Abstract

Large-scale quantification of soil organic carbon (C) and nitrogen (N) stocks and their distribution between particulate (POM) and mineral-associated (MAOM) organic matter is deemed necessary to develop land management strategies to mitigate climate change and sustain food production. To this end, diffuse reflectance mid-infrared spectroscopy (MIR) coupled with partial least square (PLS) analysis has been proposed as a promising method because of its low labor and cost, high throughput and the potential to estimate multiple soil attributes. In this paper, we applied MIR spectroscopy to predict C and N content in bulk soils, and in POM and MAOM, as well as soil properties influencing soil C storage. A heterogeneous dataset including 349 topsoil samples were collected under different soil types, land use and climate conditions across the European Union and the United Kingdom. The samples were analyzed for various soil properties to determine the feasibility of developing MIR-based predictive calibrations. We obtained accurate predictions for total soil C and N content, MAOM C and N content, pH, clay, and sand (R2> 0.7; RPD>1.8). In contrast, POM C and N content were predicted with lower accuracies due to non-linear dependencies, suggesting the need for additional calibration across similar soils. Furthermore, the information provided by MIR spectroscopy was able to differentiate spectral bands and patterns across different C pools. The strength of the correlation between C pools, minerals, and C functional groups was land use-dependent, suggesting that the use of this approach for long-term soil C monitoring programs should use land-use specific calibrations.