Laboratory-based evaluation of optical performance for a new soil penetrometer visible and near-infrared (VisNIR) foreoptic

Abstract

In situ visible and near infrared (VisNIR) spectroscopy has been gaining interest as a proximal soil sensing technique to rapidly and inexpensively measure soil properties. For this study, our goal was to design a VisNIR soil penetrometer foreoptic that matched the optical performance of a commercial ASD (Analytical Spectral Devices Inc., Boulder, CO, USA) foreoptic and was compatible with any spectrometer having a fiber optic input (e.g. commonly used ASD field spectrometers). Space available for penetrometer optics was constrained by a several practical design requirements: (a) ability to collect in situ soil VisNIR reflectance and insertion force simultaneously; (b) thick probe walls strong enough for insertion into dry soil with high clay content; and (c) a narrow enough cross-sectional diameter for insertion using a standard soil coring hydraulic push system. In this paper, we evaluate the optical and chemometric performance of our design relative to a commercial ASD foreoptic under controlled conditions in a laboratory setting, in order to isolate the optical performance of the probe from complications associated with in situ use. Both the VisNIR penetrometer and ASD Contact Probe, attached to an ASD AgriSpec spectrometer in the lab, were used to interrogate 389 milled and pressed surface and subsoil samples taken from two US continent-scale transects. The reflectance spectra obtained from these two foreoptics were highly correlated (r>0.95 for most wavelengths) and the 1st derivatives of the spectra were well correlated for all but the detector splice points (r>0.9 for most wavelengths). Partial least squares regression (PLSR) was used to calibrate and validate VisNIR models predicting soil organic carbon content for both foreoptics. Using reflectance data, our VisNIR penetrometer yielded quantitative SOC predictions with standard error of prediction (SEP=0.19log[gkg−1]) and ratio of prediction to determination (RPD=2.34) highly similar to the ASD contact probe (SEP=0.18 and RPD=2.45). Regression coefficients of PLS calibration models from both foreoptic were comparable and highly correlated (r=0.91). These results suggest that our penetrometer foreoptic provides VisNIR reflectance spectra comparable to a widely-used commercial foreoptic. In future work, we deploy this penetrometer in the field and evaluate the performance of this foreoptic where soil heterogeneity, smearing and moisture can impact the quality of VisNIR spectra.