Assessing ice storm damage to hardwood forest canopies using the Advanced Solid-State Array Spectroradiometer (ASAS) and Landsat TM imagery

A major ice storm in January 1998 damaged 17 million acres of forestland in the Northeastern United States and Quebec Province. In July 1998 the airborne multiangle hyperspectral sensor, Advanced Solid-State Array Spectroradiometer (ASAS), acquired imagery for nine ice damaged target areas in the White Mountain National Forest (WMNF). Ice damage was also measured in situ at 288 sample plots. Radiance values for nadir ASAS images were normalized and the blue shift of the red edge was represented as the radiance value at 714 nm (714nr). The 714nr for 210 sites in 3 damage classes differentiated severely damaged sites from moderately and lightly damaged sites. The classification map of 714nr in 3 damage classes produced an overall accuracy of 54%. The effectiveness of off-nadir ASAS imagery (+/-45°, +/-26° view angle) to differentiate levels of ice damage was evaluated. It was determined that the red edge features of normalized ASAS data were not significantly affected by view angle, and that when compared to off-nadir views, the nadir view was most effective for detecting damage. Three vegetation indices, Vegetation Index (VI), NDVI, and TM 5/4, were applied to a single post-event (August 1998) TM image and classification mapping and accuracy assessment methods were applied. The overall accuracies of the 3 indices to map 3 levels of damage were 68% for VI, 72% for NDVI, and 76% for TM 5/4. Finally, the effectiveness of the ASAS 714nr approach was compared to 3 Landsat TM change detection analyses using pre- and post-event differences in VI, NDVI, and TM 5/4 to detect ice damage. The overall accuracies of the 3 indices to map 3 levels of damage using change detection were 78% for VI, 82% for NDVI, and 78% for TM 5/4