Improving the selection of appropriate spatial interpolation methods

Abstract

In many natural sciences as well as in spatial data processing data describing continuous fields are a major information type. One of the main challenges when using such field representations in analysis and modelling arises from restrictions caused by the discretization that occurs when sampling the field. As a consequence, interpolation is often required before using the data in a particular application. In this paper the family of kriging methods is used as an example to illustrate the decision-making process when selecting appropriate interpolation methods. The analysis of the properties of various kriging methods shows that the decision-making process should be based on a considerable body of information, including implicit and external knowledge, i.e., information not related and derivable from the data.

The acquisition of the information needed and the examination of the relevant data characteristics needs can be a demanding procedure, including many statistical tests and other means of exploratory data analysis. Extended Exploratory Data Analysis is a strategy to support the decision process when selecting an interpolation method. This standardized procedure supports the derivation of implicit information and navigates users through the decision process. In combination with Virtual Data Sets Extended Exploratory Data Analysis results in more reliable estimated field values.