Combining X‐ray Computed Tomography and Visible Near‐Infrared Spectroscopy for Prediction of Soil Structural Properties

Abstract

Soil structure is a key soil property affecting a soil's flow and transport behavior. X-ray computed tomography (CT) is increasingly used to quantify soil structure. However, the availability, cost, time, and skills required for processing are still limiting the number of soils studied. Visible near-infrared (vis-NIR) spectroscopy is a rapid analytical technique used successfully to predict various soil properties. In this study, the potential of using vis-NIR spectroscopy to predict X-ray CT derived soil structural properties was investigated. In this study, 127 soil samples from six agricultural fields within Denmark with a wide range of textural properties and organic C (OC) contents were studied. Macroporosity (>1.2 mm in diameter) and CT matrix (the density of the field-moist soil matrix devoid of large macropores and stones) were determined from X-ray CT scans of undisturbed soil cores (19 by 20 cm). Both macroporosity and CT matix are soil structural properties that affect the degree of preferential transport. Bulk soils from the 127 sampling locations were scanned with a vis-NIR spectrometer (400-2500 nm). Macroporosity and CT matrix were statistically predicted with partial least squares regression (PLSR) using the vis-NIR data (vis-NIR-PLSR) and multiple linear regression (MLR) based on soil texture and OC. The statistical prediction of macroporosity was poor, with both vis-NIR-PLSR and MLR (R 2 < 0.45, ratio of performance to deviation [RPD] < 1.4, and ratio of performance to interquartile distance [RPIQ] < 1.8). The CT matrix was predicted better (R 2 > 0.65, RPD > 1.5, and RPIQ > 2.0) combining the methods. The results illustrate the potential applicability of vis-NIR spectroscopy for rapid assessment/prediction of CT matrix .Abbreviations: CT, computed tomography; CT matrix , density of field-moist soil matrix devoid of large macropores and stones; HU, Hounsfield units; MLR, multiple linear regression; OC, organic carbon; PCA, principal component analysis; PLSR, partial least squares regression; RPD, ratio of performance to deviation; RPIQ, ratio of performance to interquartile distance; vis-NIR, visible near-infrared.

Combining Visible−Near‐Infrared and Pedotransfer Functions for Parameterization of Tile Drain Flow Simulations

“…The required preparation for this method is simple, involving only air‐drying of the soil and sieving down to 2 mm. Visible near‐infrared spectroscopy has been successfully used to predict soil properties such as the complete soil texture distribution (Hermansen et al, 2017), soil structure (Katuwal et al, 2018), and soil specific surface area (Ben‐Dor et al, 2008; Knadel et al, 2018). Furthermore, studies have shown its ability to predict the wet part of the soil‐water retention curve (Babaeian et al, 2015; Pittaki‐Chrysodonta et al, 2018; Santra et al, 2009).…”

Saturated water conductivity estimation based on X-ray CT images – evaluation of the impact of thresholding errors

“…CT may also be used for imaging the whole soil core samples where a voxel size between 20 µm and 50 µm is achieved (Jarvis et al, 2017;Rab et al, 2014;Vaz et al, 2011). A resolution of this order allows for macropore observation (Jarvis, 2007;Katuwal et al, 2018;Müller et al, 2018) and root visualization (Daly et al, 2018;Sander et al, 2008). However, it can also be used for imaging soil aggregates with a voxel size of 1 µm or smaller (Józefaciuk et al, 2015;Voltolini et al, 2017b).…”

“…Another approach involves locally-adaptive segmentation methods Katuwal et al, 2018;Martín-Sotoca et al, 2018;Porter and Wildenschild, 2010). Some of these have been designed especially for binarising the soil media Martín-Sotoca et al, 2018) which utilize spatial information beside the gray level value to assign each voxel to a pore-space or soil matrix.…”

Comparing Visible–Near‐Infrared Spectroscopy and a Pedotransfer Function for Predicting the Dry Region of the Soil‐Water Retention Curve

“…Visible‐NIRS has been applied to predict different soil properties including basic soil properties such as soil organic C, clay, and water content (Stenberg et al, 2010). However, more recently the capability of the method to successfully predict other soil properties was also tested in studies on particle size distribution (Hermansen et al, 2017), soil structure (Katuwal et al, 2017), soil binding capacities (Paradelo et al, 2016), and water repellency (Knadel et al, 2016). Furthermore, vis–NIRS models have been developed to predict the wet end of the soil water retention curve (Santra et al, 2009; Babaeian et al, 2015; Pittaki‐Chrysodonta et al, 2018).…”