The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
chlorophyll concentration is restricted to the visible region and
the red edge, with greatest SD (2.69) around 710 nm. Next, to
gain insights in the dynamics of the chlorophyll-generated
variability, the Coefficient of Variation (CV) was calculated:
CV.«=-, (1)
The CV Cab is a useful statistic for comparing the degree of
dispersion, from one data series to another, even if the means
are drastically different from each other. Figure lb shows the
CV C ab of th e PROSPECT-generated spectra with varying
chlorophyll content. The derived CV Cab is characterized by a
peak in the red edge at 710 nm, which is the response of
chlorophyll a absorption.
wavelength [nm] wavelength [nm]
Figure 1. a) Averaged reflectance of a PROSPECT-simulated
needle, understory and bark representing the spectral properties
of PV, background and NPV relevant for FLIGHT
parameterization. The grey band represents the interpolated SD
related to the Cab range, b) associated CVCab
2.4 Generation of FLIGHT data set
Variable
Unit
Generic
field
observations
Range of variation
LUT
Min Max ste
P
Within-crown
NPV-PV 3
%
0.7
0.2
1.0
0.1
Crown LAI
-
2.5
1
10
0.5
b
CC
%
0.6
0.2
0.80
0.1
LAD
Spherical
Stand
structure
Tree height
m
11.93 ±2.9
Crown radius
m
0.88
Trunk height
m
7.0
Trunk
m
0.179 (at
diameter
ground)
Table 1. Ranges of within-crown structural variables for
generation of LUT and field observations of stand variables
relevant for FLIGHT parameterization (a: NPV=1-PV; b: 0.5
until LAI: 5 then steps ofl.O.). LAD: leaf angle distribution.
Having introduced spectral variability at needle level, the
analysis shifts to canopy level through coupling with FLIGHT.
Automatic simulations were realized based on Look Up Tables
(LUT). Prior to configuring the look-up tables, it is of
importance that the relationship between confounding factors
(e.g. structure, woody elements) and chlorophyll content is
established for any given canopy structure or composition that
may occur during a development phase. Key biophysical
components that vary throughout stand development were
selected, being canopy LAI, crown coverage (CC) and, to
accommodate for a varying canopy composition, within-crown
NPV and PV proportions. We used collected stand architectural
data (e.g., trunk height, tree height, trunk radius, and crown
radius) from the SPREAD campaign to parameterize FLIGHT.
Their major characteristics are summarized in table 1. The
simulated stands were horizontally distributed on a flat terrain
according to a Poisson distribution with crowns of irregular
conical shape and cylindrical trunks. Within the individual
crows a spherical leaf angle distribution of the phytoelements
was assumed. Additional parameters were fixed to model
default or field measurement as described in Kôtz et al. (2004).
Canopy reflectance was simulated as observed from nadir in 18
spectral bands corresponding with specifications of CHRIS in
Mode 3 (land). CHRIS aboard PROBA (Project for On-board
Autonomy) satellite is an experimental smallsat that has the
capability to provide combined hyperspectral and
multidirectional sampling with high spatial resolution (~17 m)
of selected terrestrial targets (Barnsley et al., 2004). In the
‘land’ mode wavebands are selected specifically to monitor
vegetation cover. The spatial size of a CHRIS pixel precisely
matched to the FLIGHT scene dimensions (17 m). The solar
angle was set as during CHRIS-PROBA overpass over Swiss
National Park on 2004-06-27 (9 S : 24.0°, <f> s \ 162.8°, see Verrelst
et al., 2008 for details). All spectral data were convolved to
CHRIS using the CHRIS band centers and full-width-half-
maxima (FWHM).
To make sure that explicitly relationships between reflectance
spectra and chlorophyll content are studied with no other
influencing factors than canopy structure and woody elements,
only the vegetation signal free of any contamination is needed.
To achieve uncontaminated reflectance spectra, a solution is to
apply a total correction for background signal by setting the
background layer equivalent to a perfect absorber canopy
background condition. In this case the optical characteristics
and associated Cab range of the experimental overstory canopy
signal was directly linked with biophysical parameters, without
any atmosphere or background contamination. Once the
initialization was done, one million rays penetrated in each
experimental canopy. Finally a total of 7938 forest scene
simulations (9 Cab x 14 crown LAI x 9 NPV-PV x 7CC)
executed by PROFLIGHT provided the spectral sampling for
the subsequent analysis of the contribution of woody elements
and needle Cab content at stand level.
3. RESULTS
In FLIGHT, the radiant fluxes interacted with the woody
elements and the chlorophyll-containing foliage which resulted
into upscaled chlorophyll-induced spectral response. The
averaged BRF ( BRF ) and chlorophyll-induced spectral
dispersion (CV C ab) was once more calculated, now at stand level.
Figure 2a and 2b shows the averaged BRF and associated CV Cab
for the CHRIS bands and the SNP structural configurations
(table 1). With a perfect absorbing background, only the
scattering caused by the overstory canopy attributes that
escapes into nadir direction are detected. This assures an