on which sensor causes this effect. 
1000 1200 
>t over Brazil, for 
1000 1200 
ept for AVHRR 
ot show any cyclical 
that the two data sets 
ization of vegetation 
y a reduction in the 
On the other hand, 
it presents only one 
>tted in figure 8, are 
e have no indication 
over Libya, for METEOSAT derived over Libya, for METEOSAT, using AVHRR 
coefficient. 
S. CONCLUSION 
In this study we developed a method aimed at monitoring the evolution of the calibration factor of a sensor. This method 
relies on the analysis erf long term series of reflectance measured over radiometrically stable areas. Though preliminary, 
this work has outlined that, if we are able to account for all the effects that perturbate the satellite signal, we can extract 
from the reflectances variations an information about the decaying of the sensor. Since the amplitude of the variation in 
reflectance due to this decaying is generally small, at least of the same amplitude as other perturbating effects 
(atmosphere, directional), it appears rather difficult to isolate what is due to the drift. The only difference that appears 
between all these effects being their time period, one has to take advantage of it This was done by accounting for 
atmospheric effects using a climatology, supposed to remove from the signal the seasonal variations due to atmosphere, 
then by correcting directional effects, for which the period is both seasonal and short term (orbit repeatability). After 
these various corrections, the obtained s ignal can be used to derive the variations of the calibration factor of the sensor. 
This has been done over several sites, representative of various surfaces, latitudes and climates. The observed variations 
are always of the same amplitude as those observed by other authors, and show similar general behavior, even if large 
differences appear on certain sites. These discrepancies may be partly due to the large difference in the level of the 
reflectances: e.g. the Australian site reflectance is half worth the one of the Libyan site. Therefore, the contribution of the 
atmospheric scattering is more important, may induce more variability and hamper the retrieval of model parameters. 
Still, the point that appears when evaluating the result of these computations is that this method allows to identify the 
directional behavior of the surface as well as the temporal evolution of the sensor. As it could be expected, brighter 
targets seem more appropriate for the application of this method since the effects of aerosol scattering is minimized over 
such areas. But the more important appears to be the homogeneity at a large scale of the selected targets. This allows to 
overcome misregistration problems and a more efficient filtering. The needed accuracy in all the preprocessings is such 
that it can hamper badly the retrieval of all the parameters and lead to rather large unaccuracy in the evolution of the 
calibration factor. Still, at this time, this accuracy is difficult to measure. It also appeared that, METEOSAT being less 
sensitive to atmospheric effects, the monitoring of its gain is easier. 
After applying the described method to two different sensors (AVHRR and METEOSAT) we attempted the 
intercalibration of the two instruments. This could be achieved but lead to a calibration factor for METEOSAT visible 
channel about 30% higher than the one derived by other authors. At this time, and due to the lack of information related 
to the selected target, we can not decide from which step of the procedure this difference is due. 
Despite all these uncertainties, the method appears able to derive consistent information useful in the normalization 
process. It is definite that a better knowledge of both the surface and the atmosphere, and daily, one km resolution, 
AVHRR data should lead to better results. These conclusions should be considered for the calibration of upcoming large 
field of view sensors. 
ACKNOWLEDGEMENTS 
This work is supported by the Progr amme National de Télédétection Spatiale, Région Midi-Pyrénées, CNES, CNRS and 
Ministère de 1 Enseignement Supérieur et de la Recherche. 
45