One of the classical weaknesses of remote sensing algorithms is their statistically limited validation data sets. Validation procedures are generally carried out over sparse ground-based data sets which are orders of magnitude smaller in number than the image products they are meant to validate and often of a physical scale that is inconsistent with the footprint scale of a single pixel. The multialtitude regression methodology seeks to address this problem by exploiting the flexibility and programmability of an airborne sensor whose potential for acquiring algorithm test data is much more commensurate with the validation needs of high altitude or satellite sensors. The multialtitude regression procedure was employed to validate a single-altitude aerosol optical depth (AOD) inversion algorithm which uses an atmospherically resistant vegetation index criterion to select dense dark vegetation (DDV) pixels in a boreal forest image acquired by the CASI (Compact Airborne Spectrographic Imager) imaging spectrometer. The multialtitude regression procedure permits the extraction of AOD images that are reasonably independent of the surface bidirectional reflectance factor (BRF) required as input to the DDV-based AOD inversion algorithm. This independence of surface BRF is a fundamental requirement for the validation of DDV-based algorithm since the basic weakness of this algorithm is its sensitivity to surface BRF variations. The results indicate that the multialtitude regression procedure is an effective tool for validating DDV inversion algorithms. ¿ 1999 American Geophysical Union