12:00 〜 12:15
[SCG52-12] Toward a quantitative analysis of inelastic crustal deformation: performance evaluation of elastic prediction method
キーワード：Inelastic deformation, Elasticity and inelasticity, Intraplate processes
In the interseismic time period, the stress accumulation in intraplate crust induced by inelastic deformation at depth will be released as a form of future earthquakes, it is considered to be responsible for the stress build-up for intraplate earthquakes. Thus it is important to evaluate the inelastic deformation in the local area. However, there is no standard method to separate the inelastic deformation from the local deformation field so far. Therefore, it is necessary to develop a general method from geodetically observed signals such as GNSS, which we call the elastic prediction method.
In this method, we assume the deformation observed is the sum of elastic and inelastic deformation. We choose an optimum boundary selection to describe the contributions from outer deformation sources as boundary conditions. By assuming a perfectly elastic medium and plane stress condition inside the boundary, the surface velocities for interior sites induced by the outer deformation sources can be predicted with boundary conditions. We predict the interior velocities with a 2D interpolation approach developed by Sandwell and Wessel (2016) based on the Green functions of an elastic plate under the plane stress condition. The residual velocities between predictions and observations for interior sites indicate the contributions from the interior deformation source. With a residual model, we can finally evaluate the inelastic deformation at depth.
We first applied the method to simulation data to check the feasibility of the elastic prediction method. The simulation data is uniform strain rate data with an inelastic deformation source added with random Gaussian errors. We tested three factors that could affect the prediction accuracy: (1) the errors of the data, (2) the boundary site density, (3) the distance of the boundary selection to the interior sites, and then evaluated the performance with RMSE values.
The simulation results show that the prediction errors increase with the increase of random errors but still within the acceptance compared with the large random errors. But for very small signals such as 0.1mm/yr of slipping rate and 0.8mm/yr added random errors, the RMSE is about 0.26mm/yr, the surface response is too small to detect. Also, we found that it’s better to use more boundary sites to improve the prediction accuracy. Meanwhile, we should set the boundary sites far enough to keep the effect from the interior deformation source into minimum.
Next, we applied the elastic prediction method to the San Jacinto fault, real observation data to check how is the performance. After we got the residual velocities of the interior sites, we constructed a residual model using a 2D dislocation model. The fitting result shows that 10mm/yr of slipping rate roughly fits with the residual velocities and is reasonable for the San Jacinto fault, which is also consistent with the previous studies of ~10-12mm/yr slipping rate (Sharp, 1981; Rockwell et al., 1990; Wesnouski et al., 1991; Bennett et al., 1996; Kendrick et al., 1994). Notably, there is a shifting of the fault trace in the residual figure of fault parallel component, this might be caused by factors such as imbalance of the boundary at two sides, the existence of extra inelastic deformation sources inside the boundary, or the fault dip angle. Further improvements of the method will be done with more real observation data tests.
In this method, we assume the deformation observed is the sum of elastic and inelastic deformation. We choose an optimum boundary selection to describe the contributions from outer deformation sources as boundary conditions. By assuming a perfectly elastic medium and plane stress condition inside the boundary, the surface velocities for interior sites induced by the outer deformation sources can be predicted with boundary conditions. We predict the interior velocities with a 2D interpolation approach developed by Sandwell and Wessel (2016) based on the Green functions of an elastic plate under the plane stress condition. The residual velocities between predictions and observations for interior sites indicate the contributions from the interior deformation source. With a residual model, we can finally evaluate the inelastic deformation at depth.
We first applied the method to simulation data to check the feasibility of the elastic prediction method. The simulation data is uniform strain rate data with an inelastic deformation source added with random Gaussian errors. We tested three factors that could affect the prediction accuracy: (1) the errors of the data, (2) the boundary site density, (3) the distance of the boundary selection to the interior sites, and then evaluated the performance with RMSE values.
The simulation results show that the prediction errors increase with the increase of random errors but still within the acceptance compared with the large random errors. But for very small signals such as 0.1mm/yr of slipping rate and 0.8mm/yr added random errors, the RMSE is about 0.26mm/yr, the surface response is too small to detect. Also, we found that it’s better to use more boundary sites to improve the prediction accuracy. Meanwhile, we should set the boundary sites far enough to keep the effect from the interior deformation source into minimum.
Next, we applied the elastic prediction method to the San Jacinto fault, real observation data to check how is the performance. After we got the residual velocities of the interior sites, we constructed a residual model using a 2D dislocation model. The fitting result shows that 10mm/yr of slipping rate roughly fits with the residual velocities and is reasonable for the San Jacinto fault, which is also consistent with the previous studies of ~10-12mm/yr slipping rate (Sharp, 1981; Rockwell et al., 1990; Wesnouski et al., 1991; Bennett et al., 1996; Kendrick et al., 1994). Notably, there is a shifting of the fault trace in the residual figure of fault parallel component, this might be caused by factors such as imbalance of the boundary at two sides, the existence of extra inelastic deformation sources inside the boundary, or the fault dip angle. Further improvements of the method will be done with more real observation data tests.