2 METHODS
Field data were collected for 20 plots at each site within 10 days of the acquisition of
remote sensing data. Fresh leaf samples were collected from five trees of each overstory
species in each plot. Nitrogen concentrations in all samples were determined by CHN
analysis and lignin by a sequential extraction/sulfuric acid digestion procedure [3]. Ten
litterfall collection baskets were placed on each plot to determine the relative canopy
biomass of each species. Litter was sorted by species, oven dried, and weighed. Foliar
mass was determined for deciduous species by litter sample weight. For evergreen species,
litter weight was multiplied by a measured foliar retention time (in years). Total canopy
nitrogen and lignin concentrations were calculated as mean concentration per species
weighted by foliar mass per species.
Remote sensing data were acquired using AVIRIS on 15 June 1992 at Harvard Forest
and 21 June 1992 at Blackhawk Island. AVIRIS records data in 224 contiguous spectral
bands covering the spectral range of 400 - 2500nm with a spectral resolution of \0mn.
The spatial resolution of this data is 20m [4].
Atmospheric corrections of the AVIRIS data were done by the ATREM program [5, 6].
This program uses information in each AVIRIS radiance spectra to parameterize a radia
tive transfer model that is then used to convert radiance data to ground reflectance by
removing the effects of atmospheric gases, water vapor and aerosols. In addition to the
ATREM correction, a secondary correction was made based on the difference between a
ground calibrated Blackhawk Island scene and the same ATREM corrected scene [7, pers
comm. K. Heidebrecht, R. Clark].
The atmospherically corrected AVIRIS data was transformed to first difference spectra.
This transformation serves two functions: 1) to remove the effects of baseline shifts and
2) to discriminate between overlapping bands [8, 9]. The mean of the first difference
spectra from a 2x2 array of pixels overlapping each of the 40 field locations was used in
this analysis.
3 RESULTS
Plot data from both Harvard Forest and Blackhawk Island were combined for final lignin
and nitrogen calibrations. The data for each plot were randomly assigned to either a
calibration or validation sample set. Calibration plots were used in the regression calcu
lations with validation plots remaining separate to evaluate predictions. Canopy nitrogen
concentrations were predicted from first difference AVIRIS spectra using equation 1:
%Nitrogen = 0.486 -f (0.001 * 783 nm) + (0.003 * 1640nm) (1)
The relationship between field measured and AVIRIS predicted nitrogen for calibration
and validation plots is shown in figure 1. Absorption at 764nm corresponds with both a
third overtone N-H absorption feature and a chlorophyll absorption feature. Chlorophyll
content in foliage is highly correlated with total protein, and hence total nitrogen, content.
1640nm is a first overtone of a C-H absorption band. This equation is then applied to each
202