In the present paper, the effect of shadows in the classification of three successional stages of a tropical dry forest (TDF) in Mexico, using hyperspectral and multi-angular CHRIS/PROBA images, is evaluated. An algorithm based on the cosine of the angle of solar incidence on the terrain is applied to correct the effect of topography on CHRIS/PROBA reflectances. Previous to the removal of shadows caused by topography, CHRIS/PROBA images were atmospherically corrected in BEAM software. Vegetation maps of the study site were generated using non-parametric decision trees, defining four main classes: late, intermediate and early stages of forest succession within a tropical dry forest, and riparian forests. By comparing the vegetation maps before and after shadow removal in CHRIS/PROBA spectral data, it was observed that the late stage of succession and riparian forests are overestimated for the non-corrected images while intermediate and early stages of succession are underestimated. Errors in classification are more important for the large CHRIS/PROBA viewing angles. Therefore, the removal of shadows caused by topography is necessary for an accurate classification of successional stages in tropical dry forests.