Quantitative estimation of soil texture components using spectroscopy in the visible-near infrared region

Knowledge about soil texture is very important in agricultural studies due to its direct impact on other soil properties. However, determining the soil texture in vast areas requires a lot of time and money. For this reason, researchers are looking for ways to determine this important feature of the soil on a large scale. One of these methods is the use of surface soil spectrometry. In this method, the choice of calibration method significantly affects the accuracy of measuring the characteristics of the surface. In this study, the performance of two regression techniques, namely, partial least-squares regression (PLSR), principal component regression (PCR) were compared to identify the best method to assess sand, silt and clay. For this purpose, 50 soil samples from Tehran province were collected and used as a data set for Calibration and Validation. Soil samples with different moisture levels (oven dry, 5, 10, 15 and 20 w/w) were scanned using a FieldSpec Pro Spectroradiometer with a measurement range of 350–2500 nm. The spectra were subjected to three pre-processed techniques, e.g., Savitzky–Golay (SG) smoothing, first derivative with SG smoothing (FD-SG), Normalization with SG smoothing (Normal-SG). The R2 results from cross-validation indicated that the PLSR model had a better performance than PCR. Normal + SG pre-processing method for clay loam texture and SG method for sandy clay loam texture showed better estimation of measured properties. The amount of R2 for clay was 0.74, 0.81, 0.97 and 0.87, respectively, in moisture content of oven dry, 5, 15 and 20% in   clay loam texture And 0.95 and 0.61 at oven dry and 5% levels in sandy clay loam. Silt was better predicted by R2 0.67 in moisture content of 5% in clay loam texture and R2 0.97 in moisture content of 20% at sandy clay loam texture. Sand was also predicted (R2= 0.86 and 0.72) in moisture content of 5 and 10% in clay loam texture.