In this study, the potential of combining NIR and MIR spectral range for the determination of Saturate, Aromatic, Resin and Asphaltene (SARA) contents in heavy oils was evaluated. The work aimed at better understanding the contribution of each spectral range and, at highlighting complementary information for the description of SARA contents. A further objective was to observe if more accurate models could be obtained combining the two spectral domains. For this purpose, two multiblock methods, MB-PLS and S-PLS, were chosen. In a first step, PLS models were developed for each spectral range and results were compared on the basis of a statistical test. No significant difference was found for the prediction of resins and asphaltenes. On the contrary, for the prediction of saturate and aromatic contents, NIR performed better. In a second step, MIR and NIR spectral range were combined using MB-PLS and S-PLS. A significant improvement was obtained for the prediction of resin content using MB-PLS. In addition, the contributions of CAH out of plane bending vibrations and aromatic C=C stretching vibrations for the description of highly unsaturated compounds could be highlighted in MIR. (C) 2014 Elsevier Ltd. All rights reserved.