model). Because the Green's functions of these two models are different with each other in terms of variations with epicentral distance (Sun and Okubo, 2002; Dong et al., 2014), inversion results of coseismic slip distributions may vary depending on which model is chosen for the inversion. However, no previous study has compared the difference in the effect of near-field data on the inversion results obtained by the two models. In addition, no one has fully discussed the effects of the difference in the slip distributions obtained by the two models on forward calculation results, although Wang et al. (2010) and Zhou et al. (2012) compared several forward results calculated using the two models with slip distributions obtained mainly from seismic inversions. Therefore, we calculated slip distributions and deformation fields for the 2011 Tohoku earthquake using the S and O models, to investigate the differences in the inverse/forward results between the two models.
We first performed fault slip inversions from crustal deformation data. We here used the S and O models to prepare the corresponding Green's functions, and inverted the following two datasets to the fault slip distributions: (1) only terrestrial crustal deformation data observed by GEONET, and (2) seafloor crustal deformation data (Kido et al., 2011; Sato et al., 2011) together with the GEONET data. We obtained four inversion results named S1, S2, O1 and O2 according to the combination of a model and a dataset used for each inversion. Moment magnitudes were estimated to be (S1) 9.18, (S2) 9.15, (O1) 8.98 and (O2) 8.97. This result shows that the magnitude values of the S model are greater than those of the O model by approximately 0.2. Also, maximum slips were calculated to be (S1) 78.36 m, (S2) 48.94 m (O1) 39.28 m and (O2) 46.90 m. This result indicates that when near-field seafloor data are added to the inversions, the maximum slip increases in the O-model case while it decreases in the S-model case. These features can be explained by the characteristics of the Green's functions that the deformation of the S model becomes larger in the near field and smaller in the far field than that of the O model.
We also performed the forward calculation for the data set of far-field crustal deformation and terrestrial gravity change, using the slip distributions obtained from the inversions. We here used the slip distributions of S2 and O2 in the previous paragraph, and prepared the Green's functions corresponding to the S and O models. We obtained four data sets of forward calculations named S2-S, S2-O, O2-S and O2-O, according to the combination of a slip distribution and an elastic model used for each forward calculation. We then quantitatively compared the calculated data with the observed one that had not been included in the inversions, by estimating the RMS residual between the observed and calculated data for each forward calculation. We found that the S2-S's residual was the smallest of the four forward calculations. This result shows that the S's elastic model should be used consistently to reproduce both of the slip distribution and observed data accurately. Moreover, the RMS residual of O2-S was found to be smaller than that of O2-O. This result means that the Green's function of the S model should be used for more accurate forward calculation, even if it is not used in the inversion.
We first performed fault slip inversions from crustal deformation data. We here used the S and O models to prepare the corresponding Green's functions, and inverted the following two datasets to the fault slip distributions: (1) only terrestrial crustal deformation data observed by GEONET, and (2) seafloor crustal deformation data (Kido et al., 2011; Sato et al., 2011) together with the GEONET data. We obtained four inversion results named S1, S2, O1 and O2 according to the combination of a model and a dataset used for each inversion. Moment magnitudes were estimated to be (S1) 9.18, (S2) 9.15, (O1) 8.98 and (O2) 8.97. This result shows that the magnitude values of the S model are greater than those of the O model by approximately 0.2. Also, maximum slips were calculated to be (S1) 78.36 m, (S2) 48.94 m (O1) 39.28 m and (O2) 46.90 m. This result indicates that when near-field seafloor data are added to the inversions, the maximum slip increases in the O-model case while it decreases in the S-model case. These features can be explained by the characteristics of the Green's functions that the deformation of the S model becomes larger in the near field and smaller in the far field than that of the O model.
We also performed the forward calculation for the data set of far-field crustal deformation and terrestrial gravity change, using the slip distributions obtained from the inversions. We here used the slip distributions of S2 and O2 in the previous paragraph, and prepared the Green's functions corresponding to the S and O models. We obtained four data sets of forward calculations named S2-S, S2-O, O2-S and O2-O, according to the combination of a slip distribution and an elastic model used for each forward calculation. We then quantitatively compared the calculated data with the observed one that had not been included in the inversions, by estimating the RMS residual between the observed and calculated data for each forward calculation. We found that the S2-S's residual was the smallest of the four forward calculations. This result shows that the S's elastic model should be used consistently to reproduce both of the slip distribution and observed data accurately. Moreover, the RMS residual of O2-S was found to be smaller than that of O2-O. This result means that the Green's function of the S model should be used for more accurate forward calculation, even if it is not used in the inversion.