International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
2.3 Spectral indices 
Several spectral indices have been reported in the literature and 
proven to be well correlated with vegetation biophysical 
parameters such as LAI, and biomass. Tremendous efforts have 
been devoted to improve vegetation indices and render them 
insensitive to variations in illumination conditions, observing 
geometry, and soil properties. A few studies have been carried 
out to assess and compare various vegetation indices in terms of 
their stability and their prediction capability for LAI (Baret and 
Guyot, 1991; Broge and Leblanc, 2000; Haboudane et al., 
2004). Other research has dealt with modifying some vegetation 
indices to improve their linearity with, and increase their 
sensitivity to LAI (Nemani et al., 1993; Chen, 1996; Brown ef 
al., 2000). Consequently, some indices have been identified as 
best estimators of LAI because they are less sensitive to the 
variation of external parameters affecting the spectral 
reflectance of the canopy namely soil optical properties, and 
atmospheric conditions (Broge and Leblanc, 2000), as well as to 
changes of leaf intrinsic properties such as chlorophyll 
concentration (Haboudane et al., 2004). Based on these studies 
we have selected three spectral indices briefly presented below. 
Normalized difference vegetation index NDVI (Rouse et al., 
1974) 
(NIR—R)/(NIR+R) (1) 
Modified second soil-adjusted vegetation index MSAVI2 (Qi et 
al., 1994) 
  
Lox nimi Jo NR -8* (IR-R) ©) 
Modified second triangular vegetation index MTVI2 
(Haboudane et al., 2004) 
1.5*[1.2*(NIR-G) -2.5*(R-G)] (3) 
  
J(2* NIR-1y —(6* NIR-5*4R) - 0.5 
A detailed discussion on these indices can be found in Broge 
and Leblanc (2000) and Haboudane et al. (2004). In the 
formulae G, R and NIR denote canopy reflectance in the green 
(550 nm), red (670 nm), and near-infrared (800 nm), 
respectively. 
Performance evaluation of these indices was based on 
measured canopy spectra corresponding to a wide range of LAI 
(0.13 to 6.50) measured for three crop types (corn, beans, and 
peas), under different conditions (four different years); thus, 
empirical relationships between LAI and the indices were 
determined. Their prediction capacity for estimation of LAI was 
then assessed using CASI  hyperspectral images and 
corresponding ground truth for LAI from three other crop types 
(wheat, soybean, and corn). 
110 
3. RESULTS AND ANALYSIS 
3.1 Spectral Indices Behaviour 
The main difference between these three indices resides in the 
saturation effect when LAI increases: NDVI reaches 4 
saturation level asymptotically when LAI exceeds 2, while 
MSAVI2 and MTVI2 show a better trend without a clear 
saturation at high LAI levels (up to 6). This explains, in part, 
why MSAVI2 has proven to be a better greenness measure 
(Borge and Leblanc, 2000). To illustrate this effect, we plotted 
the indices against the near-infrared (NIR) reflectance (Figure 
1) to compare each index's ability to depict LAI variations. The 
rationale for this analysis was that the NIR reflectance is 
strongly affected by changes in vegetation structural descriptors 
rather than by pigment variation. 
Distinct behaviours are observable in Figure 1: NDVI 
became saturated when NIR reflectance exceeded 0.35 to 0.40 
depending on the crop, while MSAVI2 and MTVI2 appear to be 
much more responsive to NIR reflectance increase. Indeed, 
MSAVI and MTVI2 showed only a slope change when NIR 
reflectance reached 0.40 for corn, and 0.55 for beans and peas. 
Moreover, NDVI seemed to be strongly affected by the red 
reflectance; thus, both red reflectance and NDVI approached 
asymptotic values, and showed virtually no further change, 
when NIR reflectance has exceeded 0.35 to 0.40. Conversely, 
NIR reflectance, which continued to increase substantially, 
induced virtually no change in NDVI trend. In contrast, MTVI2 
and MSAVI2 appeared to be more sensitive to NIR reflectance 
increase, but less affected by the lack of dynamic in the red 
reflectance. 
These distinctive behaviours are illustrated by indices 
values scattering in response to red reflectance variability. 
While MSAVI2 is the less sensitive to this effect, NDVI 
showed the highest level of scattering particularly in the case of 
corn and peas canopies (Figure 1 a & c). As for the dynamic 
range, MTVI2 exhibited more sensitivity to NIR reflectance 
than the other indices. 
  
  
  
0.90 
W Red 
0.80 0 NDVI 
^ MSAVI 
« 0:70 9 MTVI2 
9 
ë 
= 060 
= a 
$ 0.50 
un 
= 
2 0.40 
a 
3 
5 0.30 
A 
= e 
& 0.20 : 
0.10 oi? " 
A nu 
T p" agio n? 
0.00 
0.10 0.20 0.30 0.40 0.50 0.60 0.70 
  
  
  
NIR reflectance 
  
International 
es 
1.00 
0.90 
0.80 - 
0.70 
0.60 
0.50 
0.40 
0.30 
Red reflectance / Spectral Indices 
0.20 
910 dr 
  
0.00 
[+] 0.10 
0.90 - 
0.80 
0.70 
UA 2 
c © 
c 
E 
© 
Red Reflectance / Spectral Indices 
  
0.1 
Figure 1. 
MTVI2 to 
(Corn (a), 
of 2003. 
3.2 Relat 
The relati 
unique; t 
chlorophy 
canopy cl 
measure ' 
plays a me 
Our re 
equations 
type-depe 
was found 
for corn, | 
present th 
and meast