In most forestry remote sensing applications in steep terrain, simple photometric and empirical corrections are confounded as a result of variable stand and species structure with terrain and the anisotropic reflective properties of vegetated surfaces. To address these problems, we test two new topographic correction approaches based on Sun-Canopy-Sensor (SCS) geometry. SCS is more appropriate than strictly terrain-based corrections in forested areas since it preserves the geotropic nature of trees (vertical growth with respect to the geoid) regardless of terrain, view and illumination angles. The first SCS approach accounts for diffuse atmospheric irradiance based on the C-correction (SCS+C). Secondly, a new multiple forward mode (MFM) canopy reflectance model based correction (MFM-TOPO-COR) is introduced which normalizes topographically induced signal variance as a function of forest stand structure and sub-pixel scale components, while also maintaining proper SCS geometry. These two new techniques are compared to existing correction methods (cosine, <i>c</i> correction, Minnaert, statistical-empirical, SCS, and <i>b</i> correction) in a Rocky Mountain forest setting in western Canada. The ability of these eight correction methods are tested and compared for removing topographically induced variance and for improving the classification accuracy of a SPOT image over this sub-alpine and alpine forest area. The new MFM-TOPO-COR canopy reflectance model correction method is shown to provide the greatest improvement in classification accuracy within a species and stand density based class structure. The potential and limitations of this new approach are critically discussed.