Assessing a VisNIR penetrometer system for in-situ estimation of soil organic carbon under variable soil moisture conditions

Abstract

Soils are one of the main terrestrial carbon sinks and contain more carbon than the combined total in the atmosphere and terrestrial vegetation. A major challenge is quantifying how much carbon is stored in the soil and how much it has been sequestered due to changed management practices. In this study, a visible and near-infrared (VisNIR) penetrometer system was developed to measure in-situ soil VisNIR reflectance spectra to a depth of 90 cm into the soil profile. A spectral library of VisNIR soil spectra collected in air-dry and ground states were used to calibrate Cubist regression models for estimating soil organic carbon (SOC) contents. The External Parameter Orthogonalisation (EPO) transformation was applied to the in-situ spectra to remove soil moisture and other in-situ effects from the spectra. This penetrometer system was used to estimate SOC contents at three sites on three different days under variable soil moisture states. The penetrometer EPO performed better compared to the EPO constructed from moist cores. After applying EPO transformation matrix, soil wetness did not have any influence on soil SOC content estimations. The SOC contents were estimated using the in-situ VisNIR system were validated against lab-measured SOC content with mean R 2 , RMSE, and bias of 0.88, 0.32%, and 0.15%, respectively. It is concluded that the VisNIR penetrometer can potentially be used to measure SOC content rapidly and cost-effectively in-situ with a fine depth resolution. • A penetrometer system was used to acquire in-situ soil VisNIR spectra. • A spectral library from agricultural regions of NSW was used for calibration. • External Parameter Orthogonalisation (EPO) reduced confounding soil moisture effect. • External validation was performed to measure soil organic carbon contents.