Preliminary 3D inversion of MT data in Pohang including sea effects using conjugate gradients

We develop a three-dimensional (3D) inversion procedure that efficiently inverts magnetotelluric (MT) data. The inversion algorithm employs conjugate gradient (CG) relaxation to solve a sequence of linearized inverse problems, which are formulated based on the Gauss-Newton (GN) method. The implementation of CG techniques requires only the effect of the sensitivity matrix or its transpose multiplying a vector. Thus, the procedure allows us to avoid two computationally intensive tasks of conventional GN-based inversion procedures, which are performed at each step of GN iteration, i.e., the computation of the sensitivity matrix and complete solution of a linear system on the model space. In Korea, several MT surveys have been conducted for low temperature geothermal exploration in 2002 and 2003 in Pohang that is a coastal city locating south east of Korea. Since MT data obtained from the surveys are basically affected by sea effects, it is important to precisely consider sea effects in inversion for a more accurate interpretation of the MT data. For the interpretation of MT data in Pohang, a GN-based inversion including sea effects was made by excluding seawater from the inversion domain in order to fix the resistivity, while including sea in the modeling domain for the simulation of sea effects. However, since the GN based inversion process needs to use larger inversion blocks than modeling blocks to reduce computation burden, the coast line and bathymetry described by inversion blocks may not be able to explain sea effects on MT data in Pohang. However, In the CG approach, inversion blocks are identical to modeling blocks within the inversion domain. As a result, inversion blocks of the approach can describe coast lines better than those of conventional GN-based inversion procedures. Therefore, the CG-based procedure uses the same blocks for both inversion and modeling, the procedure developed in this study can be the best scheme to interpret MT data distorted by sea effects.