Surface Roughness and Near-Infrared Reflectance Sensing of Soils.

Near-infrared reflectance sensing (NIRS) of soils has been the object of considerable interest and research in the last few years, motivated in large measure by the prospects this method seems to offer to replace conventional, time-consuming soil analysis techniques, to design new “proximal” sensors in support of precision agriculture, and to increase significantly the amount of information that can be obtained from satellite-based sensors. With respect to these last two objectives, a practical difficulty is that soils in the field, unlike the sieved, repacked soil samples used in the laboratory, are generally moist and have uneven surfaces, especially after tillage. Whereas the effect of moisture on NIRS has been analyzed in some detail, little is known at this point on the effect of surface roughness. In this context, the present research focuses on the application of NIRS, under laboratory conditions, to “aggregates” of average sizes between 0.04 and 8 mm, obtained in 5 different soils with contrasting features, and repacked in Petri dishes. NIRS measurements were performed when the systems were air-dry, and after rewetting to near-saturation. In virtually all cases, except at the finest aggregate size in two soils, the near-infrared reflectance decreased regularly as aggregate size increased. Addition of moisture resulted in all cases in further decreases in reflectance, which obliterated to varying extent the dependence of the reflectance on aggregate size. For most cases, whether dry or near-saturated, computation of the first derivative of the NIR spectra, especially when preceded by moving-average or wavelet-based smoothing, resulted in transformed signals that were virtually independent of surface roughness in a number of spectral regions. These observations suggest that in the range of aggregate sizes considered, it might be possible practically to circumvent the dependence of NIRS on surface roughness.