A geochemical survey of the concentrations of about 50 elements in surface soils from Wisconsin, USA is currently being conducted as a state-funded ancillary effort of the USGS Geochemical Landscapes project. This data will be used to produce databases and GIS-based maps of elemental distributions for eventual public use. Analytical methods employed in the Wisconsin survey are NAA (neutron activation analysis), XRF (X-ray fluorescence), ICP-MS (Inductively Coupled Plasma-Mass Spectrometry) and ICP-OES (Inductively Coupled Plasma-Optical Emission Spectroscopy). When undertaking a survey of this magnitude, questions arise as to the choice of sampling sites being truly representative of a larger landscape and how sampling designs can assist to alleviate this uncertainty. Most of Wisconsin has been glaciated repeatedly and is covered with tills that can be assigned to particular glacial intervals and movements. Till geochemistry may be more easily determined due to the homogenization of the materials during the glacier movements. However, the southwestern quadrant of the state [known as the Driftless Area (DA)] has escaped glaciation and has bedrock dominated topography underlying varying thicknesses of loess. A preliminary evaluation of statewide soils shows that while the concentrations of elements in the total soils show significant spatial variation, the concentrations of elements in the clay-size fraction (<2 micrometers) are relatively uniform, especially in soils from the glaciated areas. The variation in element concentrations in the total soils appears to be directly related to variations in quartz contents in the soils. The soils in the DA are much older than soils from recent (12 kyr old) glacially-derived till, outwash and lacustrine sediments, having formed from both loess deposits (which have been deposited and eroded most pervasively over the past 120 kyr, correspondent to glacial advances) and mixed dolostone/sandstone bedrock. The DA soils in Wisconsin occupy a two-story landscape with productive ridge and valley soils separated by steep and wooded shallow soil sideslopes and rockland. To investigate the variability caused by landscape variables, surface soils in a small watershed (about 15 km2) in the DA were intensively sampled and analyzed to determine the relationships between element concentrations in soils with soil series and position in the landscape. The common bedrock for this watershed is dominantly dolostone and sandstone, often disaggregated by periglacial effects of the last glacial maximum. Relationships between geochemical-element concentrations and geographical-soil characterization data, such as altitude and terrain curvature (profile and planform), were established. The associations between the soil properties used to generate soil mapping units and the element composition of soil were tested to determine the confidence of the geochemical uniformity within a soil series. Differences were evident in some parts of the watershed where loess-derived soils are thin over the residual (and much older) sub-soils, which have a much higher content of redox-sensitive and phyllosilicate-bound elements. These results can potentially aid the classification of soils as well as helping to fine-tune precision management plans for landscapes for agriculture or urban development. Reliable patterns of element compositions can serve as a useful tool to study soil formation processes and potential problems that might arise through disturbances. For example, hydrochemical transport processes increase the content of clay-size particles in the B horizon and tend to enhance the loss of cations by leaching. The total variance of major and trace element concentrations can clearly be viewed as related to spatial components within the watershed. The spatial variance patterns with respect to each element were achieved by statistical methods, such as ANOVA (analysis of variance) and PCA (principle component analysis).First, a discrete version of this cumulative variance as a function of increasing distance can be obtained by balanced ANOVA based on a hierarchical nested sampling design with different distances as subclasses or levels. Samples are grouped into several levels according to the distance between sample groups in the watershed. In addition, samples in other counties of Wisconsin are considered as the highest level to check the regional (or glaciation) effect on variance. Secondly, element concentrations were classified into several groups according to their variance pattern. For example, elements subject to variations in concentration due to waste disposal or localized mineral deposits have a large variance on a small scale and behave differently from major elements, which can be described mineralogically. Both ANOVA and PCA can be used for grouping purpose. With the support of geostatistics, these results can be expanded to the soils originating from the same geological processes, which are the southwest corner of Wisconsin and conterminous non-glaciated area in Minnesota.