Soil survey information with quantified accuracy is relevant to decisions on land use and environmental problems. To obtain such information statistical strategies should be used for collecting and analysing data. A survey project based on a statistical sampling strategy requires a soil survey scheme specifying which sites are to be sampled, which data are to be recorded and how they are to be analysed statistically. The efficiency of such a scheme is determined by the accuracy of the survey results and the cost of operation. This accuracy and cost depend mainly on the method of determination and the sampling design in the scheme.
This study aimed at formulating the basic design considerations of a knowledge-based system (KBS) to assist in the design of soil survey schemes. This system should incorporate pedological and statistical knowledge. The domain of the system has provisionally been limited to surveys for which a design-based approach, i.e. the use of classical sampling theory, is appropriate.
Initially, the domain of the system has been structured in three layers: (i) an entity structure clarifying the position of the system in a soil survey project; (ii) a model describing the design process as a number of interrelated steps, and (iii) a conceptual framework defining the main concepts and their relations.
Further analysis made it possible to specify the tasks in which the KBS should assist: definition of the survey request, selection of prior information, design of outlinear schemes, evaluation and optimization of outlinear schemes, generation of a report, and evaluation a posteriori .
The system will primarily assist in the statistical decisions in the design process. Since there was no suitable classification of sampling designs available, a hierarchical framework of sampling designs has been constructed, in which sampling designs are grouped into types of designs, and types are grouped into classes of designs. Furthermore the main classes of sampling designs treated in the literature have been ordered in a taxonomy. Decision trees have been developed to guide the selection of an appropriate sampling approach (designbased versus model-based), and, in the case of a design-based approach, to guide the search for an appropriate class of sampling designs.
To ensure that the available means for a project, such as budget, personnel, and equipment, are used adequately schemes should be evaluated and optimized beforehand. Models related to the features of sampling designs have been developed for predicting the accuracy and cost of survey schemes, the so-called prior evaluation. Furthermore the use of dynamic programming is proposed to search for the optimal sampling design within an outlinear scheme. The procedure enables objective comparison of schemes taking into account differences in spatial variability or sampling cost among sub-regions.
Finally, basic design considerations are presented consisting of an initial requirements definition, a description of the intended use of the KBS, and a specification of the components for an actual KBS. Five components are distinguished: a database, a knowledge base, a model base, a problem-solving model, and a user interface. The system will assist in its own maintenance through continuous storage of knowledge from executed projects. This will facilitate the re-use of information. A KBS which is based on these basic design considerations will assist in controlling the quality of soil survey projects.
|Qualification||Doctor of Philosophy|
|Award date||28 Oct 1994|
|Place of Publication||S.l.|
|Publication status||Published - 1994|
- soil surveys
- land evaluation
- soil suitability
- soil analysis
- computer simulation
- simulation models