Modeling metal uptake by selected vegetables from urban soils in Europe: uncovering key soil factors using partial least squares regression (PLS-R)

Lundgren, Maria; Hough, Rupert Lloyd; Beesley, Luke; Troldborg, Mads; Trakal, Lukáš; Moreno-Jiménez, Eduardo; Dai, Deliang; Augustsson, Anna

Abstract

Partial Least Squares Regression (PLS-R) was introduced as a method for modeling the uptake of six potentially toxic elements (PTEs)- Ba, Cd, Cu, Ni, Pb, and Zn- by lettuce, chard, and carrot. Data were obtained from a pot experiment where these crops were cultivated in urban soils of various characteristics. The models consider soil concentrations of PTE, Al, Ca, Fe, K, Mg, Mn, Na, P, S and pH, SOM, CEC, and soil texture as predictors. Initially, eighteen metal- and crop-specific models with all predictors were developed, using selectivity ratios (SRi) to identify influential variables for predicting PTE soil-to-crop transfer. Reduced models were then created using only predictors with high SRi. Key variables for predicting PTE soil-to-crop transfer included soil PTE concentration, pH, Fe and Mn soil concentrations, and soil texture. Out of eighteen models, sixteen were suitable for predicting correlations and assessing PTE accumulation in crops, while eight were accurate for quantitative predictions. This study shows that PLS-R is a robust method for modeling soil-to-crop transfer of metal contaminants, even with multicollinear predictors. PLS-R also helps identify key variables, providing insights into the mechanisms of PTE accumulation in crops, which is crucial for effective risk assessments.

• Journal Article Type: Article
• Acceptance Date: Jan 26, 2025
• Online Publication Date: Feb 17, 2025
• Deposit Date: Mar 20, 2025
• Publicly Available Date: Mar 20, 2025
• Journal: Human and Ecological Risk Assessment
• Print ISSN: 1080-7039
• Electronic ISSN: 1549-7860
• Publisher: Taylor and Francis
• Peer Reviewed: Peer Reviewed
• DOI: https://doi.org/10.1080/10807039.2025.2464109