Abstract
Gridding algorithm proposed by Shiono et al.(1987) is designed to determine the optimal geologic surface that minimizes the augmented objective function Q=J(smoothness)+α(penalty) * R(goodness of fit) based on an exterior penalty function method. The algorithm provides a powerful mean to make use of inequality and slope information as constraints of the surface. However, as the solution depends on the value of penalty and number of grid, a series of trial and error is required to find the most proper values for each set of data. We revised the Fortran program from a viewpoint of rationalization of trial and error. The revised program evaluates the smoothness of surface in a form of numerical integration and the goodness of fit in a form of residual mean of squares, and shows the smoothness of surface and the goodness of fit for each solution on the display. Through calculations for a sparsely distributed data, it is confirmed that the solution depends only on a value of penalty but not on number of grid as expected from the evaluation of smoothness. Through the examination of a densely distributed inequality data, we found that the revised program is efficient to find a proper value of penalty and number of grid.
