A high-density regional-scale soil geochemical survey comprising 727 samples (one sample per each 5 × 5 km grid) was carried out in the Parauapebas sub-basin of the Brazilian Amazonia, under the Itacaiúnas Basin Geochemical Mapping and Background Project. Samples were taken from two depths at each site: surface soil, 0-20 cm and deep soil, 30-50 cm. The ground and sieved (< 75 µm) fraction was digested using aqua regia and analyzed for 51 elements by inductively coupled plasma mass spectrometry (ICPMS). All data were used here, but the principal focus was on the potential toxic elements (PTEs) and Fe and Mn to evaluate the spatial distribution patterns and to establish their geochemical background concentrations in soils. Geochemical maps as well as principal component analysis (PCA) show that the distribution patterns of the elements are very similar between surface and deep soils. The PCA, applied on clr-transformed data, identified four major associations: Fe-Ti-V-Sc-Cu-Cr-Ni (Gp-1); Zr-Hf-U-Nb-Th-Al-P-Mo-Ga (Gp-2); K-Na-Ca-Mg-Ba-Rb-Sr (Gp-3); and La-Ce-Co-Mn-Y-Zn-Cd (Gp-4). Moreover, the distribution patterns of elements varied significantly among the three major geological domains. The whole data indicate a strong imprint of local geological setting in the geochemical associations and point to a dominant geogenic origin for the analyzed elements. Copper and Fe in Gp-1 were enriched in the Carajás basin and are associated with metavolcanic rocks and banded-iron formations, respectively. However, the spatial distribution of Cu is also highly influenced by two hydrothermal mineralized copper belts. Ni-Cr in Gp-1 are highly correlated and spatially associated with mafic and ultramafic units. The Gp-2 is partially composed of high field strength elements (Zr, Hf, Nb, U, Th) that could be linked to occurrences of A-type Neoarchean granites. The Gp-3 elements are mobile elements which are commonly found in feldspars and other rock-forming minerals being liberated by chemical weathering. The background threshold values (BTV) were estimated separately for surface and deep soils using different methods. The ’75th percentile’, which commonly used for the estimation of the quality reference values (QRVs) following the Brazilian regulation, gave more restrictive or conservative (low) BTVs, while the ‘MMAD‘ was more realistic to define high BTVs that can better represent the so-called mineralized/normal background. Compared with CONAMA Resolution (No. 420/2009), the conservative BTVs of most of the toxic elements were below the prevention limits (PV), except Cu, but when the high BTVs are considered, Cu, Co, Cr and Ni exceeded the PV limits. The degree of contamination (Cdeg), based on the conservative BTVs, indicates low contamination, except in the Carajás basin, which shows many anomalies and had high contamination mainly from Cu, Cr and Ni, but this is similar between surface and deep soils indicating that the observed high anomalies are strictly related to geogenic control. This is supported when the Cdeg is calculated using the high BTVs, which indicates low contamination. This suggests that the use of only conservative BTVs for the entire region might overestimate the significance of anthropogenic contamination; thus, we suggest the use of high BTVs for effective assessment of soil contamination in this region. The methodology and results of this study may help developing strategies for geochemical mapping in other Carajás soils or in other Amazonian soils with similar characteristics.