1. The fitting of an analytic surface (the regional) to data points as a representation of the ‘‘order’’ in the data, as opposed to the erratic element (the residual). The objective may be to analyze trends in the data or to interpolate between data points. The underlying assumption is that the data may be decomposed into a relatively low-order, smooth surface plus more-or-less random noise. The number of independent parameters in the analytic surface should be appreciably less than (usually less than 10 percent of) the number of data points. Also called surface fitting. 2. The 2D Fourier analysis of a surface into spatial frequency components (or wavelengths) so that one can determine preferred orientations and wavelengths.