The finite-difference method (FDM) for solving three-dimensional (3-D) resi
stivity problems has traditionally used a graded, rectangular grid whose sp
acings change independently in orthogonal coordinate axis directions. Small
cell sizes are used to represent the field around external sources or fine
resistivity features. The cell sizes are increased gradually toward the bo
undaries of a computational domain. Typically, cells can have very large as
pect ratios, especially near the computational domain boundaries. Large rou
nd-off errors and slow convergence of (iterative) numerical solutions to th
e finite-difference (PD) equation system may result. In this paper, we pres
ent an upgridding approach to improve the efficiency of the FDM with a conv
entional rectangular grid. The upgridding process coalesces cells of extrem
al shapes in the directions of short dimensions to reduce cell aspect ratio
s and the total number of unknowns. Our experiments with a set of 3-D resis
tivity models show that the upgridding FDM can reduce the computation time
by nearly half relative to using the FDM with a graded, rectangular grid.