Predictions of soil surface and topsoil organic carbon content through the use of laboratory and field spectroscopy in the Albany Thicket Biome of Eastern Cape Province of South Africa

• Published in: Geoderma 2011-11, Vol.167, p.295-302
• Main Authors: Nocita, Marco, Kooistra, Lammert, Bachmann, Martin, Müller, Andreas, Powell, Mike, Weel, Silvia
• Format: Article
• Language: English
• Subjects: agricultural soils; Agronomy. Soil science and plant productions; Albany thicket biome; Biological and medical sciences; Earth sciences; Earth, ocean, space; EnMAP; Exact sciences and technology; Fundamental and applied biological sciences. Psychology; Geochemistry; in-situ characterization; infrared reflectance spectroscopy; least-squares regression; meta analysis; nir spectroscopy; nitrogen; Organic carbon; river floodplains; sequestration; Soil and rock geochemistry; Soil spectroscopy; Soil surface; Soils; spectrometry; Surficial geology; Topsoil
• ISSN: 0016-7061; 1872-6259
• DOI: 10.1016/j.geoderma.2011.09.018

In recent years it has been shown that laboratory and field visible near infrared spectroscopy (VNIRS) allows for the accurate prediction of soil organic carbon (SOC) — more rapidly, less expensively, and at larger scales than conventional soil laboratory methods. VNIRS might find application in the restoration assessment of the degraded, semi-arid subtropical thickets of the Albany Thicket Biome (ATB) of the Eastern Cape Province of South Africa. During the twentieth century, the semi-arid forms of the ATB suffered heavy browsing by goats, transforming the dense closed-canopy shrubland into an open savannah-like system. This paper presents a study dealing with SOC estimation of soil surface (0–5 mm) and topsoil (0–200 mm) in the degraded ATB, through the combination of soil spectroscopy and partial least square regression (PLSR). Spectroscopic measurements and soil samples were collected along a transect in the ATB. The PLSR models developed with laboratory and field spectra gave good predictions of SOC, with root mean square error of validation (RMSEV) < 5.0 and 5.5 g C kg − 1 , respectively. The use of the full visible near-infrared spectral range gave better SOC predictions than using either visible or near-infrared separately. The resampling simulation of the field surface spectra to the 232 channels of the satellite-born EnMAP sensor gave good SOC predictions for laboratory conditions (RPD > 2), but low accuracy (RMSE: 9.88 g C kg − 1 ) for field model. The results of this research study indicated that, for the ATB, (i) combining soil spectroscopy and PLSR does favor accurate prediction of SOC, (ii) the predictions of surface SOC can be used as a proxy of topsoil SOC, and (iii) there is potential for future application of satellite-born hyperspectral data for SOC content predictions. ► PLSR models developed with laboratory and field spectra accurately predicted SOC ► Vis-NIR spectral range predicted SOC better than both Vis and NIR ► The predictions of surface SOC can be used as a proxy of topsoil SOC in natural lands ► Resampling of laboratory spectra to the EnMAP channels gave good SOC predictions ► There is potential for future application of hyperspectral data for SOC predictions