The soil system demonstrates a high degree of spatial heterogeneity. Contemporary soil surveys attempt to capture this variability by describing and characterizing soil profiles and interpolating between sampled locations. Hence, relatively few locations are sampled because of time and cost constraints; thus soil map accuracy is insufficient for many applications. Rapid, inexpensive soil characterization methods are required. Recent research has shown the effectiveness of Visible-Near Infrared Diffuse Reflectance Spectroscopy (VNIR-DRS) to provide a non-destructive rapid prediction of soil physical, chemical, and biological properties of air-dried ground soil samples in the laboratory (Shepard and Walsh, 2002). However, the capability of VNIR-DRS for in situ soil analysis is less explored (Sudduth and Hummel, 1993).

The objective of our research was to evaluate the accuracy and precision of VNIR-DRS derived predictions of soil clay and carbon content, using scans of field-moist soil cores simulating in situ characterization. Seventy-two cores, from Central Texas, were scanned at field-moist and air-dried water content. Water potential measurements of field-moist cores were taken with a thermocouple psychrometer. Once scanned, the soil samples were characterized for clay and carbon content. The spectral data were processed by averaging on 10-nm intervals and by taking the first derivative. To calibrate the model, 70 percent of the cores were chosen randomly. The remaining 30 percent were used for model validation. Partial least squares regression analysis was used to model the correlation between soil properties and processed reflectance data. Preliminary results of measured versus predicted clay percentage yielded an r2 of 0.84 and root mean squared deviation of 6% clay content. The results of this study indicate that VNIR-DRS may be useful in measuring some properties of soil profiles in situ, which would allow for quantifying soil properties at high vertical and horizontal resolutions quickly and more cost effectively.