SIE: A Digital Soil Mapping Tool for Soil Scientists.

Generally there are two approaches to DSM. One aims at truly automatic, objective, and quantitative mapping, taking advantage of the techniques in statistics, geostatistics, machine learning, and data mining, and generally relying heavily on densely sampling. The other tries to fit within the conventional soil survey and mapping framework, and aims to effectively utilize the soil scientist’s knowledge, while reduce the inconsistency and cost associated with the traditional manual process. We articulate two advantages of the knowledge-based approach. First, it does not rely on dense field sampling. Second, it might be more acceptable to soil surveyors. These advantages justify the adoption of this approach, especially in the US and other developed countries where more soil scientists and associated soil knowledge are available. Analysis of the soil scientist’ knowledge provides guidelines for the development of DSM tools that implement the knowledge-based approach. We analyze soil scientists’ knowledge from the perspectives of scale and space. We distinguish global knowledge that covers the entire mapping area and local knowledge that is only applicable to certain local regions. We also distinguish knowledge represented by environmental values in parametrical space and knowledge represented by locations in geographical space. Rule-based reasoning (RBR) is proposed for handling the global knowledge in parametrical space, global case-based reasoning (CBR) for the global knowledge in geographical space, and local CBR for the local knowledge in geographical space. The final soil mapping products should represent an integration of knowledge and inferences of all different types. A software tool, named Soil Inference Engine (SIE), was developed to be used by soil scientists and facilitate an eight-step integrated RBR-CBR process. SIE was tested in northern Vermont and proved to be effective. The soil scientist working on the project was generally satisfied with the results from SIE, in terms of both quality and cost.