Estimates of forest canopy fuel attributes using hyperspectral data

Increasingly severe forest fires in the west have triggered a high demand for accurate and timely information on forest fuel attributes. There is great interest in the potential for using recent advances in high spectral resolution remotely sensed imagery to estimate fuel characteristics. We combined field forest inventory and field spectroscopy in the Colorado Front Range with airborne imaging spectrometer measurements of the region to test their capacity to estimate fire related forest attributes including canopy cover, forest type, and burn severity in ponderosa pine (Pinus ponderosa) and Douglas-fir (Pseudotsuga menziesii var. glauca) dominated forests. Spectral angle mapper and mixture-tuned matched filtering techniques were tested for mapping fuel attributes. Estimates of canopy cover using spectral angle mapper techniques found 61% agreement with observed values, while mixture-tuned matched filtering estimates of forest canopy cover matched 78% with field observations. The distinction of ponderosa pine versus Douglas-fir is crucial for predicting fire spread in the Rocky Mountains; we found that spectral discrimination of these species was also promising, with an accuracy of 53–57%. The average canopy cover of mixed conifer forest in the area is 38.6%, 24.7% contributed by ponderosa pine and 13.9% by Douglas-fir. The values of canopy cover ranged from 53% to 56% in US Forest Service planned fuel treatment areas, among the highest in the region. Recent forest fires have created approximately 684 km2 of burned area, with very low canopy cover (13–22%).