Throughout the central hardwood and southern Appalachian forest regions, managers are using prescribed fire to achieve a suite of management objectives. When management occurs on federal lands, monitoring is mandated to determine whether objectives have been met, yet limited funding restricts the number, size and spatial array of monitoring plots. Remote sensing data have the potential to provide accurate and relatively inexpensive information examining landscape-scale responses to fire in this landscape. We used Landsat satellite imagery to (1) assess landscape patterns of forest response to burning, and (2) analyze the relationships between image-derived indices and field-measured forest characteristics (e.g., canopy openness, crown health). We used Landsat Thematic Mapper 5 images from 2002-2009 to calculate several spectral vegetation indices (SVI; e.g., wetness, greenness) and changes through time. Forest characteristics and their response to burning were measured in 93 forest plots located at random across three study sites, each with three treatments, unburned, burned twice, and burned four times, starting in 2003. Buffered polygons encompassing each plot were created from aerial photos and matched with plot-scale locations to capture a larger area of similar vegetation. Analysis of variance and regression were used to test landscape- and plot-scale responses.
Results/Conclusions
Prior to statistical analysis, every pixel was assigned a ‘landscape position’ using a derived metric based on elevation, slope, aspect and hillslope shading to represent sub-xeric, intermediate, and sub-mesic landscape positions. Differences in SVIs among burned and unburned sites were significant for sub-xeric and intermediate landscape positions but not for sub-mesic positions, coinciding with our plot-based analysis of fire impacts to stand structure. Differences in SVIs were greatest the first year after burning, and were increasingly difficult to detect in subsequent years, coinciding with regrowth of vegetation. Significant associations of plot-level SVIs with measured forest characteristics, such as basal area or stem density, were limited to sub-xeric locations. The management objective of the prescribed burning in our study was to remove midstory stems of fire-sensitive species, and resulted in fairly small changes to the overstory. Prescribed fire used to create a range of habitats with more open canopies, such as oak woodlands and savannas, would be more easily detected using this approach.