1109 
superior to both SAVI and MSAVI in the case of dark soils, it is always the worst of all the indices in the case 
of medium and bright soils. 
5-CONCLUSIONS 
A number of spectral vegetation indices have been proposed to analyse satellite remote sensing data. 
The objective of this work was to quantitatively evaluate the performance of a representative subset of these 
indices with respect to their capability to estimate the fractional vegetation cover. 
NDVI was selected because of its wide use in the literature. WDVI, SAVI and MSAVI were selected because 
they were designed specifically to be less sensitive than NDVI to soil color and brightness variations. GEMI 
was evaluated because it has been constructed to be less sensitive than NDVI to both atmospheric and soil 
effects. Objective criteria to evaluate the performance of these indices were set up and applied to these indices. 
Albedo values were computed with a coupled soil-plant canopy model, both at the surface and at the 
top of the atmosphere. Ten soil types and three continental atmospheres were considered, to represent a wide 
range of situations that could be expected in global applications. The results were computed over two ranges of 
fractional vegetation cover (0-50% and 50-100%), and reported in the form of signal to noise (S/N) ratios. The 
signal has consistently been taken as the variation in the index value over the specified range of fractional 
cover, while the noise has been computed from the uncertainty in index value resulting from the lack of a 
priori knowledge on the type of soil or composition of the atmosphere. Since an average atmospheric 
correction could be applied to account for the presence of a standard atmosphere (mostly Rayleigh scattering), 
the S/N at the top of a turbid atmosphere was computed on the basis of comparisons with albedos that would 
have been observed in the abscence of aerosols. 
It has been shown that, within the framework of the model assumptions made in this study, the widely 
used NDVI and the WDVI are generally the poorer predictors of the fractional vegetation cover than either of 
the newer indices : SAVI,MSAVI and GEMI. SAVI and MSAVI are best suited for laboratory and field studies 
because of their significantly reduced sensitivity to soil type, while the GEMI is a good predictor of fractional 
vegetation cover, especially in the lower range of values and when the estimation of this parameter must be 
made on the basis of measuurements taken in space. 
It would be interesting to try to validate these results on the basis of field measurements, although this 
would be a very complex endeavor. This study has shown that it is possible to evaluate objectively the 
performance of vegetation indices with respect to the evaluation of a specific surface parameter, under a variety 
of soil and atmospheric conditions. To the extent that a single vegetation index is being applied globally and 
repetitively to address climatic and environmental issues, it is clear that the evaluation and use of advanced 
indices designed to be relatively insensitive to perturbations is most profitable. 
Acknowledgments: 
The authors are grateful to Dr. Y. Kaufman of NASA/GSFC for being the first to suggestthat the concept of 
signal to noise may be useful to evaluate vegetation indices. In their analysis, the authors have benefited from 
formal and informal discussions during a meeting organized at NASA GSFC in June 1992. M. Verstiaete is 
also acknowledging the continuing support of the Global Change research program of the Institute for Remote 
Sensing Applications, and in particular of its Exploratory Research Program. 
References 
Avissar, R. and M.M. Verstraete (1990): The representation of continental surface processes in mesoscale 
atmospheric models, Reviews of Geophysics, 28, 35-52. 
Baret F. & Guyot G., (1991): Potentials and limits of Vegetation Indices for LAI and APAR Assessment, 
Remote Sens. Environ., 35:161-173. 
Bowker D.E., Davis RE., Myrick D.L., Stacy K. & W.T. Jones (1985) : Spectral reflectance of natural 
targets for use in remote sensing studies, NASA reference publication 1139. 
Chehbouni A.,Kerr Y.H., Qi J., Huete A.R & Sorooshian S. (1993) : Toward the development of 
Multidirectional vegetation index, accepted in Water Res. Research. 
Clevers J.G.P.W. (1988): The application of a weighted infrared-red vegetation index for es timating leaf area 
index by correcting for soil moisture, Remote Sens. Environ. 25:53-69.