To support the development of protocols for the proposed North American Soil Geochemical Landscapes project, whose objective is to establish baselines for the geochemistry of North American soils, two continental-scale transects across the United States and Canada were sampled in 2004. The sampling employed a spatially stratified random sampling design in order to estimate the variability between 40-km linear sampling units, within them, at sample sites, and due to sample preparation and analytical chemical procedures. The 40-km scale was chosen to be consistent with the density proposed for the continental-scale project. The two transects, north–south (N–S) from northern Manitoba to the USA–Mexico border near El Paso, Texas, and east–west (E–W) from the Virginia shore north of Washington, DC, to north of San Francisco, California, closely following the 38th parallel, have been studied individually. The purpose of this study was to determine if statistically significant systematic spatial variation occurred along the transects. Data for 38 major, minor and trace elements in A- and C-horizon soils where less than 5% of the data were below the detection limit were investigated by Analysis of Variance (ANOVA). A total of 15 elements (K, Na, As, Ba, Be, Ce, La, Mn, Nb, P, Rb, Sb, Th, Tl and W) demonstrated statistically significant (p<0.05) variability at the between-40-km scale for both horizons along both transects. Only Cu failed to demonstrate significant variability at the between-40-km scale for both soil horizons along both transects.

The patterns of relative variability differ by transect and horizon. The N–S transect A-horizon soils show significant between-40-km scale variability for 29 elements, with only 4 elements (Ca, Mg, Pb and Sr) showing in excess of 50% of their variability at the within-40-km and ‘at-site’ scales. In contrast, the C-horizon data demonstrate significant between-40-km scale variability for 26 elements, with 21 having in excess of 50% of their variability at the within-40-km and ‘at-site’ scales. In 36 instances, the ‘at-site’ variability is statistically significant in terms of the sample preparation and analysis variability. It is postulated that this contrast between the A- and C- horizons along the N–S transect, that is dominated by agricultural land uses, is due to the local homogenization of Ap-horizon soils by tillage reducing the ‘at-site’ variability. The spatial variability is distributed similarly between scales for the A- and C-horizon soils of the E–W transect. For all elements, there is significant variability at the within-40-km scale. Notwithstanding this, there is significant between-40-km variability for 28 and 20 of the elements in the A- and C-horizon data, respectively. The differences between the two transects are attributed to (1) geology, the N–S transect runs generally parallel to regional strikes, whereas the E–W transect runs across regional structures and lithologies; and (2) land use, with agricultural tillage dominating along the N–S transect. The spatial analysis of the transect data indicates that continental-scale maps demonstrating statistically significant patterns of geochemical variability may be prepared for many elements from data on soil samples collected on a 40 x 40 km grid or similar sampling designs resulting in a sample density of 1 site per 1600 km².