Developing spectrotransfer functions (STFs) to predict basic physical and chemical properties of calcareous soils

Abstract

Spectroscopy in visible (Vis) and near-infrared (NIR) provides a rapid, timely, non-destructive, low-preparation, and less expensive analysis of soil samples in comparison with traditional laboratory analysis. Therefore, the objectives of this study were to predict basic physical and chemical properties of calcareous soils using Vis-NIR spectral reflectance data by applying partial least square regression (PLSR) and stepwise multiple linear regression (SMLR) approaches and finally to develop spectrotransfer functions (STFs). The target soil properties including sand, silt, clay, pH, electrical conductivity (EC), calcium carbonate equivalent (CCE), and water-soluble sodium (Na), potassium (K), calcium (Ca), and magnesium (Mg), in 234 soil samples, and DTPA-extractable iron (Fe), manganese (Mn), copper (Cu), and zinc (Zn), in 161 soil samples, were measured using standard laboratory procedures in calcareous soils of Fars province, Iran. The spectra of soil samples in Vis-NIR region (400–2500 nm) were collected using Rapid Content Analyzer spectrophotometer (NIRS XDS model) apparatus. Results revealed the better performance of developed PLSR models compared to STFs to predict most of the target soil properties; however, their capability differences were not significant. The STFs predicted sand, clay, and CCE with very good accuracy (0.78 ≤ determination coefficient of validation dataset, R2val ≤ 0.83); pH, K, Mg, Fe, Mn, Cu, and Zn with good accuracy (0.65 ≤ R2val ≤ 0.71); and silt, EC, Na, and Ca with acceptable accuracy (0.55 ≤ R2val ≤ 0.63). It is strongly recommended to use the related STFs for predicting sand, clay, CCE, pH, K, Mg, Fe, Mn, Cu, and Zn of calcareous soils in order to save time and costs, less use of chemicals, and mapping large areas.