Soil Specific Surface Area Determination by Visible Near‐Infrared Spectroscopy. Issue 5 (20th September 2018)

• Record Type: Journal Article
• Authors: Knadel, Maria; Arthur, Emmanuel; Weber, Peter; Moldrup, Per; Greve, Mogens Humlekrog; Chrysodonta, Zampella Pittaki; de Jonge, Lis W.
• Is Part Of: Soil Science Society of America Journal. Volume 82:Issue 5(2018)
• Page Start: 1046
• Page End: 1056
• Publication Date: 2018-09-20
• Subjects: Soils -- United States -- Periodicals; Soil science -- Periodicals; Periodicals; 631.4973
• DOI: 10.2136/sssaj2018.03.0093
• Languages: English
• ISSNs: 0361-5995

Abstract

Abstract : Core Ideas: Feasibility of using vis–NIRS for SSA determination in a highly variable data set was investigated Acceptable predictive ability for SSA using PLS and iPLS was obtained. iPLS resulted in little improved accuracy for both SSA and clay iPLS indicated a higher specificity of SSA than of clay response The important wavelengths reflected the organo‐mineral composition and its interactions The soil specific surface area (SSA) affects soil physical and chemical properties. Numerous studies applied visible near‐infrared spectroscopy (Vis–NIRS) to estimate clay content (particles < 2 µm). Since SSA is better defined and more directly related to particle size distribution and mineralogy than clay content, predictions of SSA from Vis–NIRS are expected to be better than that for clay. Thus, the aims of this study were to (i) test the feasibility of using Vis–NIRS for SSA determination, (ii) compare the predictive ability of Partial Least Squares (PLS) model of SSA with that of clay, (iii) identify important wavelengths using interval Partial Least Squares (iPLS) regression, and to test if the application of iPLS improves the predictive ability of the models. A total of 550 soil samples with a wide range in SSA (3–437 m 2 g –1 ) and clay content (1–83%) was divided into a calibration and a validation set. The PLS models had similar predictive ability for SSA (ratio of performance to interquartile range, RPIQ = 1.7) and clay content (RPIQ = 1.6). Utilizing iPLS led to only limited improvement in the prediction accuracy (RPIQ of 1.8 and 1.7 for SSA and clay content, respectively), yet decreased the number of relevant wavelengths and indicated a higher specificity of SSA over the broader spectral response of clay. The important wavelengths for SSA and clay predictions were indicative of the organo‐mineral content and its interactions, including spectral response from not only iron oxides and minerals but also organic matter due to masking effect of the non‐complexed organic carbon on the mineral phases of some of the soils.