Using Landsat Phenology Curves to Characterize Post-Fire Forest Responses in South Carolina, USA

Characterizing forest responses to disturbance over large geographic areas represents one of the most challenging aspects of ecosystem monitoring. Traditional remote sensing methods often assess annual or biennial forest change after a disturbance, selecting one image for every year or two years for the study period. However, by using multiple images per year, researchers can examine intra-annual vegetation patterns, or phenology. Phenology provides information on the timing of vegetation events, such as the onset of greenness and the amplitude of NDVI, which can then be used to classify vegetation communities and characterize land cover change over time. Using all available images collected by Landsat 5, 7, and 8 for the study area in South Carolina, I compared intra-annual fluctuations in various spectral indices in pre- and post-fire Landsat pixels, using nearby unburned pixels as an approximate control group, at varying levels of fire severity. Additionally, this research provides baseline pre- and post-fire phenology estimates for the two dominant forest groups in the study region, loblolly-shortleaf pine and oak-gum-cypress. The methods I developed take advantage of the freely available Landsat archive and can be used to characterize forest recovery following a variety of disturbances in the southeastern U.S. and other regions. Future research could examine the feasibility of using phenology metrics to develop predictive species maps at various timesteps following fire events in the region and to develop successional models for this landscape. Hopefully, this research will add to our understanding of how forests are responding to and recovering from fire in a human-impacted region of the U.S.