The spectral response characteristics of the soils and their possible estimation by using partial least square regression (PLSR) analysis

The objective of this study was to investigate the potential of diffuse reflectance spectroscopy technique to predict several important soil parameters such as soil moisture (SM), hygroscopic water (HW), soil carbonates (SC) and specific surface area (SSA) through descending layers of the soil stratigraphy. For these purposes a field spectrometer (Analytical Spectral Device, ASD) with a probe (sub-surface spectral head device - 3S-HeD) for reading subsoil reflectance data were used. 145 samples, taken from different Israeli soils at different depth and a spectral library containing laboratory and field in-situ collected spectra (by using 3S-HeD device) were used to generate calibration and validation models in the study based on the Near Infrared Analysis (NIRs). These models were generated by the correlation between spectral characteristics and chemical soil properties separately for each soil property, using partial least square regression (PLSR) analysis.The regression coefficients between measured and predicted values of the soil properties varied between 0.93 and 0.99 in the calibration and validation stages, RMSE (Root Mean Square Error) between 0.09 and 4.6, which showed that NIRS method had potential to accurately predict these attributes in soils. Even though the prediction results for SSA are relatively poor (RMSE varied between 27 and 30) accurate linear relationships between the laboratory measured and predicted values exist (it is seen by high correlation r values in the calibration and validation stages). In addition, to predict studied properties down the profile by using 3S -HD device, eighteen drill holes locations were selected from the surface IS-based map within the semi-arid agricultural area, using clustering isodata methods. For each drill holes ten layers were spectrally measured down to 100 cm depth. The results we got from the spectroscopic analysis were logical given the environmental conditions. Further research is needed to compare our results with chemical lab tests of the same soil samples, and to repeat this study in other locations. This new approach is worth investing in as it is cost effective for such a variety of applications currently required.