Main content area

Using AVIRIS to assess hemlock abundance and early decline in the Catskills, New York

Pontius, Jennifer, Hallett, Richard, Martin, Mary
Remote sensing of environment 2005 v.97 no.2 pp. 163
Tsuga canadensis, decline, forest trees, insect pests, Adelges tsugae, monitoring, hyperspectral imagery, image analysis, remote sensing, chlorophyll, spatial data, detection, spectral analysis, wavelengths, New York
In order to aid land managers in monitoring and controlling the ongoing hemlock woolly adelgid outbreak, more accurate landscape scale tools are required to locate the hemlock resource, identify infestation and spot early decline. To this end, NASA's Airborne Visible Infra-red Imaging Spectrometer was flown over the infestation front in the Catskills region of New York during the summer of 2001. Mixture Tuned Matched Filtering in ENVI was used to "unmix" spectra and quantify the hemlock signature contribution to each pixel. The resulting percent hemlock basal area coverage correctly identified hemlock dominated pixels (> 40% basal area) with 83% accuracy. Key wavelengths and health indices were examined to determine if a subset of wavelengths could accurately predict an 11-class decline rating system. A linear regression based on reflectance at a chlorophyll sensitive wavelength (R683 nm), coupled with a water band index (R970/R900), was able to predict decline with 85% accuracy. The extreme accuracy at the low (0-3) end of the range indicated that these wavelengths might be used to assess early decline, before visual symptoms are apparent in the field.