Early detection of delayed tree mortality postfire using hyperspectral remote sensing

Implementation of effective postfire recovery management requires the detection of delayed tree mortality in order to prevent high economic loss and unexpected ecological impacts. Traditional methods to detect mortality include laborious direct readings of cambial tissue or indirect correlative analysis from crown scorch volume. Integrating remote sensing techniques can provide timely information over larger geographic areas. Multispectral data lack the sensitivity needed to detect delayed tree mortality, but hyperspectral data may be able to detect these subtle changes in canopy reflectance resulting from physiological stress, specifically phloem girdling. This study examined the use of ground-based hyperspectral data to differentiate between three stress treatments on ponderosa pine; fire induced phloem girdling, dehydration, and a rewetting cycle to dehydrated trees. The spectral reflectance properties of Pinus ponderosa were examined over the visible, near-infrared, and short wave infrared wavelength regions (350-2500 nm). Relationships among the spectral data, stomatal conductance, chlorophyll fluorescence, and pre-dawn leaf water potential were analyzed through time. The multivariate analysis of variance revealed a significant multivariate group by time interaction  (Pillai's Trace =1.57, p < 0.04). Results suggest that hyperspectral remote sensing can be used to provide earlier detection of delayed tree mortality postfire as well as differentiate between lethally dehydrated trees and those that can rejuvenate once watered. Ground-based data may eventually be extrapolated from a leaf to a canopy level using airborne hyperspectral sensors allowing forest managers to implement effective postfire recovery strategies, direct salvage logging, and/or model eventual snag habitat.