Processing of soil survey data.

This thesis focuses on processing soil survey data into user-specific information. Within this process four steps are distinguished: collection, storage, analysis and presentation. A review of each step is given, and detailed research on important aspects of the steps are presented.Observation density, type of soil attributes and selection of observation sites are important aspects in the collection of soil data. The effect of observation density on the accuracy of spatial predictions was investigated in an acid sulphate soil area in Indonesia. It was found that a similar accuracy could be obtained with a marked reduction in observation density.Most attributes collected in soil survey are on an ordinal measurement scale. Commonly used statistics, such as mean, standard deviation and semivariance, and most spatial interpolation techniques are not permissible for this type of data. Ordinal data from a soil survey in Costa Rica are used to illustrate processing possibilities. For instance, the spatial-difference-probability function was proposed for describing the spatial structure of ordinal data.Over the past twenty years the storage of soil survey data in information systems has been receiving much attention. Digital storage is essential for rapid analysis of data. The soil information system of The Netherlands is described.Seven main categories of soil data analysis can be distinguished. Examples of some categories are presented. The differences between interpreted soil maps on scales of 1 : 10 000, 1 : 25 000 and 1 : 50 000 for predicting moisture deficits and changes in crop yield were investigated. No differences in quality were found between the three maps when predicting average values for an area. The best predictions for point locations, however, were obtained with the 1 : 10 000 map.Also a comparison was made between a thematic map produced by spatial prediction from point data (kriging) and one derived from a general-purpose soil map. The thematic map contain attributes that are important for water movement in the soil. No significant difference in purity was found between the two maps. When combining soil data with other spatial data a vector to raster conversion of the soil map is often necessary. Several sheets of the soil map of The Netherlands 1 : 50 000 of different complexity were investigated for the magnitude of the rasterizing error. The regression equations determined related map complexity to rasterizing error. The rasterizing error of a complex map may be as high as 20% for a raster cell size of 4 mm * 4 mm.Two display methods are introduced for the presentation of uncertainty in soil data. The first method yields an isoline map with empirical confidence limits based on the use of kriging and associated estimated kriging variance. The second method yields a map showing the probability that a certain threshold value is exceeded. When presenting soil data in the form of a map, the complexity of the map pattern has an important influence on its readability. Six complexity measures for maps were compared. The fragmentation index was selected as the best measure for evaluating map complexity