25
ig 2008
A MODELING APPROACH FOR STUDYING FOREST CHLOROPHYLL CONTENT IN
RELATION TO CANOPY COMPOSITION
J. Verrelst 3 ’*, M.E. Schaepman 3 , J.G.P.W. Clevers 3
a Centre for Geo-Information, Wageningen UR, Wageningen, The Netherlands - (Jochem.Verrelst, Michael.Schaepman,
Jan.Clevers)@wur.nl
Commission VII, WG VII/1
KEY WORDS: Chlorophyll Content, Non-Photosynthetic Vegetation, PROSPECT, FLIGHT, Canopy Structure, Woody Elements
ABSTRACT:
Foliar concentration of the main photosynthetic pigments chlorophyll a and b (Cab) is widely regarded as a generic bioindicator of
the actual plant status. However, when the scale moves up to stand level, relationships between the spectral response and leaf
chemistry tend to break down due to confounding factors such as canopy structure, woody elements and background contributions.
Especially in old-growth forests large numbers of standing and fallen dead wood are generated. We questioned the role of woody
elements in the retrieval of Cab content on the basis of synthetic reflectance data through coupling of leaf-level (PROSPECT) and
canopy-level (FLIGHT) radiative transfer models. For a wide range of forest stands the Ca6-induced dispersion (Coefficient of
Variation: CV Cab ) and total spread (Standard Deviation: SD) was calculated. The magnitude of CV Cab and SD provides information
about the Cab-related spectral spread and can therefore be regarded as stand-specific indicators of the theoretical Cab detectability.
Results demonstrate that in dense canopies woody elements are key players in suppressing the Cab-related spectral spread. Apart
from composition canopy structure also exerts influence: e.g. an overstory with crown coverage (CC) of 60% and a crown LAI of 1.5
propagated greatest spectral spread. In sparse stands (e.g. CC<40%) the background contribution is the dominant confounding factor.
The impact that woody elements exert in the theoretical retrieval of Cab content was quantified for four distinct real-world
coniferous forest types.
1. INTRODUCTION
Foliar concentration of the main photosynthetic pigments
chlorophyll a and b (Cab) is widely regarded as a generic
bioindicator of the actual plant status, such as stress condition
(Lichtenthaler et al., 1996) and vegetation gross primary
productivity (Gitelson et al., 2006). Various leaf and canopy
experiments have indicated that imaging spectroscopy is a
potentially powerful tool for assessing variation in chlorophyll
content of trees (Ustin et al., 2004). However, when the scale
moves up to stand or landscape level, relationships between the
reflected electromagnetic radiation and leaf chemistry tend to
break down (e.g. Trotter et al., 2002). Then the subtle scattering
and absorption properties of the foliar chemistry are
confounded by whole-tree features such as the foliage structural
arrangement, woody elements and background reflectances (e.g.
Asner, 1998).
At landscape level, a common approach to deal with sub-pixel
complexity is to unmix a pixel into its most distinct, ‘p ure ’
endmembers. For instance, a vegetated surface might be
decomposed into fractions of photosynthetic vegetation (PV),
non-photosynthetic vegetation (NPV) and bare soil. Although
pixel unmixing into NPV-PV-bare soil endmembers facilitated
to study ecosystem dynamics (e.g. Asner et al., 2004; Jia et al.,
2006), it does not provide understanding of the true complexity
between the interaction of phytoelements and radiant energy.
Alternatively, leaf chemistry estimates retrieved from optical
remote sensing can be investigated using radiative transfer
models, which describe the transfer and interaction of solar
radiation inside the canopy based on physical laws and thus
* Corresponding author.
provide an explicit connection between the phytoelements and
the canopy reflectance. While much work exists in the realm of
radiate transfer modeling, the relative importance of woody
elements (NPV) in the context of quantifying canopy
chlorophyll content however has not adequately been evaluated.
Only recently the influence of the 3D structure of trunks and
branches in a coniferous canopy on reflectance has been
explicitly quantified (Malenovsky et al., 2008). Yet this work
comprised a young production forest (e.g. <30 year), where
woody elements are exclusively part of the standing trees and
concentrated in the lower part of the canopy. By contrast, in
old-growth forests a surplus of woody parts in the form of lying
and standing deadwood are scattered within the canopy layer
and on the forest floor and can encompass 18-40% of total
woody biomass (Siitonen 2001). In these older forests woody
parts play a significant role in determining canopy reflectance
(Asner, 1998), as they are an important photon absorbing and
scattering component. Apart from changing canopy
composition, the forest also increases in heterogeneity during
aging (Franklin et al., 2002); making structural attributes
equally important drivers of the canopy spectral response (Song
et al., 2002). This paper reports on the propagation of canopy
compositional and structural effects when inferring chlorophyll
content assessments on the basis of synthetic reflectance data.
We used the leaf-level model PROSPECT (Jacquemoud et al.,
1996) coupled with the canopy-level ray-tracing model
FLIGHT (North, 1996). The coupled models allow controlling
simultaneously foliar chemistry and biophysical variables. The
objective of the present study was to assess the influence of
NPV and forest structure on the determination of leaf
chlorophyll concentration from synthetic reflectance