Gradient Approximation on Uniform Meshes by Finite Differences and Cubic Spline Interpolation

摘要

For the approximation of gradients from data values at vertices of a uniform grid, we compare two methods based on cubic spline interpolation with a classical method based on finite differences. For univariate cubic splines, we use the so-called de Boor's Not a Knot property and a new method giving pretty good slopes. Then these methods are used on parallels to the axes for estimating gradients on bivariate grids. They are illustrated by several numerical examples.