Deriving and validating point spectrotransfer functions in VIS-NIR-SWIR range to estimate soil water retention

Parametric soil water retention functions are important inputs for soil water conservation models. The objective of this study was to derive some spectrotransfer functions (STFs) to estimate soil water retention functions, using spectral reflectance. Consequently, proximal spectral reflectance of the soil samples within the VIS-NIR and SWIR (400-2500 nm) ranges was measured by using a handheld spectroradiometer apparatus. After (pre)processing the spectral data, the correlation between all wavelengths and soil water contents at matric potentials of -15000, -10000, -5000, -3000, -1000 and -330 cm were determined. Then, the transfer functions in the form of partial least square regression (PLSR) and bootstrap aggregation with PLSR (Bagging-PLSR) were derived. These new transfer functions were called point spectrotransfer functions (PSTFs). Results indicated that the derived PSTFs can well estimate soil water retention characteristics. The derived PLSR-based STFs provided more accuracy (RMSE=0.012 to 0.022 cm3 cm-3) than Bagging-PLSR based STFs (RMSE=0.012 to 0.029 cm3 cm-3) for estimating soil water contents at all designated matric potentials. On the other hands, performance of PSTSs for soil water contents at -15000, -10000 and -5000 cm were more reasonable than that of water contents at -3000, -1000 and -330 cm. It may be then concluded that PSTFs can be used as a reasonable indirect method for estimating soil water retention particularly at high soil water pressure heads.