541 
ESTIMATION OF TOTAL ATMOSPHERIC WATER VAPOUR 
CONTENT OVER OCEANS FROM 
ERS1/ATSR DATA. 
C. OTTLE, S. LE MAGUER, C. FRANÇOIS 
L. EYMARD and L . TABARY 
C.E.T.P 
Centre Universitaire 
10 - 12 Av. de l'Europe 
78140 Vélizy, France 
ABSTRACT 
Different studies have shown that the estimation of the surface temperature from Split-Window algorithms can 
be greatly improved by a rough knowledge of the total atmospheric water vapour. A new algorithm is proposed 
to estimate this variable from infrared measurements in the 11pm region. This algorithm has been adjusted 
with ATSR/ERS1 data, taking advantage of the simultaneous measurements of the microwave radiometer. The 
first results show that the atmospheric water vapour column can be estimated with an accuracy better than 0.5 
g/cm 2 which is widely sufficient for our purpose. 
Further studies are necessary now to test the robustness of this algorithm whose simplicity of 
application seems promising. 
KEY WORDS : Précipitable Water, Thermal 1R , ATSR/ERS1, Surface Temperature 
1 - INTRODUCTION 
Surface temperature is a parameter which is important to map and monitor at large scale and over long time 
periods. This parameter can be obtained from space from thermal infrared satellite data after elimination of the 
atmospheric absorption which is principally caused by the water vapour in this spectral region. 
Different methods may be applied to estimate this effect but the most easy to use is the Split-Window 
method because it does not require any other ancillary data than the radiometric brightness temperatures. 
As a matter of fact, the combination of two measurements in two different spectral bands, both 
sensitive to water vapour only, allow to estimate the atmospheric absorption. The surface temperature is linked 
to the two brightness temperatures by a linear relationship. But the coefficients of this relationship are not 
constant and depend on several parameters like the surface emissivity, the mean air temperature and the 
tropospheric water vapour, as shown for example by Becker (1987). Consequently, it appeared necessary to 
take into account these parameters in the algorithms. Recently, different local Split-Window (SW) algorithms 
have been proposed which require the knowledge of local surface and atmospheric conditions (Sobrino et al., 
1991,1994, Li andBecker, 1993, Prata, 1993, Ulivieri et al„ 1993). 
Since it is not possible to obtain these parameters at the pixel scale easily, the local SW algorithms 
can not be applied routinely for most applications. For this reason, in a previous study, Ottlé and Vidal-Madjar 
(1992) proposed 3 different sets of coefficients depending only on the surface emissivity and on the satellite 
viewing angle, calculated for 3 different classes of atmosphere namely tropical, temperate and polar, on the 
TIGR dataset containing 1207 radiosoundings ( Scott and Chédin, 1981). But this classification was proved 
not suitable because the latitude can not be in most cases a choice criteria and large errors ( greater than 4K for 
high land surface temperatures) may appeared in the case where the atmospheric water vapour is far from the 
mean of the atmospheric class considered, since it is mostly the total water vapour which differentiate these 3 
classes. Therefore, we have proposed another set of coefficients for ATSR and AVHRR data for 8 classes of 
water vapour, ranging between 0.26 and 5 g/cm 2 ( François and Ottlé, 1994). At the same rime, we have 
investigated the ways to estimate the total atmospheric water vapour from infrared data. ATSR/ERS1 is a 
good tool to test such methods because the infrared instrument is coupled with a microwave passive radiometer