DBPapers
DOI: 10.5593/SGEM2015/B32/S13.057

USING VNIR-DRS AND MAGNETIC SUSCEPTIBILITY TO PREDICT SOIL ORGANIC CARBON CONTENT

A. Klement, A. Kapicka, O. Jaksik, R. Kodesova, H. Grison
Thursday 24 September 2015 by Libadmin2015

References: 15th International Multidisciplinary Scientific GeoConference SGEM 2015, www.sgem.org, SGEM2015 Conference Proceedings, ISBN 978-619-7105-37-7 / ISSN 1314-2704, June 18-24, 2015, Book3 Vol. 2, 425-432 pp

ABSTRACT
In this study we attempt to apply two techniques (reflective spectroscopy in the visible and near infrared part of electromagnetic spectrum, VNIR DRS, and mass-specific magnetic susceptibility, MS) to estimate oxidable organic carbon content (Cox). Tested soil type was Haplic Chernozem. Samples were taken from the entire soil profile (depths of 0 to130 cm), which consisted of two layers: organic matter horizon and soil substrate (loess). In total, 59 samples were collected. The soil spectra curves (of air dry soil) were measured in the laboratory using spectrometer FieldSpec®3 (400 – 2 500 nm). Partial least squares regression (PLSR) and support vector machines (SVM) were used for modelling of the relationships between soil spectra and measured oxidable organic carbon content. The magnetic susceptibility measurements were carried out using Bartington MS2B for low and high frequency. Simple regression was carried out to relate MS and Cox values.
The results showed that the best predictions of the oxidable organic carbon contents were obtained using VNIR DRS and PLSR technique (R2 = 0.92, RMSE = 0.08). Predictions using SVM technique were less accurate (R2 = 0.78, RMSE = 0.16). In the case of the magnetic analysis, the strongest linear relationship was found between soil organic carbon and mass specific magnetic susceptibility at high frequency (R2 = 0.91, RMSE = 0.12) in comparison to that measure at low frequency (R2 = 0.89,
RMSE = 0.13). In general, the results confirmed that the measurement of soil spectral characteristics and magnetic susceptibility measurements are promising novel techniques for digital soil mapping and could be utilized for prediction of soil degradation due to erosion processes.

Keywords: Visible and near infrared diffuse reflectance spectroscopy, magnetic susceptibility, soil organic carbon, partial least squares regression, support vector machine

Home | Contact | Site Map | Site statistics | Visitors : 145 / 353063

Follow site activity en  Follow site activity SOILS  Follow site activity Papers SGEM2015   ?

CrossRef Member    Indexed in ISI Web Of Knowledge   Indexed in ISI Web Of Knowledge
   

© Copyright 2001 International Multidisciplinary Scientific GeoConference & EXPO SGEM. All Rights Reserved.

Creative Commons License