Naturally induced electromagnetic fields (magnetotelluric fields) are highly dependent on both the vertical and horizontal conductivity structure of the earth and are greatly influenced by three-dimensional features. Three-dimensional magnetotelluric modeling algorithms are critical for correctly interpreting observed data, but the current generation of algorithms are computationally expensive and time-consuming. The objective of this proposal is to develop three-dimensional magnetotelluric modeling algorithms that can model fully inhomogeneous earth models and that are quick and can be run on modern scientific workstations. The algorithm to be developed through this project will use conjugate direction relaxation algorithms coupled with appropriate preconditioners and a multiple scaling technique.