1. Scope. Within the German joint PF research project preagro, soil surveys have been performed on more than sixty fields of pilot farms all over Germany. The Digital Farm Soil Maps (“Digitale Hof-Bodenkarten”) were the basis for Variable Rate Technology (VRT) applications of the project, especially site-specific N-fertilization. Additionally, geostructured variabilities of various soil and nutrient parameters were studied by variogram analyses on selected project fields in order to plan soil sampling for PF more rationally. 2. Sites and Methods. By support of detailed maps of soil conductivity (EC), soil rating data (SR), digital terrain models (DTM) and relative yields (Y%), and calibrated by >1600 1.5m soil auger profiles we delineated soil types (acc. to German classification) and soil texture classes precisely within the fields. These were situated in different landscapes with soils derived from young (2 farms) and old (1) glacial tills, from loess (2), sand/till/loess-covered claystones (1), limestones (2) and tertiary debris (1). From these, six fields were selected to study by a stratified, regular sampling design the variograms of soil and nutrient parameters (clay, organic matter OM in Ap; available P, K, Mg, pH in 0-30cm and 30-60cm; soil types, texture classes and AWC of profiles). 3. Results from soil surveys. Patterns of soil types and texture classes of all fields show that mean polygon sizes of soils in young morainic (and alluvial) soilscapes tend to be <1 ha, while soil boundaries in soilscapes on mesozoic clay- and limestone slopes often comprise >1ha. Due to glacial ero-sion rinns in underlying sandurs, tills or claystones, soilscapes of smoothing loess covers have more intricated patterns than expected from relief. 4. Geostatistics from variogram analyses. From >50 variograms the common statistics (nugget, sill, range, model type) were ana-lyzed and discussed elsewhere (Herbst&Lamp 2002). A simple geo-statistic MD, the median distance at 50% of the geostructered semivariance, allows to give an overview of variogram results (fig.) with consequences: (1) available nutrients tend to have geostructural variances at <30m distances, near those of pedocells (approx. nugget): they can not be sampled effi-ciently; (2) designs for permanent soil parameter (e.g. clay, AWC) in hilly morainic soilscapes (Thumby, Kassow) should plan >2 samples/ha; (3) only in soilscapes of mesozoic and loess materials a sample density of ~1/ha may be considered.