We have developed a model for determining the spectral difference Ae=e 4 -es from AVHRR data, which is 
described in detail in Coll et al. (1993b). The method is based on the temperature difference T 4 -T 5 , which is due 
to the differential atmospheric attenuation and the spectral variation of emissivity. These effects are coupled in 
the satellite images, so that it is not possible to separate them with AVHRR data only. The method proposed 
here requires a proper description of the atmosphere existing between the surface and the sensor, namely the 
vertical profiles of temperature and humidity, so that the water vapour differential absorption can be accurately 
assessed. These data can be available from radiosounding measurements coincident in space and time to the 
satellite overpass. An advantage of this method is that the emissivity difference is obtained at the satellite spatial 
scale, and with the AVHRR spectral responses. The method proposed here can be applied to the different natural 
surfaces of interest in remote sensing (agricultural areas, forests, different types of soils) when radiosoundings are 
available. Thus a spectral characterization of these surfaces can be obtained at different seasons, which can be 
further applied in a routine mapping of LST by means of the split-window method, without the aid of 
radiosoundings or additional information. 
Figure 3,- Coefficients (b 4 +b ¡)/2 and byb 4 of equation (9) versus the water vapor content, W. 
The temperature T,* (see equation 2) contains the information regarding to surface emissivity, 
although a small contribution of the downwelling atmospheric radiance is included. This temperature can be 
calculated from T¡ (equation 1) using the coincident atmospheric profiles and a radiative transfer model to obtain 
ti(0) and L¡^(0). The resulting temperature difference T 4 *-Ts* is a measure of Ae. According to Coll et al. 
(1993b), it is possible to write: 
, (T4*-T5*)-(l-e)(b5-b 4 ) n) 
(b 4 +bs )/2 
where b¡ are coeficients that depend mainly on the water vapor content and are given in Coll et al. (1993b). For 
applying equation ( 8 ), an estimate of the mean emissivity is necessary; however its influence is small since the 
difference b 5 -b 4 is small (see Figure 3). As shown in Coll et al. (1993b), an accuracy of 0.005 in Ae can be 
achieved if near simultaneous and coincident radiosounding measurements are used. 
3 - RESULTS AND DISCUSSION 
The models described above have been applied to satellite data in the framework of the research projects EFEDA 
(Bolle et al., 1993) and HAPEX-Sahel (Goutorbc et al., 1993). The areas of study were, respectively, an 
agricultural area in Castilla-La Mancha, Spain, and the Sahel in Niger. Besides, ground measurements of 
emissivity and temperature, and radiosonde measurements of the atmospheric profiles have been used. 
3.1. EFEDA Project 
Ground emissivity measurements of the various surfaces existing in the EFEDA pilot areas were performed 
using the box method during the field campaign in June, 1991. With these data, a digitized map of emissivity 
789