the effects of topography on the radiometric properties of multispectral scanner (mss) data are examined in the context of remote sensing of forests in mountainous regions. two test areas in the coastal mountains of british columbia are studied: one near the anderson river and the other near gun lake. the predominant forest types are douglas fir and lodgepole pine and ponderosa pine. both regions have rugged topography with elevations ranging from 275 to 1,500 m and 670 to 1,990 m, respectively. lambertian and non-lambertian illumination corrections are formulated, considering atmospheric effects and topographic variations. terrain slope and aspect are determined from a digital elevation model, and atmospheric parameters are derived from a model atmosphere computation. in the lambertian approximation, a cosine correction is used, but it is shown to be inadequate for larger angles of incidence. semi-empirical functions based on cosines of incident and reflected angles are used to remove topographic effects. correlations between topographic parameters and mss radiance values are investigated for different forest types. multitemporal landsat mss data and 11-channel airborne mss data are analyzed. slope-aspect correction algorithms are implemented in software at the canada centre for remote sensing. geometric rectification is necessary to relate image geometry to map coordinates. a special technique involving flight line modeling is used for aircraft data. feature selection based on divergence criteria shows terrain parameters compare favorably with mss data in distinguishing forest classes. maximum likelihood classification results for mss data corrected for slope-aspect effects show little improvement over uncorrected data. this outcome is discussed to better understand physical principles and image processing methodologies.the effects of topography on the radiometric properties of multispectral scanner (mss) data are examined in the context of remote sensing of forests in mountainous regions. two test areas in the coastal mountains of british columbia are studied: one near the anderson river and the other near gun lake. the predominant forest types are douglas fir and lodgepole pine and ponderosa pine. both regions have rugged topography with elevations ranging from 275 to 1,500 m and 670 to 1,990 m, respectively. lambertian and non-lambertian illumination corrections are formulated, considering atmospheric effects and topographic variations. terrain slope and aspect are determined from a digital elevation model, and atmospheric parameters are derived from a model atmosphere computation. in the lambertian approximation, a cosine correction is used, but it is shown to be inadequate for larger angles of incidence. semi-empirical functions based on cosines of incident and reflected angles are used to remove topographic effects. correlations between topographic parameters and mss radiance values are investigated for different forest types. multitemporal landsat mss data and 11-channel airborne mss data are analyzed. slope-aspect correction algorithms are implemented in software at the canada centre for remote sensing. geometric rectification is necessary to relate image geometry to map coordinates. a special technique involving flight line modeling is used for aircraft data. feature selection based on divergence criteria shows terrain parameters compare favorably with mss data in distinguishing forest classes. maximum likelihood classification results for mss data corrected for slope-aspect effects show little improvement over uncorrected data. this outcome is discussed to better understand physical principles and image processing methodologies.