The conventional methods for measuring soil properties are time-consuming and expensive. Over the past few decades, reflectance spectroscopy has been widely applied as a rapid and cost-effective technique to estimate soil properties. However, less attention has been given to estimating soil engineering properties using soil spectral data. To evaluate the efficiency of the technique, selected physical and mechanical properties of 220 soil samples from two study areas in Ghorveh-Dehgolan and Zrêbar regions in the Kurdistan province, west of Iran, were measured by standard laboratory methods. Spectral measurements were also performed on air-dried soil sub-samples using a laboratory spectroradiometer apparatus. After recording the spectra, the different pre-processing methods were tested to improve the estimations. Partial least squares regression (PLSR) was used to establish relationships between soil engineering properties and the spectra transformed via Savitzky–Golay first derivative. Based on the results, the prediction accuracy was very good for clay activity (Ac) and clay content (modeling efficiency (EF) = 0.75 and the ratio of performance to deviation (RPD) > 2), and good for cation exchange capacity (CEC) and effective diameter size (D10) (EF > 0.68 and RPD > 1.81). Reasonable predictions were also obtained for Atterberg limits, silt and sand contents, particle density (PD), diameters corresponding to 30% and 60% finer (D30 and D60, respectively), median particle size (D50), and geometric mean diameter of soil particles (dg) and their standard deviation (δg) (EF: 0.51–0.67, RPD: 1.44–1.76); however, the total porosity (n) and bulk density (BD) of soils were poorly predicted. The reliability and good quality of the estimations was probably due to the variability in the properties of the different soils collected from the two distinct geographical regions. Overall, the results confirmed the potential efficacy of Vis-NIR spectroscopy for a rapid estimation of different important soil engineering properties.
