Three‐dimensional crustal structure in central Taiwan from gravity inversion with a parallel genetic algorithm

Abstract

The genetic algorithm method is combined with the finite‐element method for the first time as an alternative method to invert gravity anomaly data for reconstructing the 3D density structure in the subsurface. The method provides a global search in the model space for all acceptable models. The computational efficiency is significantly improved by storing the coefficient matrix and using it in all forward calculations, then by dividing the region of interest into many subregions and applying parallel processing to the subregions. Central Taiwan, a geologically complex region, is used as an example to demonstrate the utility of the method. A crustal block 120 × 150 km2 in area and 34 km in thickness is represented by a finite‐element model of 76 500 cubic elements, each 2 × 2 × 2 km3 in size. An initial density model is reconstructed from the regional 3D tomographic seismic velocity using an empirical relation between velocity and density. The difference between the calculated and the observed gravity anomaly (i.e., the residual anomaly) shows an elongated minimum of large magnitude that extends along the axis of the Taiwan mountain belt. Among the interpretive models tested, the best model shows a crustal root extending to depths of 50 to 60 km beneath the axis of the Western Central and Eastern Central Ranges with a density contrast of 400 or 500 kg/m3 across the Moho. Both predictions appear to be supported by independent seismological and laboratory evidence.

Precollisional tectonics and terrain amalgamation offshore southern Taiwan: Characterizations from reflection seismic and potential field data

“…Crustal thicknesses revealed from OBS experiments west of Hengchun Peninsula and Manila Trench suggested a progressive southeastward crustal thinning from an extended continental crust of 11 km thick to an oceanic crust of 5 to 6 km thick [19,20] . Zhang et al [21,22] recently analyzed the density structure of the Central Ridge of Taiwan through parallel computation and inversion of 3D gravity data. Various earthquake hypocenter relocations and tomographic studies [23][24][25][26][27] are significant in delineating the subduction of the SCS plate beneath the Philippine Sea plate, the Eurasian continental subduction beneath the Taiwan orogen, as well as the transition from the subduction to collision in southeastern Taiwan.…”

Three-dimensional modelling of gravity data using finite differences

“…Many of the methods developed thus far discretize the model in terms of a regular rectangular mesh in which each cell has a uniform density (e.g., Bear et al, 1995;Li and Oldenburg, 1998;Portniaguine and Zhdanov, 1999;Boulanger and Chouteau, 2001;Nagihara and Hall, 2001;Chasseriau and Chouteau, 2003;Zhang et al, 2004). This style of mesh has a number of advantages.…”

“…An example of the use of the forward solver in a minimum-structure inversion algorithm is given in Farquharson, 2008. ) Forward modelling of gravity data via numerical solution of Poisson's equation, by a finite-element technique rather than a finite-difference one, is used by Zhang et al (2004) in their 3-D inversion program. Zhang et al discretize the subsurface using a regular rectangular mesh, and use a genetic algorithm for the inversion.…”

Gravity inversion using the frequency characteristics of the density distribution

“…Yao et al (2003) discussed problems for gravity inversion with GA to enhance inversion effi ciency. Zhang et al (2004) proposed a hybrid method using the finite element method (FEM) and GA to invert gravity anomaly data. Cai and Wang (2005) developed the FEM calculation method of complex heterogeneous regions for forward density modeling problems.…”