342 
chlorophyll concentration using model simulation: for very low concentrations the sensitivity is higher at the 
maximum absorption located around 675 nm. Conversely, for medium to high concentrations, leaf reflectance 
sensitivity to chlorophyll concentration is higher at 550 nm. Other empirical approaches are based on the use of 
several wavebands. For example, Chapelle et al. (1992) related the ratio between the reflectances in two 
wavebands to chlorophyll content. Spectral features in the red edge have also been widely studied and used to 
estimate chlorophyll content (Baret el al ., 1992; Filellaand Penuelas, 1994). 
Carotenoid concentration provide complementary information on canopy physiological status 
(Margalef, 1974). Carotenoid pigments protect the photosynthetic reaction centers from excess light. Carotenoid 
and chlorophyll concentration are generally coarsely related (Baret et al., 1988). However, increases of carotenoid 
relative concentration are often observed when plants are subjected to stress, and Sanger (1972) showed that 
carotenoids persist longer than chlorophyll a in senescing leaves. Thus, estimates of the ratio of carotenoids to 
chlorophyll a provide and interesting additional avenue for plant ecophysiology studies. Penuelas (1984a and b) 
and Penuelas et al. (1993a and b) show that the ratio of carotenoids to chlorophyll a generally rises in decaying 
plants and decreases in healthy plants. Carotenoids and chlorophyll pigments absorb concurrently in the 300 to 
500 nm spectral region (Margalef, 1974). It is therefore difficult to retrieve carotenoid concentration 
independently from chlorophyll concentration using reflectance measured at one given wavelength. However, 
unlike chlorophyll, carotenoids do not absorb strongly in the red. Penuelas et al. (1993a, b) studying sunflowers 
and aquatic plants, found that, both at the leaf and plant levels, the Normalized Difference Pigment index 
( NDPl=R red -R b i lu jR r ed + Rblue > was highly correlated with the ratio between total carotenoids and chlorophyll a. 
The aim of this paper is to investigate several approaches to retrieve the ratio between 
carotenoid and chlorophyll concentrations ( C c /C a ) from leaf reflectance measured in few wavebands. This may 
provide fast and easy ways to characterize the physiological and phenological status of complex ecosystems. We 
will test some empirical approaches and develop some theoretical background that will help designing an 
alternative semi-empirical approach. Particular attention will be paid to the optimal choice of wavelengths 
required to estimate C c lC a . This investigation is based on data sets collected along various experiments. 
2.BACKGROUND AND ALGORITHMS USED 
2.1. Background 
Leaf optical properties are governed by leaf surface and internal structural properties as well by concentration, 
distribution and intrinsic optical characteristics of leaf material. Leaf material optical property is characterized by 
the complex refraction index. The real part describes the scattering processes, while the complex part describes 
the absorption processes. Experimental results show that the real refraction index varies only slightly with 
wavelength. Therefore, Vanderbilt et al. (1993) showed that the specular reflection observed over leaf surfaces 
could be considered as independent on the wavelength at least in the visible and near infrared domain. Maas and 
Dunlap (1989) as well as Jacquemoud and Baret (1990) showed that reflectance and transmittance spectra of 
albino leaves (leaves deprived of photosynthetic pigments) were almost flat in the visible and near infrared 
domains. On the other hand, the complex refraction index that corresponds to the absorption coefficient depends 
strongly on wavelength. Generally absorption is described through a Beer's law, in which the absorption 
coefficient K is assumed to be: K=Sk l C i , where i refers to a particular biochemical component such as 
chlorophyll or carotenoid, ¿¿and C ( are respectively the corresponding specific absorption coefficient and 
concentration. Baret et al. (1988) showed that leaf reflectance could be approximated by the following semi 
empirical model: 
R=R s +S.exp(-Sk t C;) (1) 
R s is the reflectance value obtained for very high absorption coefficient value corresponding to high 
concentrations and high specific absorption coefficients. It is composed of 2 terms: (i) the specular component 
generated from simple scattering on leaf surface that is almost wavelength independent, and (ii) the residual 
multiple scattering that originated from leaf internal mesophyll when concentrations C i tend toward great values. 
This second component might depend on wavelength. However, results from Baret et al. (1988) suggested that 
this second component is small in the visible domain. The S parameter describes the structural effects on leaf 
reflectance: When there is no absorption (Sk/Oj-O), leaf reflectance R tends toward R s +S. It follows from 
equation (1) that the relationship between concentrations C, and reflectance at a given wavelength might change 
from one leaf to another according mainly to leaf surface and internal structure.