Typescript, etc. / Digitized by Kansas State University Libraries

Typescript, etc. / Digitized by Kansas State University Libraries

No description available.

Computerized soil information systems (SIS) store and retrieve, much more efficiently than soil maps, the soil information collected from field survey which is essential for land-use planning. The soil is only observed and sampled at a limited number of locations, depths and for a limited number of properties. Information not recorded during survey is missing and if needed must be generated or predicted. New geostatistical techniques for spatial analysis and interpolation of soil data, i.e. the semi-variogram and Kriging, can now be included as on-line capabilities of SIS to equip users with a powerful tool for prediction of the missing information. When there are no records of the wanted property, its values can be estimated by a function on other recorded properties acting as its surrogates. A strategy for model development is provided for the construction of surrogate functions based on multiple regression and curve fitting techniques, to generate the information missing. When no records of the soil property at the depth wanted exist, the values at the required depths are interpolated by a function of the property on the soil depth. Equal-area spline curves reconstruct, piecewise, the depth function quite closely, enabling their use for interpolation of values and depths in a variety of formats. The equal-area spline algorithm is a capability of the Oxford SIS (OXSIS). When the wanted site was unvisited during survey, the information missing is provided by spatial prediction. The predictions may come from means of soil classes or mapping units from conventional survey, or from Kriging interpolation based on spatial analysis by the semi-variogram. In order to select the best predictive tool, the success of these techniques in different situations of sampling effort and variablity were compared. Semi-variograms depicted the spatial structures of 8 selected soil properties. Anisotropic variation in 4 of them was induced by strong trends. Where the semi-variogram was isotropic Kriging was the best tool for prediction if spatial dependence is strong. Fitting elliptical functions to find a model for anisotropy did not give satisfactory results. Where anisotropies or trends precluded ordinary Kriging, map unit means and class means, in that order, gave the best predictions. Success in prediction was related to the structures in the semi-variogram, which when used for reconnaissance helps to infer which technique will give the best predictions so that survey is designed accordingly. Accounting for the trends removed anisotropies and Kriging of de-trended data was possible. Partitioning trends by stratification based in soil mapping units gave a greater improvement in predictions than modelling trends by bicubic spline surfaces and then Kriging the residuals from trend. After trend removal, Kriging did not always make the best predictions and means from classes seemed equally as good as Kriging or even better in some cases. These results indicate that a critical point is to ascertain how to best sample to estimate a reconnaissance semi-variogram for survey design to provide the information missing necessary for land-use planning.

published_or_final_version / Business Administration / Master / Master of Business Administration

published_or_final_version / Community Medicine / Master / Master of Public Health

No description available.

No description available.

No description available.

No description available.