Comparison of Hyperspectral Aviris and Landsat Tm Imagery for Estimating Burn Site Pine Seedling Regeneration Densities in the Central Plateau of Yellowstone National Park

Knowledge of seedling regeneration in burn sites is a valuable aspect of ecosystem monitoring. Remote sensing methods are ideal for such monitoring since large spatial extents can be monitored at multiple time scales. The goal of this project was to asses how well hyperspectral AVIRIS data can discriminate the differences between seedling densities in postfire regenerating sites, as compared to Landsat TM data. The field data were collected in the burn sites (legacies of the 1988 fires) of the Central Plateau, Yellowstone National Park during the summer of 1997. Timber cruising methods were used to count the seedlings at 24 sites. Other data reported included the basal area of the dead wood in the study plots. The project incorporated statistical methods (discriminant analysis), AVIRIS hyperspectral image and Landsat TM image classification and analysis methods (endmember selection/ evaluation and Spectral Angle Mapper classification). The statistical analysis showed that the hyperspectral imagery classified the seedling density classes with almost a 20 % improvement in accuracy compared to using only the Landsat TM imagery. This improvement was solemnly based on the hyperspectral content of the data, as the spatial variability was kept constant, by resampling the AVIRIS imagery to Landsat TM imagery resolution (30 m per pixel resolution). When the spatial content of the AVIRIS data were applied the improvement was close to 10 %, suggesting that the spatial information, summarized by a texture measurer, contributed additional information to the classification of seedling densities. Other relationships observed included the correlation between spectral and spatial data and the basal area of dead wood in the study plots. One of the future goals of this work is to incorporate multi-temporal analysis, where the rate in change of the density classes will be applied to determine the rate of regeneration of the stands.