Proximal sensing of soil pH and lime requirement by mid infrared diffuse reflectance spectroscopy

2001

R.A. VISCARRA ROSSEL^1,2, D.J.J. WALVOORT^1,3, A.B. McBRATNEY¹, L.J. JANIK⁴, J.O. SKJEMSTAD⁴

¹Australian Centre for Precision Agriculture, McMillan Building A05, The University of Sydney, NSW 2006, Australia
²INRA Laboratoire de Science du Sol, 35042 Rennes, France
³Wageningen University & Research Centre, Wageningen, The Netherlands
⁴CSIRO Land and Water, Glen Osmond SA 5064, Australia

In G. Grenier and S. Blackmore (Eds), ECPA 2001, Third European Conference on Precision Agriculture, Vol. 1, Agro Montpellier, pp. 497-508.

ABSTRACT

We describe the use of mid-infrared (MIR) diffuse reflectance spectroscopy in combination with partial least squares regression (PLS) for the prediction of soil pH and lime requirement. Delete-one-jackknife cross-validation was used to determine the optimal number of factors for the most accurate prediction of each soil property. Predictions of soil pH in 0.01M CaCl₂ (pHCaCl₂), H₂O (pHH₂O) and lime-requirement buffer (pH_buffer) were compared to those derived from laboratory extractions. Comparisons of lime requirement were made using pH_buffer measurements and a response surface regression. The mid-infrared with partial least squares predictions were less accurate. The nugget variance of lime requirements derived using both the conventional and the MIR showed that the latter may be a more precise analytical technique. Kriging of the MIR data produced a map suitable for variable-rate liming.

INTRODUCTION

Increasingly, spectroscopic methods are being considered as possible alternatives to enhance or replace conventional laboratory methods of soil analysis (e.g. Janik et al.,1995; Viscarra Rossel & McBratney, 1998a). The reasons for this may be that the techniques are rapid, timely, less expensive than conventional soil analysis (after initial set-up costs), and are occasionally more accurate than conventional laboratory methods (e.g. McCauley et al., 1993). The development of a proximal spectrometric soil sensing system would be an useful step towards real-time, continuous site-specific management, which is the ultimate aim of precision agriculture (PA). The aims of this work are: (i.) to determine the utility of MIR soil analysis with partial least-squares regression (MIR-PLS) for the prediction of soil pH and lime requirement, and (ii.) to assess its adaptability for use in precision agriculture. The MIR portion of the electromagnetic spectrum occurs in the 2500 to 25000 nm wavelength range, corresponding to a wave number range from 4000 to 400 cm^-1. Mid-infrared spectroscopy has a particular advantage over other spectroscopic techniques for soil analysis (e.g. near-infrared) as it is sensitive to the intense fundamental vibrations of both the organic and inorganic phases of the soil. A review of the technique may be found in Janik et al. (1998).