Modelling Leverage of Different Soil Properties on Selenium Water-Solubility in Soils of Southeast Europe

Abstract

Selenium (Se) is a nonmetal that is essential for humans and other animals, and is considered beneficial for plants. The bioavailability of Se strongly influences its content in the food chain. Soils are the main source of Se, and their Se content primarily influences its availability, along with other soil properties. A field survey was conducted on soils of Southeast Europe, specifically in Croatia (Osijek), Bosnia and Herzegovina (Sarajevo, Banja Luka, Mostar, and Prud), and Serbia (Novi Sad). Soil samples were taken from the arable soil layer (0–30 cm depth), and two types of Se availability were measured: Se extracted using pure HNO3 (SeTot) and Se readily extracted in water (SeH2O). Only soils from the Mostar area had Se concentrations above deficit levels (0.5 mg kg−1), with the highest values of cation exchange capacity (CEC), soil organic matter (SOM) measured as loss of ignition (LOI), total C, total N, ZnTot and CdTot. The connections between the chemical characteristics of the soil and SeH2O were investigated. Principal component analysis (PCA) explained 73.7% of the variance in the data set in the first three principal components (PCs). Using the provided data, we developed a partial least squares (PLS) regression model that predicted the amount of SeH2O in the soil, with an accuracy ranging from 77% to 90%, depending on the input data. The highest loadings in the model were observed for LOI, CEC, total C, total N, and SeTot. Our results indicate the need for biofortification in these key agricultural areas to supplement the essential dietary requirements of humans and livestock. To efficiently and economically implement biofortification measures, we recommend utilizing regression models to accurately predict the availability of Se.