Among natural factors driving the soil content in metallic trace elements (MTE), the geochemical background is often deemed as a sound base for the detection of contaminations. Numerous studies have shown the difficulties to take into account the multi-dimensionnality of the spatial variability of most soil properties. As far as a detailed soil map is available for Belgium, the relevance of its information was investigated at various spatial scales in the Walloon Region (Southern Belgium), from the plot to the entire region, to evaluate the feasibility of a multi-scale mapping of geochemical background. A database of soil profiles (gathering both agricultural and forest soils) was used to analyse the relationships between the MTE total content (measured after HF-HClO4-HCl dissolution) and other general characteristics such as pH, Total Organic Carbon content (TOC), Particle Size Distribution (PSD), and cationic exchange capacity (CEC). The factors of MTE spatial distribution inside and between solums were investigated through multivariate statistical analysis. The spatial dependence of these factors was then studied over plot, field, farm, county and region scales, and compared to the spatial structures of the soil map parameters. At the solum level, strong relationships between most of the studied MTE were found, as well as between stone charge and fine earth contents. The PSD is the main driving factor of MTE content in non-carbonated soils, whatever the land use. When considering soil associations, such as catenas, the slope redistribution processes must be taken into account. However, the PSD remains the main driving factors and, as the soil map contains information about that property (mainly the textural class and the profile development), a cartography of the geochemical background based on the use of the soil map and pedotransfer functions or regression models appears relevant. At the regional scale, some long-range spatial structures could be identified. These result mainly from the geological structure of the Walloon region, where the lithological zonation is rather clearly marked. The soil map (texture and nature of stone charge) appears relevant at that scale too but does need a generalization process and fails however to differenciate the intra-type soil spatial variability. Some basic processes of integration of point measurements and qualitative information have been explored. They should be further investigated in the future.