Evaluation of ikonos satellite imagery for detecting ice storm damage to oak forests in Eastern Kentucky
Ice storms are a recurring landscape-scale disturbance in the eastern U.S. where they may cause varying levels of damage to upland hardwood forests. High-resolution Ikonos imagery and semiautomated detection of ice storm damage may be an alternative to manually interpreted aerial photography. We evaluated Ikonos multispectral, winter and summer imagery as a tool for detecting forest canopy damage to oak dominated forests resulting from an ice storm that occurred in the Daniel Boone National Forest in February 2003. The objectives of this exploratory study were to determine if classification accuracy was affected by: (1) spectral band, (2) size of training window, and (3) season of imagery. Within a 100 km2 study area, field sites representing three land cover types (forested with none to light canopy damage, forested with moderate to heavy canopy damage, and non-forested) were established or identified to provide image training signatures. Classification accuracies averaged between 65 and 70 percent overall for the three cover types and was highest (>80 percent) for the non-forest type. Results of this pilot study suggest that Ikonos imagery is useful for detecting ice storm damage to upland hardwood forests based on multispectral analysis. Additional study is needed, however, to determine if classification results from the small training areas can be expanded to produce a landscape-scale map of estimated forest damage with accuracy adequate for resource management purposes.