Satellite Remote Sensing of Forest Disturbances Caused by Hurricanes and Wildland fires

Abstract

The impact of forest disturbances, especially those caused by hurricanes and wildland fires, on forest ecosystems and forest carbon sequestration has become more pronounced in recent years. Remote sensing of these two natural hazards and their impacts received increasing attention among research communities for the carbon cycle study and forest management. The research presented in this dissertation is dedicated to promoting remote sensing of forest disturbances caused by hurricanes and wildland fires from space. This study developed an algorithm for rapidly assessing post-hurricane forest damage using MODIS measurements, without relying on intensive ground inventory or sampling. The performance of five commonly used vegetation indices as post-hurricane forest damage indicators was investigated, among which the Normalized Difference Infrared Index (NDII) was found the optimal vegetation index. This new algorithm was validated by ground measurements. The validation showed that the relative change of pre- and post-hurricane NDII was linearly related to the damage severity estimated in the ground inventory with a coefficient of determination of 0.79 and p value < 0.0001. This approach was applied to evaluate forest damage severity and the impacted forest region caused by Hurricane Katrina. Based on the MODIS enhanced contextual algorithm and a smoke detection algorithm, an improved algorithm for monitoring low intensity fires in regional scale was developed. Sources of omission errors in the MODIS active fire product were diagnosed using a wildland fire database. This database was a collection of spectral signatures of low intensity fires, including fires missed by the MODIS enhanced contextual algorithm. This algorithm was applied and evaluated by case studies, which showed that this improved algorithm was more suitable for regional low intensity fire detection in the southeastern United States. These algorithms and findings contribute to studies of the natural hazard detection, carbon cycle study and forest management. The rapid assessment algorithm can provide timely information on forest live fuel loading change, impacted regions and damage severity. The availability of change information on fuel loading will allow the quantitative study of hurricane impacts on forest fire danger. The improved fire detection algorithm can provide more accurate information on wildland fire events and decision support for firefighting activities and forest management.