Development of a Danish national Vis-NIR soil spectral library for soil organic carbon determination

The development of a soil spectral library is a prerequisite for using vis-NIR spectroscopy for soil analysis. The aim of this study is twofold: (i) to create a Danish national soil library with a main focus on building a model for SOC determination and (ii) to test the feasibility of using the library to predict SOC from new dry and moist samples, before and after spiking the library with additional local samples. Soils spectra were obtained using Labspec 5100 (350-2500 nm) vis-NIR sensor. The best validation model resulted in a root mean square error of prediction (RMSEP) value of 0.31, R-2 value of 0.79, and ratio of performance to deviation (RPD, defined as the ratio between the standard deviation and the RMSEP) value of 2.3. The prediction on independent data resulted in very similar results: RMSEP = 0.29, R2 of 0.81 and RPD of 2.4. Before spiking the library, moderate SOC predictions were obtained for the new dry soils. Prediction on moist samples however, required prior spiking the global library with moist samples from the new location.