A Java-based intelligent advisor for selecting a context-appropriate spatial interpolation algorithm

The choice of an appropriate interpolation technique is an imposing task. There are many techniques to choose from depending on the characteristics of the phenomenon, the purpose of the study and also the expertise of the user. This paper describes the implementation and architecture of a Java-based intelligent advisor to assist a generic user, ranging from the casual to the specialist, in selecting the interpolator most appropriate for a given task and data set. The software integrates procedural knowledge from disparate sources through an expert system shell, and uses multiple sensory channels to present and abstract contextual knowledge from the user regarding both the data and the task at hand. The system has been tested for use in a meteorological domain in the first instance. The interpolation methods currently assessed by the system are multiple forms of kriging, thin plate smoothing splines, inverse-distance weighting and trend surface/polynomial analysis. r 2004 Elsevier Ltd. All rights reserved.