Remote detection of forest structure in the White Mountains of New Hampshire: An integration of waveform lidar and hyperspectral remote sensing data

The capability of waveform lidar, used singly and through integration with high-resolution spectral data, to describe and predict various aspects of the structure of a northern temperate forest is explored. Waveform lidar imagery was acquired in 1999 and 2003 over Bartlett Experimental Forest in the White Mountains of central New Hampshire using NASA\u27s airborne Laser Vegetation Imaging Sensor (LVIS). High-resolution spectral imagery from 1997 and 2003 was likewise acquired using NASA\u27s Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). USDA Forest Service Northeastern Research Station (USFS NERS) 2001-2003 inventory data was used to define basal area, above-ground biomass, quadratic mean stem diameter and proportional species abundances within each of over 400 plots. Field plots scaled to LVIS footprints were also established. At the smallest scale, metrics derived from single LVIS footprints were strongly correlated with coincident forest measurements. At the larger scale of USFS NERS plots, strong correlations encompassing the full variability of the Forest Service data could not be established. Restrictions set by species composition and land-use, however, significantly improved both the descriptive and predictive power of the regression analyses. Higher amplitude values of 1999 LUIS ground return metrics obtained within two years of the January 1998 ice storm, were found to provide a spatial record of higher levels of canopy damage within older, unmanaged forest tracts. Subjected to repeated disturbance of intermediate severity over the time frame of decades, these particular tracts, predominately found on southeastern aspects, simultaneously support by levels of sugar maple abundance and low levels of sugar maple coarse woody debris. LVIS height metrics were used here to establish a statistical relationship with coarse woody debris data. The integration of waveform lidar with hyperspectral data did enhance the ability to remotely describe a number of common measures of forest structure. Compositional abundance patterns, however, were not improved over use of AVIRIS data alone. Maps predicting species abundance patterns (primarily derived from AVIRIS data) with coincident patterns of stem size (derived from LVIS data) can be created for several of the dominant tree species of this region. The results are the near equivalent of a field-based forest inventory