Premium
Free Iron Oxide Determination in Mediterranean Soils using Diffuse Reflectance Spectroscopy
Author(s) -
Richter N.,
Jarmer T.,
Chabrillat S.,
Oyonarte C.,
Hostert P.,
Kaufmann H.
Publication year - 2009
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/sssaj2008.0025
Subject(s) - silt , soil texture , texture (cosmology) , soil water , mineralogy , soil science , diffuse reflectance infrared fourier transform , absorption (acoustics) , analytical chemistry (journal) , soil test , materials science , chemistry , geology , environmental chemistry , geomorphology , biochemistry , composite material , photocatalysis , artificial intelligence , computer science , image (mathematics) , catalysis
Soil Fe oxides occur in almost all soils and reflect different environmental conditions by the high variability of their mineralogy and concentration. Quantitatively determining this important pedogenic indicator enables diffuse reflectance spectroscopy (DRS) based on material‐specific absorption characteristics. This paper presents a methodology that directly links free Fe oxide content (Fe d , citrate‐dithionite extractable Fe) with the diagnostic Fe absorption band near 900 nm (Fe‐NIR). In addition, we investigated the influence of soil texture on the spectral characteristics and prediction accuracy. We showed that the Fe absorption bands of clay‐dominated soil samples were, in general, deeper than sand‐dominated samples with comparable Fe d content. Based on the Fe‐NIR absorption depth, we created two texture‐dependent Fe d prediction models, retrieving the best Fe d estimates for the sand calibrated model ( R 2 v = 0.87, rel. MSE v = 13.9%). Due to the high texture variability in sand, silt, and clay fractions of the clay–silt dominated samples, the clay–silt calibrated model produced good predictions ( R 2 v = 0.70, rel. RMSE v = 19.0%). The soil texture appeared to have no significant influence on model stability but did affect the prediction accuracy. Constant Fe d contents were over‐ and underestimated when applying the texture‐dependent models to other texture groups. The texture‐independent model was stable and performed well ( R 2 v = 0.76, rel. RMSE v = 18.1%). These results are highly relevant to the subsequent spatial assessment of free Fe oxide content as an indicator for soil development from hyperspectral remote sensing data.