THERMAL IR REMOTE SENSING OF ATMOSPHERIC TRANSMITTANCE 
AND WATER VAPOR FROM AVHRR DATA 
J. A. SOBRINO 1 , Z.-L. LI 2 , F. BECKER 2 and V. CASELLES 1 
1 University of Valencia, Department of Thermodynamics. 50, Dr. Moliner, 46100 Burjassot (Spain) 
2 GSTS/LSIT/ENSPS, 7 rue de l’Université, 67000 Strasbourg (France) 
ABSTRACT 
This paper describes how the spatial covariance ( 045 ) and variance ( 044 ) of image brightness temperature 
measured in Channels 4 (=10.3-11.3 |im) and 5 (=11.5-12.5 pm) of the Advanced Very High Resolution 
Radiometer (AVHRR) on board the NOAA-11 can be used to derive the atmospheric transmittance for each 
channel and the total column content of water vapor. The technique starts from the relationship exhibited between 
the ratio of atmospheric transmittances in the two channels of the split-window, and the observed variations of 
satellite brightness temperatures. This relationship is obtained under the condition that the atmosphere and the 
surface emissivity are unchanged over N neighboring pixels where the surface temperature changes. LOWTRAN- 
7 code was used to simulate remote sensing of atmospheric transmittance and water vapor over a wide range of 
situations and to test the accuracy of the method. The method was applied to AVHRR data over different regions 
and compared with other measurements (radiosounding or HIRS). Finally, it is shown, that the use of this 
information allows deriving a novel extension of the split-window technique which represents an improvement 
over the conventional algorithms for satellite-derived surface temperature. 
KEY WORDS: Atmospheric transmittance, AVHRR, Split-Window, Water vapor. 
1- INTRODUCTION 
In order to improve the quality of remote-sensing data there is a need to estimate the atmospheric transmittance 
and/or the total water vapor from the image itself. In fact measurements of these parameters allow to improve 
the accuracy of remotely sensed surface temperature (ST) which is a necessity for many applications, notably 
agrometeorology, climatic and environmental studies. However the accurate determination of ST constitutes a 
difficult task. The principal sources of distortions arc atmospheric attenuation and emission of thermal radiation 
and the non-black-body nature of the surface. Different techniques have been developed to eliminate these 
perturbations, the most popular being the split-window technique (McClain el al„ 1985, etc). This technique 
takes advantage of the differential absorption in two adjacent spectral windows, centred at 10.8 pm and 12 ^im 
for the AVHRR instrument, to correct for atmospheric effects and describes the surface temperature in terms of a 
simple linear combination of brightness temperatures Tj and T 2 as measured in both thermal channels. 
T s =ao+aiTi+a 2 T 2 , where T s is the surface temperature measurement, and ao, aj, and a 2 are coefficients that have 
been chosen to minimize the error in the ST determination. With this point of view numerous studies have been 
made to determine the coefficients over the sea surface (McMillin, 1975; Deschamps an Phulpin, 1980; Barton 
etal., 1989) and land surface (Price, 1984; Becker and Li, 1990; Sobrino et al., 1991). 
Although successful results have been obtained using these approaches, the algorithms given in 
these works do not respond to the stringent accuracy requirements for ocean (0.3 K) and land surface studies (1 
K). In this direction Harris and Mason (1992), and Sobrino et al. (1994) demonstrated that the inclusion of the 
ratio of transmittances in Channels 4 and 5 of AVHRR, or, the total water vapor content in the algorithms 
permits the elimination of a significant quantity of error in the retrieval of SST and constitutes a large 
improvement over the currently used split-window algorithms. In recent years, numerous methods have been 
proposed to derive precipitable water and the channel transmittance ratio from satellite-based sounding radiometers 
such as the IRIS (Infra-red Interferometer Spectrometer) on board the NIMBUS-4 satellite (Prabhakara et al., 
1979), the HIRS (High-resolution Infra-red Radiation Sounder) that is the infra-red part of TOVS (TIROS-N 
Operational Vertical Sounder) on board the NOAA satellite, the VISSR Atmospheric Sounder (VAS) on the 
GOES satellite (Chesters et al., 1983), etc. In practice, however, especially when the AVHRR data are the only 
available, it is more useful to apply methods that use the radiometric temperatures in two-infrared channels. 
Thus, Klesspies and McMillin (1990) obtain the ratio of transmittances from the channel brightness temperature 
differences by assuming that the atmosphere and surface emissivities in Channels 4 and 5 are invariant. Jedlovec 
(1990) proposes an extension of this technique that uses the ratio of the spatial variance of the channel brightness 
temperatures. Dalu (1987) and Schluessel (1989) presented other approaches that are based on the relationship 
exhibited between water vapor and the difference of brightness temperatures between the two infrared channels.