2.2 Acquisition of satellite thermal images and atmospheric radiosoundings 
1.0 
0.8 
0.6 
0.4 
0.2 
0.0 
14 
^METRICS and 
iare soil. 
Six cloudless images have been obtained: 2 LANDSAT5-TM images (20 July and 5 August), 2 ERS1-ATSR 
images and 2 NOAA11-AVHRR images (24 and 27 July). On ATSR and AVHRR images, the test site is 
fortunately situated or nearly situated at the centre of the image and the viewing angle effect can be neglected. 
We have collected the standard meteorological radiosounding data (performed at 12h U.T. at Nimes, 20 km N.W 
of the test site) corresponding to the dates of the images acquisition. In order to avoid the non-coincidence of the 
radiosounding and the satellite overpass times, 3 special radiosoundings were performed on 27 July at the time of 
ERS1-ATSR overpass and on 20 July and 5 August at the LANDSAT5-TM overpass times. 
3 ATMOSPHERIC CORRECTION 
Split-Window methods 
We have used 10 sets of published Split-Window coefficients (Table 1). Their references are from the RAL 
(Rutherford Appleton Laboratory, 1993, personal communication, denoted RAL93), Li et al. (1993, denoted 
Li93), Becker and Li (1990, denoted Becker90), Ottld et al. (1992, denoted Ottl692), Kerr (1992, denoted 
Kerr92), Price (1984, denoted Price84), Deschamps et al. (1980, denoted Deschamps80), Ulivieri et al. 1985, 
1992 (denoted Ulivieri85 and Ulivieri92 respectively) and the NESDIS (1992, denoted NESDIS92;) described in 
the paper of Li et al. (1993). As our ground measurements were performed in the 9.8-11.4 pm band, and as we 
will just compare the ground level brightness temperature estimated from satellite data with filed measurements, 
the emissivity can be set to 1 . 0 . 
Table 1: Ten sets of Si 
ni it-Window coefficients tested (Ts = A + B T4 + C T5) 
Authors 
A 
B 
c 
Sensors considered 
RAL93 
-1.652 
3.677 
-2.671 
ERS1-ATSR 
Li93 
-0.226 
3.630 
-2.630 
NOAA11-AVHRR 
Becker90 
1.274 
3.630 
-2.630 
NOAA9-AVHRR 
Ottlé92 
0.858 
3.218 
-2.218 
NOAA9-AVHRR 
Kerr92 
3.100 
3.100 
- 2.100 
NOAA9-AVHRR 
Price84 
0.000 
4.300 
-3.300 
NOAA7-AVHRR 
Deschamps80 
- 2.200 
3.600 
-2.600 
NO AA7 -AVHRR 
NESDIS92 
-0.155 
3.673 
-2.657 
NOAA9-AVHRR 
Ulivieri85 
-0.880 
4.000 
-3.000 
NOAA7-AVHRR 
Ulivieri92 
0.000 
2.800 
-1.800 
NOAA 11-AVHRR 
of the camera 
Use of LowtranJ code 
Correction using the LOWTRAN 7 radiative transfer code were achieved by iterations. With the same input 
radiosounding data, we adjust the input ground temperature to get the same brightness temperature derived from 
satellite measurements. For this method, we have used the 7 radiosounding data acquired by the Nîmes 
meteorological station and the 6 standard atmospheric models included in the Lowtran7 code. 
New method using the measurements in 2 channels (ATSR and AVHRR) 
A Split-Window method uses generally the same coefficients independently of the atmospheric conditions. This 
will produce some errors when the real atmospheric condition is quite different from the average atmospheric 
condition in which the Split-Window coefficients were obtained. However, the spectral difference of brightness 
temperatures measured in the two channels equivalent to AVHRR4 and AVHRR5 of NOAA sensors, can be 
directly related to the real atmospheric condition. It is the reason why we have developed a new method 
combining the Split-Window basic idea and a radiative transfer code (Lowtran7). 
The procedure used is the following: 
(1) Calculation of the two ground-level brightness temperatures (Ts(4), Ts(5)) by using 
respectively the two satellite measurements (channel 4 and 5 for AVHRR image) with help of 
Lowtran7 code with each of the 6 standard atmospheric models included in the Lowtran7 code;