The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
28
explicit connection with the structural configurations but, in
turn, also leads to low BRF . Considering the CV Cab , a shape
similar to leaf level appeared but then with a markedly higher
magnitude (around 35-40%), which can be fully explained by
absence of any other confounding factor.
Figure 2. a) BRF with interpolated Ca6-induced SD (grey
band), b) associated CV Ca b at canopy level as simulated by
FLIGHT (crown LAI: 2.5; PV: 70%; CC: 60%).
To exploit the propagated variability along gradients of canopy
variables, we chose one specific wavelength where the subtle
Cab signal is prominently expressed. Wavelengths associated
with chlorophyll content are to be found in the 550 or 700 nm
regions, where higher contents of chlorophyll b and chlorophyll
a, respectively, are required to saturate the absorptance
(Thomas & Gausman, 1977). CHRIS wavebands associated
with these chlorophyll regions are A^: 530, 551, 570 and 697,
703, 709 nm respectively. For each of these wavebands the
CV C ab were plotted along the NPV-PV gradient. The gradient
starts from 0% NPV onwards whereby the proportion of canopy
woody elements (NPV) gradually increases (figure 3). From
this set of chlorophyll-sensitive wavebands, the A^ 531 nm
band exhibited greatest decrease all along the NPV gradient
implying greatest sensitivity to woody elements, while the A^
709 nm band gave sign of a flatter response implying certain
robustness to woody elements. Since apart of chlorophyll,
anthocyanin also absorbs around 550 nm (Sims et al., 2002), we
opted to continue working with the radiant flux at 709 nm.
Figure 3. Waveband-dependent trends of CV Cab (%) along the
NPV gradient (%PV= 100-%NPV). Values were averaged over
the gradients of crown LAI and crown closure.
3.1 Detectability of chlorophyll content
The applicability of observed trends in the context of real-world
canopies is analyzed next. As pointed out above, woody parts
can significantly govern the canopy cover of coniferous forests.
Being aware of varying NPV fractions at sub-pixel level the
underlying idea of this modelling exercise is that modelling
results may provide canopy-specific indicators regarding the
suitability of retrieving chlorophyll content from reflectance
data. The Standard deviation (SD) indicates the absolute spread
of the reflectance spectra. Due to the fixed Cab range the
magnitude of the SD is explicitly governed by structural
attributes. Consequently, given the assumption that a greater
spectral spread allows easier detection and mapping of
chlorophyll content, the SD results provides thus a theoretical
indicator of the stand-dependent chlorophyll content
detectability. To exemplify this, four structurally distinct real-
world forests were selected (figure 4).
Each of the coniferous stands holds near-optimal conditions to
retrieve chlorophyll content in terms of CC and crown LAI with
the exception of the heavily affected spruce stand in the
Sumava Mountains. They encompass a CC of 60% or higher
and a varying crown LAI. The homogenous, dense young
spruce stand (figure 4a), for example, is characterized by high
crown LAI value (7.8) which results into a slight suppressed SD
as a consequence of the LAI-dependent chlorophyll absorptions.
The other test sites encompass a lower LAI which implies a
slightly greater spectral spread and thus detectability.
When positioning the four stands along the NPV-PV gradient,
however, only then the true Cab detection feasibility appears.
The virtual absence of woody cover in the outer canopy of the
young spruce stand implies that the full Cab-related variability
can reach a sensor without any contamination. Canopy structure
such as the high LAI would be the limiting factor here, but only
to a minor extent (figure 4e). By contrast, the old-growth forest
in SNP gave sign of a more complicated canopy structure and
composition (figure 4b). As a consequence of dead and partly
dead trees a fractional NPV cover may reach up to 30% at
CHRIS pixel resolution. Due to the scattered woody parts
within the canopy, PV and NPV of this stand are
heterogeneously arranged in a quasi-random manner. The
modeling results indicate that such woody contribution may
suppress the chlorophyll-dependent spectral spread with 28%
given the SNP structural configurations (CC: 0.60; LAI: 2.5).
However, note that the canopy throughout the test site is rather
open meaning that in reality understory, which is again a
mixture of PV and NPV, will partly contribute to the
chlorophyll signal as well.
Contrarily to the latter, the infested stands in British Columbia
(figure 4c) and in the Bohemian Forest (figure 4d) are from
another order of woodiness. The British Columbian stand shows
mixtures of green trees, red-attack trees and grey-attack trees
throughout the forest (figure 4d). Subsequently, the spatial
distribution of NPV fractional cover varies dramatically. Such
spatial PV-NPV variation implies that relationships between
reflectance spectra and the needle pigment concentrations of
remaining foliage are perturbed in a pixel-specific way. The
presented results of the rapid declining SD along the NPV-PV
gradient indicate the magnitude of change (figure 4f, g). Finally,
the example of the Bohemian Forest reflects the worst case
situation: virtually the complete stand suffers from insect
defoliation. Only small patches and a few isolated trees are left
unaffected. What is left of the signal from remaining green
cover is in consequence heavily suppressed by dominating NPV
cover. No structural assessments have been carried out in this
region. With a visually assessed of remaining 40% PV and 50%
CC the modeling results indicated that only 58% of the full Cab
spread is left over.
4. CONCLUSIONS
Natural forests go through development stages with increasing
heterogeneity in the distribution of foliage and woody parts.
The direct observable effects of these trajectories are that
structural variables such as fractional cover, LAI and the
amount and arrangement of woody elements alter over time.
For instance, the old-growth forest of Swiss National Park that