Coastal forest impact and recovery from hurricane storm surge: The use of remote-sensing to assess extreme disturbance in the lower Florida Keys

Background/Question/Methods

Disturbance to coastal forests from hurricanes alters forest structure and ecosystem functions through wind damage and storm surge flooding. Global climate change and sea level rise of the 21^stcentury will likely change both the effect of and recovery from hurricanes, favoring disturbance tolerant species by exposing coastal forests more frequently to storms, while shrinking the recovery period between events. In this study, we quantify the effects of two hurricanes that struck the islands of the lower Florida Keys, Hurricane Georges (1998) and Hurricane Wilma (2005), and ask whether forest recovery, or a lack thereof, varied along an elevation gradient and between two coastal forest communities, hardwood hammock and pine rockland. Using Landsat TM data, we first compare the ability of two satellite-derived vegetation indices, the normalized differenced vegetation index (NDVI) and the normalized differenced moisture index (NDMI), to identify disturbance impact. Then, we employ empirical orthogonal function (EOF) analysis to assess spatiotemporal patterns of disturbance impact and recovery (1986 to 2011) and assess differences between these two communities and topographic elevation. Finally, we assess the relationship between vegetation indices and measured forest stand parameters, basal area per hectare and shrub cover, in permanent plots sampled in 1990 and 2012.

Results/Conclusions

Both vegetation indices showed a sharp departure from the pre-disturbance control, but NDMI showed a greater capacity to detect impacts from hurricane disturbance than that of NDVI. The magnitude of the decrease and the time to recovery of NDMI values varied by community, as well as between the two events, with much greater decreases observed post-Hurricane Wilma. The results of EOF analysis showed disturbance impact and recovery varied as a function of both elevation and community type and indicated two separate trajectories for post-disturbance recovery. While hardwood hammock NDMI values returned to pre-disturbance levels within a couple years, six years after Hurricane Wilma, pine rockland NDMI values were still depressed. A strong positive correlation was observed for NDMI and basal area per hectare in both community types. Our results confirm that proper selection of index is essential to forest disturbance monitoring, and the incorporation of EOF analysis enabled us to see diverging spatial trends in recovery in the two communities. We conclude that the incorporation of remotely-sensed data combined with a statistical technique typical of another scientific discipline, can enhance the assessment of coastal forest change from disturbance over large spatiotemporal scales.