Soil organic carbon (SOC) and other soil properties related to carbon sequestration (eg. soil clay content and mineralogy) vary spatially across landscapes. To cost effectively capture this variability, new technologies, such as Visible and Near Infrared (VisNIR) spectroscopy, have been applied to soils for rapid, accurate, and inexpensive estimation of SOC and other soil properties.  For this study, we evaluated an “on the go” VisNIR sensor developed by Veris Technologies, Inc. (Salinas, KS) for mapping SOC, soil clay content and mineralogy. 

The Veris spectrometer spanned 350 to 2224 nm with 8nm spectral resolution, and 25 spectra were integrated every 2 seconds resulting in 3 -5 m scanning distances on the ground.  The unit was mounted to a mobile sensor platform pulled by a tractor, and scanned soils at an average depth of 10 cm through a quartz-sapphire window.  We scanned eight 16.2 ha (40 ac) wheat fields in north central Montana (USA), with 15 m transect intervals.  Using random sampling with spatial inhibition, 100 soil samples from 0-10 cm depths were extracted along scanned transects from each field and a subset of samples were analyzed using standard laboratory procedures to determine SOC, clay content, and mineralogy.  Neat, sieved (<2 mm) soil sample materials were also scanned in the lab using an Analytical Spectral Devices (ASD, Boulder, CO, USA) Fieldspec Pro FR spectroradiometer with a spectral range of 350-2500 and spectral resolution of 2-10 nm. 

The analyzed samples were used to calibrate and validate a number of VisNIR models to compare on-the-go scanning vs. higher spectral resolution laboratory spectroscopy vs. standard SOC measurement methods  To construct calibrations, we used partial least squares regression (PLSR) and experimented with different spectral transformations (1^st derivative, 2^nd derivative, Kubelka-Munk, and absorbance).