Tree species mapping with spectral mixture analysis: Applications for moderate- and fine-scale spatial resolution imagery
Lucie
C.
Plourde, University of New Hampshire, lucie.plourde@unh.edu
(Presenting)
Jennifer
H.
Pontius, University of Vermont, jennifer.pontius@uvm.edu
Scott
V.
Ollinger, University of New Hampshire, scott.ollinger@unh.edu
Mary
E.
Martin, University of New Hampshire, mary.martin@unh.edu
Richard
A.
Hallett, USDA Forest Service Northern Research Station, rah@unh.edu
Characterizing the spatial distribution of tree species in forest ecosystems is central to a range of scientific and management issues. Knowledge of species-level abundance patterns can inform research concerned with biodiversity and habitat quality, both of which are increasingly threatened by loss of native species, shifts in distribution brought about by climate change, and the spread of exotic insects and pathogens. For example, two insect pests native to Asia—the hemlock woolly adelgid and the emerald ash borer—continue to cause widespread mortality of Tsuga canadensis L. Carr. and Fraxinus spp. L., both economically and ecologically important species in northern temperate forests. Accurate mapping of host tree species is thus imperative to detecting and monitoring the spread of these pests, and moreover, to gaining an understanding of their short- and long-term effects on the forest resource. Tree species distribution has also become vital to studies in biogeochemistry, where there is increasing evidence that changes in health and distribution of individual tree species may have implications to C and N cycling. Further, accurate mapping of species abundance allows for more effective parameterization of ecosystem models and thus improved model predictions.
Here we describe results from three research projects that employed spectral mixture analysis (SMA) for tree species mapping. SMA was applied to NASA’s Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) data and SpecTIR’s VNIR sensor data to determine fractional pixel composition for a range of species abundance and health conditions. Our results suggest that using SMA to estimate species abundances is an effective approach to classifying heterogeneous forests, and holds promise for a variety of ecological studies, spatial applications of ecosystem models, and forest land management.
NASA Carbon Cycle & Ecosystems Active Awards Represented by this Poster: