Development of the spatiotemporal trans-dimensional source inversion

Abstract:

The trans-dimensional inversion, which estimates the dimension of model parameters as well as values of model parameters, has been recently applied in the geophysical field (e.g. Agostinetti and Malinverno 2010; Bodin et al. 2012) including the source-process analysis (e.g. Dettmer 2014; Kubo et al. 2019 AGU; Hallo and Gallovič 2020; Tomita et al. 2020, JpGU). The advantages of the trans-dimensional inversion on source-rupture analysis are (1) that the discretization on fault, which has been subjectively assumed by a modeler in conventional source inversion, can be directly determined by data, and (2) that it does not require the smoothing constraint on slips, which significantly reduces the inversion resolution and sometimes leads to the biased solution. Dettmer (2014) and Hallo and Gallovič (2020) developed the trans-dimensional source inversion with waveform data. In these studies, the rupture history at each point on fault was modeled by a predetermined functional shape and the fault rupture evolution was expressed by slips and parameters of the predetermined function in 2D fault space. One disadvantage of this approach is that the expressible shape of the source time function is limited. To overcome this problem, in this study, we introduce the spatiotemporal trans-dimensional inversion to the source-process analysis. Here, the fault rupture evolution is expressed by slips in 2D fault space and time, which are discretized by Voronoi cells. Because the rupture history at each point on fault is expressed by a time-series of several slips, this approach has high flexibility regarding the shape of the source time function.

We will present the result that this novel approach is applied to waveform records of the 2016 central Tottori, Japan, earthquake (Kubo et al. 2017).