Manual and automated interpolation of climatic and geomorphic statistical surfaces: An evaluation

A survey of the literature reveals that many interpolation algorithms have been proposed for generating statistical surfaces of geographical phenomena from a finite set of observations and that there is no one best method for handling even a single class of phenomena. One also finds that manually-produced surfaces are often used as the standard against which computer-generated surfaces are compared, implying that the manual method yields better results. To date, little research has compared manual and computer interpolation, assessed whether expert knowledge has a consistent impact on manual interpolation, and evaluated objectively the differences between manual and computer interpolation. This study surveys expert opinion on computer interpolation, compares the metric accuracy of manual and computer-generated surfaces, and reports expert evaluations of these surfaces. Experts indicated that interpolation algorithms are in need of further development in order to produce maps that are comparable in character to those produced manually. Comparison of metric accuracy of interpolated surfaces at eliminated data points showed: 1) differences between expert manually-drawn maps are not significant; and 2) the accuracy of computer maps equals or surpasses that of expert manually-drawn maps. Accuracy notwithstanding, expert evaluation of the computer maps report: 1) dissatisfaction with the computer portrayals of the spatial characteristics of the phenomena mapped; and 2) consistently lower ratings as compared to manual maps. The study notes weaknesses in current interpolation algorithms and offers editorial guidelines for interactive resolution of these deficiencies.