Proximal sensing of soil reflectance to characterise field soil organic carbon

2002

Raphael A. VISCARRA ROSSEL^1,2, Christian WALTER^2,3 and Youssef Fouad³

¹ Australian Centre for Precision Agriculture, University of Sydney, NWS 2006, Australia
² INRA Laboratoire de Science du Sol, 65 rue de St Brieuc, 35042 Rennes, France
³ Ecole Nationale Supérieure Agronomique, Rennes, France

Symposium No.48, Paper No. 1523, 17th World Congress of Soil Science, 14-21 August, Bangkok, Thailand.

ABSTRACT

Various soil properties exhibit spectral response in the visible range of the electromagnetic spectrum and these properties have been shown to have good correlations with measurements of soil colour, e.g. soil organic carbon (SOC) (e.g. Ben-Dor et al., 1997; Lindbo et al., 1998). Soil colour using the Munsell colour system has long been used for soil identification and qualitative determinations of soil characteristics, however there is a need for quantitative measurements of soil colour. The main objectives of this work were to (i.) establish relationships between soil colour (using RGB, CIE and Munsell colour systems) and soil organic carbon (SOC), (ii.) determine whether quantitative measurements of soil colour could be used to predict SOC content of agricultural soil and (iii.) to compare predictions to spectrometric measurements made using a visible range spectrometer. Soil colour was determined using different systems for representing colour space: qualitatively using the Munsell Soil Colour Charts and quantitatively using RGB (red, green, blue) tristimulus values from soil images acquired using a digital camera. The RGB values were converted to CIE XYZ tristimulus and their resulting CIEL*a*b* (CIELAB) and CIEL*u*v (CIELUV) transforms. To establish relationships between SOC and soil colour, forty-three different soils were collected from various locations across Brittany, France. Visual measurements of Munsell soil colour and digital images of these soil samples were acquired in the laboratory under ‘ideal’ lighting conditions. The soil was also analysed for SOC using conventional laboratory analysis. Statistical relationships between Munsell value units, RGB image-intensities, CIEL*a*b* and CIEL*u*v coordinates and SOC were derived for predictions of SOC content of field soil. The visible spectra (400 nm – 700 nm) of these soils were also measured using a spectrometer and partial least-squares (PLS) regression (Martens & Naes, 1989) implemented on the spectra for predictions of SOC. Field soil was sampled from two different fields with different levels of SOC and transported to the laboratory for image acquisition, spectral analysis and SOC determinations. Predictions were validated against chemical analyses and statistics that relate the accuracy of predictions to their precision and bias were used to quantify their quality. Quantitative soil colour measurements using either RGB image-intensity values or CIELAB / CIELUV colour coordinates showed good response for SOC. Using an appropriate calibration model, accurate predictions of field SOC are possible. It appears that the CIELAB system may be more appropriate for predictions of SOC in Brittany soils, be it using either the L or the b coordinates, depending on the inherent characteristics of the soil and its range in SOC content.