Soil erodibility prediction by Vis-NIR spectra and environmental covariates coupled with GIS, regression and PLSR in a watershed scale, Iran

چکیده مقاله

Visible–Near-Infrared (Vis-NIR) spectroscopy as a rapid, cost-effective, and non-destructive technique has become an alternative approach to evaluate difficult-to-measured soil properties. Hence, this study aimed to evaluate RUSLE soil erodibility (so-called K-factor) in calcareous soils of a semi-humid watershed located in west of Iran using Vis-NIR spectroscopy combined with topography data in GIS. To this end, the K-factor from the RUSLE model and various soil properties such as texture, organic-matter (SOM), bulk density, pH, EC were determined in 120 soil samples taken from different land uses. Partial least-squares-regression (PLSR) and stepwise-multiple-linear-regression (SMLR) were employed to derive models via spectral data coupled with topographical factors for modeling the K-factor. Results showed that slope-farmlands with SOM = 1.6% and woodlands with SOM = 3.5% had the highest and lowest K-factor, respectively. The PLSR model with respect to R2 = 0.63, RMSE = 0.018 Mg h MJ−1 mm−1 and RPD =1.6 had an appropriate performance in determining the K-factor. The results also revealed that the R2 and RMSE of the Improved-SMLR model developed from spectral data combined with topographical factors (elevation and slope) increased and decreased by 11% and 15%, respectively compared with the SMLR model developed only from spectral information. We concluded that using Vis-NIR spectroscopy integrated with topographical factor could be an applicable and reliable method for the prediction of K-factor.