Carolina C. Molina Segura

Remote sensing techniques have been used effectively in both near real-time and post-event analyses of broad-extent disturbances within forested landscapes. In this project, high spatial resolution imagery (0.6 m pansharpened Quickbird data) was used to map forest disturbance in Point Pleasant Park, Halifax, Nova Scotia, Canada caused by Hurricane Juan in fall 2003. To identify location and intensity of damage, image differencing of atmospherically corrected 2002 and 2004 Normalized Difference Vegetation Indices was used. A thresholding approach, based on known forest damage intensity, was applied to classify the difference image according to non-damaged forest, moderate to severe damage, and complete canopy loss. Forest damage intensity was determined to be 16.03% moderate to severe damage and 28.12% complete canopy loss; remaining forest experienced little to no appreciable damage. An error analysis of these results reported a Kappa Statistic of 0.87. To attempt an explanation of the resulting damage patterns, physical characteristics (slope and aspect) of the park were investigated for spatial correlation with damage presence and intensity. Results suggest that the physical landscape did play some role in predisposing forest to damage; however, un investigated variables including stand age and tree condition, as well as random down bursts of wind during the hurricane, are thought to be additional explanatory factors.