Magnetic signature and X-ray fluorescence for mapping trace elements in soils originating from basalt and sandstone

Abstract

The knowledge of the lithological context is necessary to interpret trace elements concentrations in the soil. Soil magnetic signature (χ) and soil X-ray fluorescence (XRF) are promising approaches in the study of the spatial variability of trace elements and the environmental monitoring of soil quality. This research aimed to assess the efficiency of measurements of χ and XRF sensors for spatial characterization of zinc (Zn), manganese (Mn), and copper (Cu) contents in soils of a sandstone-basalt transitional environment, using machine learning modeling. The studied area consisted of the Western Plateau of São Paulo (WPSP), with soils originating from sandstone and basalt. A total of 253 soil samples were collected at a depth of 0.0–0.2 m. The soils were characterized by particle size and chemical analysis: organic matter (OM), cation exchange capacity (CEC), ammonium oxalate-extracted iron (Feo), sodium dithionite-citrate-bicarbonate-extracted iron (Fed), and sulfuric acid-extracted iron (Fet). Hematite (Hm), goethite (Gt), kaolinite (Kt), and gibbsite (Gb) contents were obtained by X-ray diffraction (XRD). Magnetite (Mt) and maghemite (Mh) contents were obtained by soil χ, while trace elements contents were obtained by XRF and predicted by χ. Descriptive analysis, the test of means, and correlation were performed between attributes. Zn, Mn, and Cu contents were predicted using the machine learning algorithm random forest, and the spatial variability was obtained using the ordinary kriging interpolation technique. Landscape dissections influenced iron oxides, which had the highest contents in slightly dissected environments. Trace elements contents were not influenced by landscape dissections, demonstrating that lithological knowledge is necessary to characterize trace elements in soils. The prediction models developed through the machine learning algorithm random forest showed that χ can be used to characterize trace elements. The similar spatial pattern of trace elements obtained by XRF and χ measurements confirm the applicability of these sensors for mapping it under lithological and landscape transition, aiming for sustainable strategic planning of land use and occupation.