Using Landsat imagery and FIA data in an original approach to map and describe forest dynamics in south-central Pennsylvania.

Clay Baros, University of Maryland Center for Environmental Science – Appalachian Laboratory, cbaros@al.umces.edu

Dr. Phil Townsend, University of Maryland Center for Environmental Science – Appalachian Laboratory, townsend@al.umces.edu

Change detection analyses in remote sensing offer an array of techniques to identify spatial patterns of variations in surface reflectance over time. This paper introduces an unconventional approach towards change detection using a lengthy record of Landsat imagery along with FIA (Forest Inventory and Analysis) plot data occurring in the deciduous forests of central Pennsylvania. From the Landsat imagery, I extracted reflectance data corresponding to the FIA plot locations for fourteen dates acquired between 1984 thorough 2000. Reflectance values for each plot within a scene were compared to the average reflectance value of like plots on common dates. This technique permitted the establishment of a scale to evaluate how each forest plot compared with the normal reflectance values of that date. Three separate tiers of the FIA data were analyzed and compared: one comparing only chestnut oak-dominated plots to their average reflectance value, one comparing only plots dominated by species of the family Fagaceae (primarily oaks), and one encompassing all deciduous plots. The magnitude of deviation from each date’s average was then tracked through time to establish a trend line detailing each plot’s propensity towards normalcy. The peaks and valleys of this trend line revealed the inclination of a plot towards change. The nature of those changes was evaluated through correlating change variables to environmental variables describing climate. Specifically, canonical correlation was used to identify the ostensible driving forces behind changes in reflectance over a given area. Finally, maps of all three forest tiers and their associated change trajectories were produced describing the change variables and their canonical variates so as to identify and compare contiguous areas of forest change.