Digital soil mapping and portable X-ray fluorescence prediction of cadmium, copper and zinc concentrations as decision support for crop production

Trace element concentrations in agricultural soil are important for crop production. Certain trace elements, e.g. copper (Cu) and zinc (Zn), are essential for crops to complete their life cycle. Other trace elements, e.g. cadmium (Cd), can be harmful to crops or the end-consumer. Hence, it is important to have maps of soil concentrations of trace elements or methods for determining concentrations in soil samples. This thesis investigated the possibility of predicting trace element concentrations (Zn, Cu, Cd) in soil samples using portable X-ray fluorescence (PXRF) measurements. It also examined usefulness of digital soil mapping (DSM) to create maps of Cu and Cd concentrations in agricultural topsoil in Sweden. Portable X-ray fluorescence models were validated at national and farm level. Predicted Zn concentrations were found to be comparable to those obtained in conventional laboratory analysis, while predicted Cd and Cu concentrations were less accurate. The most accurate PXRF models were created using non-linear machine learning algorithms, e.g. random forest. Digital soil mapping of Cd concentrations in Skåne County, combined with data from grain sampling, revealed that low Cd concentrations in winter wheat grain were associated with predicted low concentrations in soil. The map could thus be used to identify arable soils suitable for producing winter wheat for products with strict quality criteria, e.g. baby food. Digital soil mapping of Cu concentrations at national level revealed that 47% of arable soils are highly likely not at risk of Cu deficiency. Covariate importance analysis indicated importance of airborne gamma radiation measurement data in DSM of Cu and Cd concentrations