Premium
Quantification of Different Forms of Iron from Intact Soil Cores of Paddy Fields with Vis‐NIR Spectroscopy
Author(s) -
Xu Shengxiang,
Zhao Yongcun,
Wang Meiyan,
Shi Xuezheng
Publication year - 2018
Publication title -
soil science society of america journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.836
H-Index - 168
eISSN - 1435-0661
pISSN - 0361-5995
DOI - 10.2136/sssaj2018.01.0014
Subject(s) - partial least squares regression , soil water , chemistry , soil test , spectroscopy , soil science , environmental chemistry , analytical chemistry (journal) , mineralogy , environmental science , mathematics , statistics , physics , quantum mechanics
Core Ideas The spectra of intact soil cores can be used to predict various soil Fe forms. SVMR shows superiority to PLSR for determining soil Fe contents. The study provides an alternative for determination of soil Fe forms in paddy fields. Iron (Fe) occurs in almost all soils and the analysis of various forms of Fe in the soil is of great pedological interest. Very little is known, however, about how visible and near‐infrared (Vis‐NIR) spectroscopy performs in intact soil cores of paddy fields for quantifying Fe concentrations. Our objective was to evaluate the feasibility of Vis‐NIR spectroscopy of intact soil cores for rapid determination of the four Fe forms: total Fe (Fe t ), pyrophosphate‐extractable Fe (Fe p ), dithionite‐citrate‐bicarbonate extractable Fe (Fe d ), and oxalate‐extractable Fe (Fe o ). A total of 148 intact soil cores in Yujiang County, China, were sampled, and Vis‐NIR spectra (350–2500 nm) were sectioned and scanned on four horizontal surfaces (5‐cm depth intervals) of each soil core in the laboratory. Partial least squares regression (PLSR) and support vector machine regression (SVMR) models were compared using 70% of the section samples for calibration and 30% for independent validation. Results showed that the nonlinear SVMR models performed better than the PLSR models for the predictions of all Fe forms. The SVMR models produced the best predictions in the independent validation set for Fe d (RMSE P = 2.223; R 2 P = 0.88; RPD P = 2.86), Fe o (RMSE P = 0.994; R 2 P = 0.85; RPD P = 2.59), Fe t (RMSE P = 3.693; R 2 P = 0.82; RPD P = 2.32), and Fe p (RMSE P = 0.086; R 2 P = 0.79; RPD P = 2.17). It was concluded that Vis‐NIR spectroscopy coupled with SVMR is suitable for quantitatively determining different Fe forms in intact soil cores of paddy fields.