Estimating forest cover in a boreal forest test site using Thematic Mapper data from two dates

The utility of Landsat Thematic Mapper (TM) imagery for estimating forest cover in a boreal test site was evaluated. Corrections were applied to both a spring and a summer image to put the data on the same radiometric scale. A number of features were then tested, including TM bands, spectral indices, and forest cover estimated by inversion of a reflectance model. All features derived from single-date imagery gave only low correlations with measured cover (|R|<.47), although the spring image had greater utility. Reflectance simulations were used to help understand how a number of scene factors modulated cover estimates. The spring image data were found to give better separation of cover in multispectral plots, greater dynamic ranges, and were less adversely affected by differences in stand structure. The utility of using both images in combination was dependent on how the reflectance information was used. Using TM bands from both dates in model inversions did not improve the results compared to using the spring image only. However, the band differences, taken between dates in TM3 and TM5, respectively, gave the strongest correlations with cover (|R|=.56 and .64). Experiments using simulated reflectances indicated that the TM5 band difference was less adversely affected than TM5 to between-stand differences in structure and crown reflectances. Temporal reflectance information may thus have some applicability in boreal environments, where the availability of ancillary information, required for accurate inversions using reflectance data, can be very limited. However, further work needs to be undertaken to identify optimal image acquisition dates and to better understand the biophysical aspects of the temporal changes in forest reflectance in boreal conditions.