Remote sensing and regression analysis of impacts from Hurricane Maria on forest of Puerto Rico

Friday, December 14, 2018 - 13:40
Add to Calendar

Increasing hurricanes have a significant impact on tree mortality and forest structures. As the worst natural disaster on record in Puerto Rico, Category 5 Hurricane Maria traversed the island in a northwesterly direction causing widespread destruction. This study focused on Hurricane Maria’s impact on Puerto Rico’s forests and the associated factors that can explain the patterns of tree mortality. We used Google Earth Engine to generate calibrated and corrected Landsat 8 image composites for the entire island. We chose pre-Maria and post-Maria time periods that accounted for phenology. Spectral mixture analysis (SMA) were carried out on both composites to quantify the severity of forest disturbance using the non-photosynthetic vegetation (ΔNPV) spectral response. A ΔNPV map for only the forested pixels of the island showed significant spatial variability in the disturbance. By applying statistical regression, we analyzed forest disturbance with the estimated wind condition and forest structures. The factors include elevation, slope, windwardness, annual precipitation, distance to hurricane track, distance to hurricane landfall, forest canopy height, forest type, forest age, and pre-hurricane green vegetation ratio. Forest type and forest age became the most significant variables. Sierra Palm, Transitional and Tall Cloud forests, Seasonal Evergreen Forest with Coconut Palm, and Puerto Rico’s oldest forest mangrove sustained the highest disturbance. The full regression model explained 34% of forest disturbance. This study showed the great usefulness of remote sensing combined with statistical modeling in the rapid assessment of hurricane damage to forests.

Link for More Information: