International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
relative error is 11 %, 9 % and 5 % for AVHRR, 
VEGETATION and MODIS, respectively. However, RDVI, 
OSAVI and SAVI indices show a similar sensitivity in the 
bands of the same sensor. Indeed, relative error is 9% for 
AVHRR, 7 % for VEGETATION and 4 % for MODIS. 
With the difference of the four preceding indices, the NDVI 
shows a similar relative error between AVHRR and 
VEGETATION. In the extreme ozone concentration and a 
very dense cover, the relative error on the NDVI is 8 %. If we 
consider the same conditions, we show that the MODIS bands 
normalize better the ozone absorption effect on this index; the 
relative error does not exceed 4%. 
The indices developed to correct the atmospheric effects 
(ARVI and TSARVI) normalize correctly the variations of the 
ozone concentration in the atmosphere. In the extreme ozone 
concentration (0,959 g/cm3) and a very dense forest cover, the 
relative error on the ARVI (y = 0.5) does not exceed 7 % 
independently from the sensor characteristics. In the same 
condition, the relative error on the TSARVI (y = 0.5) does not 
exceed 5 % and 3 %, respectively, for VEGETATION and 
MODIS. We can conclude that the TSARVI index is the least 
affected by the atmospheric effects. Also, GEMI index has a 
fow sensitivity to the ozone absorption effect. For a dense 
cover and strong ozone concentration in the atmosphere, the 
50 
45 | 
40 | 
MSR 
Relative Mean Error ( % ) 
] 
tr 
< 
  
  
0 i LIZE 
relative error on this index is 6 %, 2 % and 1% for AVHRR, 
VEGETATION and MODIS, respectively. According to 
Asalhi (2003) and Bannari er al. (2000), the GEMI is less 
sensitive to the water vapour absorption and resistant to the 
aerosols diffusion. 
4. Conclusions 
Figure 1 summarizes the impact of a strong ozone 
concentration (0,959 g/cm3) on different vegetation indices 
considering a very dense forest cover (100 %). Except for the 
EVI, MSR and GARI, all used vegetation indices are 
characterized in general by a good resistance to the ozone 
absorption particularly in MODIS channels. This sensor 
seems to avoid perfectly the ozone absorption bands. 
Furthermore, the relative error is higher for the AVHRR 
sensor and relatively higher for the VEGETATION sensor. In 
the broadband AVHRR sensor, the relative errors vary 
between 6 % and 23% for the most resistant (GEMI) and the 
most sensitive index (MSR). Also, this figure underlines the 
perfect resistance of the TSARVI and GEMI to the ozone 
absorption considering the MODIS and VEGETATION 
Sensors. 
O AVHRR 
EJ] VEGETATION 
E] MODIS 
  
  
Figure 1: Sensitivity of different vegetation indices to ozone concentration (0,959 g/cm3) considering a dense forest cover (100 %). 
Acknowledgment 
The authors would like to thank the Natural Sciences and 
Engineering Research Council of Canada (NSERC) for 
the financial support. 
References 
Asalhi, H., 2003, Analyse de la sensibilité des indices de 
végétation au-dessus d'un couvert forestier de sapin : 
étude comparative à partir des données de simulations 
804 
entre MODIS-EOS, VEGETATION-SPOT et AVHRR- 
NOAA. Thèse de maîtrise, Département de géographie, 
Université d’Ottawa, Ottawa, Ontario, Canada, 153 
pages. 
Bannari, A, Teillet, P.M., Leckie, D.G. et Fedosejevs, G., 2000, 
Impact des conditions internes et externes aux couverts 
forestiers sur les indices spectraux dérivés de AVHRR. 
La revue Télédétection, vol. 1, n°. 3, p. 157-181. 
Bannari, A, Royer, A. et Morin, D., 1997, L'indice de 
végétation TSARVI : Transformed soil atmospherically 
Inte) 
Banı 
Bare 
Chet 
Desc 
Jack 
Kau 
Gite