space. Corrected 
et al. 1994) or in 
ditional data sets 
5) or new sensors 
onal reflectances 
siting techniques 
is not well suited 
f system features 
;ing performance 
level corrections 
ition by man and 
unmission of the 
remotely sensed 
>C, can be found 
! measurements: 
th AVHRR data 
teric water vapor 
ence Vegetation 
s by atmospheric 
Administration, 
A from NOAA- 
f the total ozone 
Threshold cm Irt, 
NDVI time series 
1 scale of C02 
t sensing: some 
onal papers. 14, 
ction of Satellite 
lamps, 1990: A 
> code. Intern. J. 
NOAA-AVHRR 
t semi-analytical 
MONITORING NOAA/AVHRR AND METEOSAT SHORTWAVE BANDS CALIBRATION 
AND INTER CALIBRATION OVER STABLE AREAS 
Cabot F., Dedieu G. and P. Maisongrande 
LERTS, Unité mixte CNES-CNRS, 18, Avenue Edouard Be lin, 31055 TOULOUSE CEDEX, FRANCE 
Abstract 
Monitoring of the calibration of shortwave sensors onboard NOAA and METEOSAT satellites has been the topic of 
various studies, involving atmospheric scattering, ocean glitter or the monitoring of radiometric ally stable targets. 
Nevertheless, most of the developed methods, principally those based on de sertie areas, suffer diverse constraints, 
particularly the need for reproducible geometric conditions. These requirements limit the possibility of a continuous 
monitoring of the gain drift of these instruments. 
The method that we developed to achieve this aim relies cm the difference in the frequency of all the phenomenons that 
contribute to the measured signal. Atmospheric effects present strong daily variations superimposed on seasonal 
evolution. Directional effects depend on orbital characteristics of the platform, instrumental characteristics and also on 
the season. At last, one can assume that the temporal evolution of the calibration of the instruments does not present any 
cyclical behavior and obey a continuous law. These considerations led us to develop a method coupling models 
describing atmospheric and directional effects. This method is particularly adapted to the monitoring of the calibration 
over long term period, using desertic stable areas. The model parameters are prescribed from ancillary data sets (e.g. 
atmospheric water vapor content) or fitted with satellite measurements. 
This method has been applied to several years of NOAA/AVHRR and METEOSAT measurements over several sites, 
considering each sensor separately in a first time and then altogether to attempt inter calibration of the two sensors. 1 
1. INTRODUCTION 
Throughout the last years, growing consideration has been given, in use of remote sensing data, to absolute calibration 
problems and gain drift monitoring. Although most of the satellites bear on board calibration devices, their use, 
especially in shortwave bands, remains difficult because of their uncontrolled ageing. 
In order to solve such a problem, some experiments are periodically held to quantify the decaying of the sensors and to 
get a realistic gain value. These experiments suffer of low repeatability (presently one is held every 6 mo nths 
approximately), atmospheric contamination, directional effects and of the very few sites being used. 
The atmospheric and directional effects that influence the signal are difficult to take into account and the variations due 
effectively to the gain drift are generally of the same order of magnitude. The only difference in the behaviors of all these 
effects is their time period. Atmospheric effects present a yearly repeatability with strong random daily variations, effects 
of the non-lambertian behavior of the surface are ruled by the orbit repeatability and season whereas gain drift is 
expected to follow quite smooth variations with no repeatability all along the satellite life. 
The method presented here takes advantage of these remarks, and is aimed at separating the variations according to then- 
time period, and when possible, correcting them to isolate the variations due to gain drift. Our objective is to 
continuously monitor the variation of the gain of the sensors using large homogeneous areas as targets. Several sites have 
been selected and observed through 4 years. Their bidirectional behavior has been studied as well as their temporal 
stability. Some of these targets seem appropriate and experiments with several satellites were held. An attempt to 
complete an intercalibration of the different sensors used was also made. 
2. DATA SET 
The data set we used was extracted from the Global Vegetation Index product provided by the National Oceanic and 
Atmospheric Administration (Kidwell 1990). This archive includes, since 1985, digital count for visible and near 
infrared channels, brightness temperatures for thermal channels 4 and 5 and geometric conditions for NOAA/AVHRR 
satellites, subsampled to a spatial resolution of 16 km on a weekly basis. We choose to focus cm NOAA/AVHRR-11. 
launched in September 1988 and still in operation at this day. 
41