Recycling organic wastes into farmland faces a double challenge: increasing the carbon storage of soil while mitigating CO2 emission from soil. Predicting the stability of organic matter (OM) in wastes and treatment products can be helpful in dealing with this contradiction. This work proposed a modeling approach integrating an OM characterization protocol into partial least squares (PLS) regression. A total of 31 organic wastes, and their products issued from anaerobic digestion, composting, and digestion-composting treatment were characterized using sequential extraction and three-dimension (3D) fluorescence spectroscopy. The apportionment of carbon in different fractions and fluorescence spectra revealed that the OM became less accessible and biodegradable after treatments, especially the composting. This was proven by the decrease in CO2 emission from soil incubation. The PLS model successfully predicted the stability of solid digestate, compost, and compost of solid digestate in the soil by using only the characterized variables of non-treated wastes. The results suggested that it would be possible to predict the stability of OM from organic wastes after different treatment procedures. It is helpful to choose the most suitable and economic treatment procedure to stabilize labile organic carbon in wastes and hence minimize CO2 emission after the application of treatment products to the soil.