The Relation between Soil Water Repellency and Water Content Can Be Predicted by Vis‐NIR Spectroscopy

Abstract

The severity of soil water repellency (SWR) varies nonlinearly with water content ( w ), and it is extremely laborious to obtain complete SWR‐ w curves, which are needed to predict the occurrence of SWR. In this study, we combined a three‐parameter moisture‐dependent SWR (MD‐SWR) model with visible near‐infrared spectroscopy (vis‐NIRS) as a fast and novel method to estimate the SWR‐ w curve. The method was applied to a data set of SWR‐ w curves determined for 71 soil samples (organic carbon [OC] content: 0.021–0.147 kg kg ‐1 , clay: 0.000–0.520 kg kg ‐1 ). The degree of SWR was measured on air‐dried soil samples (SWR AD ) and on soil samples with increasing water contents until the water content at which the soils became wettable ( w NON ) was reached. The three‐parameter MD‐SWR model was fitted to the measured SWR‐ w curves between the water content at air‐dry conditions ( w AD ) and w NON . The total SWR was then calculated as the trapezoidal integrated area underneath the SWR‐ w curves (SWR AREA ). Air‐dried soil samples were scanned with a vis‐NIR spectrometer. Each of the three MD‐SWR model parameters was correlated to vis‐NIRS spectra using partial least squares regression. The SWR AREA was predicted using two approaches. For Approach I, the SWR AREA calculated from the MD‐SWR model was predicted with a single vis‐NIRS model. Approach II utilized predicted MD‐SWR model parameter values to obtain vis‐NIRS‐predicted SWR‐ w curves between w AD and w NON , and the SWR AREA was calculated from these curves. Results show that vis‐NIRS can predict the shape of the SWR‐ w curves as well as the SWR AREA ( R 2 = 0.58 and 0.56 for Approach I and II, respectively) across a highly variable dataset from a single vis‐NIRS scanning and one SWR measurement at air‐dried conditions.