Multi-sensor fusion for the collection of soil information has been the subject of varying improvements in previous studies, but the
underlying prediction mechanisms for spectrally active and inactive properties are still poorly understood. By measuring Mid-Infrared (MIR) and X-ray Fluorescence (XRF) spectra, texture, total and labile Organic Carbon (OC) and Nitrogen (N) content, pH, and Cation Exchange Capacity (CEC) for n=117 soils from an arable field in Germany, our goal was to investigate the prediction mechanisms and benefits of model fusion. Using MIR spectra or elemental concentrations derived from XRF spectra, partial least squares regression models went through a three-step training and testing process. Two high-level fusion and two sequential hybrid strategies were also tested. MIR outperformed XRF when it came to inorganic properties (RPIQV for clay=3.4, silt=3.0, and sand=1.8) in the field under investigation, while MIR was superior for organic properties (RPIQV for total OC=7.7 and N=5.0). For these properties, there was little to no improvement in accuracy with even the optimal fusion approach. The large number of elements with variable importance in the projection scores >1 (Fe, Ni, Si, Al, Mg, Mn, K, Pb (clay only), and Cr) and strong spearman correlations (0.57 rs 0.90) with clay and silt account for the high XRF accuracy for these materials.
Share this article
Bioceramics Development and Applications received 1050 citations as per Google Scholar report
