Significant models predicting Soil Organic Carbon (SOC) and other chemical and biological indicators of soil health in an experimental farm with semi-arid Mediterranean Calcisol have been obtained by partial least squares (PLS) regression, with mid-infrared (MIR) spectra of whole soil samples used as independent variables (IVs). The dependent variables (DVs) included SOC, pH, electric conductivity, N, P2O5, K, Ca2+, Mg2+, Na+, Fe, Mn, Cu and Zn. The DVs also included free-living nematodes and microbivores, such as Rhabditids and Cephalobids, and phytoparasitics, such as Xiphinema spp. and other Dorylaimids. More importantly, an attempt was made to determine which spectral patterns allowed each dependent variable (DV) to be predicted. For this purpose, a number of statistical indices were plotted between 4000 and 450 cm\u22121, e.g., variable importance for prediction (VIP) and beta coefficients from PLS, loading factors from principal component analysis (PCA) and correlation and determination indices. The most effective plots, however, were the \u201cscaled subtraction spectra\u201d (SSS) obtained by subtracting the averages of groups of spectra in order to reproduce the spectral patterns typical in soils where the values of each DV are higher, or vice versa. For instance, distinct SSS resembled the spectra of carbonate, clay, oxides and SOC, whose varying concentrations enabled the prediction of the different DVs.