10:00 AM - 10:15 AM
[SGD02-05] A comparison between Green's functions for modeling crustal deformation and application for coseismic fault slip inversions for the 2011 off the Pacific coast of Tohoku Earthquake
Keywords:Fault slip inversion, Green's function, The 2011 earthquake off the Pacific coast of Tohoku
Crustal deformation during earthquakes has been observed by GNSS and seafloor crustal observations. Coseismic displacement can be well explained by the response to the dislocation of elastic bodies, and coseismic fault slip can be estimated by the inversion of observed displacement data. A model assuming a homogeneous half-space is often used as Green's function to describe elastic deformation due to displacement. Some half-space models consider the layered structure. There are also spherical models that consider self-gravitation and layered structure. Especially in the case of large earthquakes, spherical models are needed since the spatial scale of crustal deformation exceeds 100 km. Several studies applied the models proposed above and compared the results of the forward calculation based on the same slip, but only a few cases examined the effect of inversion results. In this study, we carried out the inversion of the following three models and compared the results.
A flat model assuming a Poisson medium (Okada (1985)), a half-space model using PREM layered structure (Wang et al. (2003), hereafter referred to as "flat layered model"), and a spherical model with self-gravitation using PREM (Tanaka et al. (2014), hereafter referred to as "spherical layered model"). We evaluate the effects of (1) curvature, self-gravitation, and (2) layered heterogeneity to the depth direction on a slip estimation for the fault slip inversion of the 2011 off the Pacific coast of Tohoku Earthquake. In the inversion, we used GNSS data from the Geospatial Information Authority of Japan (GSI) and seafloor crustal deformation data from the Japan Coast Guard, and each estimation of models was determined using ABIC.
The results show that the minimum ABIC value is the smallest in spherical layered, followed by flat layered and flat models. The RMS of horizontal displacement is also smaller in the same order. These results indicate that the ABIC values and the reproducibility of horizontal displacement significantly improved by considering layered structure, curvature, and self-gravitation. The layered structure improved the reproducibility of both horizontal and vertical displacements at distant sites, and the curvature and self-gravitation further enhanced the reproductivity beyond the epicentral distance of 400 km. The maximum estimated slips are 45 m for flat, 40 m for flat layered, and 39 m for spherical layered, with Mw of 8.98 for flat, 9.06 for flat layered, and 9.05 for spherical layered, respectively.
To clarify the cause of the difference in slip distribution, we carried out forward calculations for each model with point sources at depths ranging from 10 km to 40 km, which is the estimated main slip distribution area. At the epicentral distance of 200 km, which is off the Sanriku coast for the Tohoku-oki earthquake case, the difference between flat and flat layered models is more considerable than that between flat layered and spherical layered models, especially at a source depth of 40 km, the former is notably larger than the latter. At the epicentral distance of 400 km, which is the Sea of Japan side for the Tohoku-oki earthquake case, the displacements of the flat model do not tend to be significantly different from those of the other two models as is the case of 200 km, while the difference between the flat layered and spherical layered models, for example at a source depth of 10 km, becomes twice as large as that at the case of 200 km. Although we have not yet identified the reason for the difference in the slip distribution, the results of these forward calculations indicate that the difference between the models can be significantly detected by observation. These results indicate that the difference in Green's functions should be considered when discussing the uncertainty of the coseismic slip distribution for the M9 event.
A flat model assuming a Poisson medium (Okada (1985)), a half-space model using PREM layered structure (Wang et al. (2003), hereafter referred to as "flat layered model"), and a spherical model with self-gravitation using PREM (Tanaka et al. (2014), hereafter referred to as "spherical layered model"). We evaluate the effects of (1) curvature, self-gravitation, and (2) layered heterogeneity to the depth direction on a slip estimation for the fault slip inversion of the 2011 off the Pacific coast of Tohoku Earthquake. In the inversion, we used GNSS data from the Geospatial Information Authority of Japan (GSI) and seafloor crustal deformation data from the Japan Coast Guard, and each estimation of models was determined using ABIC.
The results show that the minimum ABIC value is the smallest in spherical layered, followed by flat layered and flat models. The RMS of horizontal displacement is also smaller in the same order. These results indicate that the ABIC values and the reproducibility of horizontal displacement significantly improved by considering layered structure, curvature, and self-gravitation. The layered structure improved the reproducibility of both horizontal and vertical displacements at distant sites, and the curvature and self-gravitation further enhanced the reproductivity beyond the epicentral distance of 400 km. The maximum estimated slips are 45 m for flat, 40 m for flat layered, and 39 m for spherical layered, with Mw of 8.98 for flat, 9.06 for flat layered, and 9.05 for spherical layered, respectively.
To clarify the cause of the difference in slip distribution, we carried out forward calculations for each model with point sources at depths ranging from 10 km to 40 km, which is the estimated main slip distribution area. At the epicentral distance of 200 km, which is off the Sanriku coast for the Tohoku-oki earthquake case, the difference between flat and flat layered models is more considerable than that between flat layered and spherical layered models, especially at a source depth of 40 km, the former is notably larger than the latter. At the epicentral distance of 400 km, which is the Sea of Japan side for the Tohoku-oki earthquake case, the displacements of the flat model do not tend to be significantly different from those of the other two models as is the case of 200 km, while the difference between the flat layered and spherical layered models, for example at a source depth of 10 km, becomes twice as large as that at the case of 200 km. Although we have not yet identified the reason for the difference in the slip distribution, the results of these forward calculations indicate that the difference between the models can be significantly detected by observation. These results indicate that the difference in Green's functions should be considered when discussing the uncertainty of the coseismic slip distribution for the M9 event.