A method for combining Landsat Thematic Mapper (TM), Advanced Very High Resolution Radiometer (AVHRR) imagery, and other biogeographic data to estimate forest cover over large regions is applied and evaluated at two locations. In this method, TM data are used to classify a small area (calibration center) into forest/nonforest; the resulting forest cover map is then used in combination with AVHRR spectral data from the same area to develop an empirical relationship between percent forest cover and AVHRR pixel spectral signature; the resultant regression relationship between AVHRR band values and percent forest cover is then used to extrapolate forest cover for several hundred kilometers beyond the original TM calibration center. In the present study, the method was tested over two large regions in the eastern United States: areas centered on Illinois and on the Smoky Mountains on the North Carolina-Tennessee border. Estimates of percent forest cover for counties, after aggregating AVHRR pixel estimates within each county, were compared with independent ground-based estimates. County estimates were aggregated to derive estimates for states and regions. For the Illinois region, the overall correlation between county cover estimates was 0.89. Even better correlations (up to r = 0.96) resulted for the counties close to one another, in the same ecoregion, or in the same major land resource region as the calibration center. For the Smokies region, the correlations were significant but lower due to large influences of pine forests (suppressed spectral reflectance) in counties outside the hardwood-dominated calibration center. The method carries potential for estimating forest cover across the globe. It has special advantages in allowing the assessment of forest cover in highly fragmented landscapes, where individual AVHRR pixels (1 km2) are forested to varying degrees.