32. Application to satellite radiometer data 
In the previous subsection, the mathematical expressions derived to obtain the channel transmittances, and the 
total water vapor were found valid in application to simulated data. However, it is clear that the very test of any 
theoretical concept is to apply it to real data, and to compare it with ground truth data. In the present section we 
try to check the validity of the above equations in application to real data. To do this, two data sets have been 
used. The first data set consist in two NOAA-11 images of a semi-arid region of NW Victoria (Australia), 
collected at night during mid-August 1990, with, in addition, atmospheric radiosoundings measured at the 
satellite overpass times (provided by A. J. Prata), and two NOAA-11 images of a mountainous region around the 
radiosounding station of Belfort-Fontaine, France, acquired respectively at 2h UT and 14 h UT on 17 September 
1992, together with the appropriate radiosoundings provided by Météo France. Using this data set the exact 
values of R 54 , T 4 > T 5 and W calculated from the radiosoundings using LOWTRAN-7 are displayed in Table 2 
with the estimated values obtained from image brightness temperatures. The second data set consists in one 
NOAA-11 image of the Iberian Péninsule, acquired at 14 h UT on 19 June 1991, for which the total water vapor 
has been calculated according to the procedure suggested by Smith et al., (1985) using HIRS-2 input profiles, 
provided by M. Arbelo. Tables 3 and 4 display, respectively, the water vapor values obtained from Eq. (15), 
considering a grid of 50x50 pixels sections, and the retrieved using HIRS data. These last values have been 
placed in the correponding boxes according to its latitude and longitude. The water vapor mean values in tables 2 
and 3 are respectively (1.34±0.42) g/cm 2 and (0.94±0.45) g/cm 2 for the coincidents boxes, while from the 
coincident radiosounding registered by the Spanish Meteorological Service (INM) inside the study area and 
processed in the LOWTRAN-7 code, the vertical water vapor is 1.62 g/cm 2 . 
In conclusion, it is clear that although the main problem of such a validation is the absence of 
good quality "in situ" measurements, the two data sets offer a first-order check on the validity of Eqs. ( 8 ), (14a), 
(14b) and (15). 
TABLE 2.-Exact values of R 54 z 4 , T 5 and W at nadir calculated by the radiosoundings using LOWTRAN-7 and 
from Eqs. ( 8 ), (14a), (14b) and (15) using satellite brightness temperatures. 
AVHRR 
LOWTRAN-7 
Image 
R54 
T4 
*5 
W 
R54 
T 4 
*5 
W 
NW Victoria 1 
0.921 
0.878 
0.809 
1.35 
0.931 
0.859 
0.800 
1.13 
NW Victoria 2 
0.912 
0.861 
0.785 
1.42 
0.924 
0.846 
0.782 
1.12 
Belfort (night) 
0.840 
0.704 
0.591 
1.69 
0.864 
0.744 
0.643 
1.44 
Belfort (day) 
0.970 
0.924 
0.897 
0.70 
0.950 
0.881 
0.837 
1.03 
TABLE 3.-a) Water vapor values along the path TABLE 4,- Water vapor values obtained from HIRS 
(Eqn. 15) for the NOAA-11 AVHRR of the Iberian data 
Peninsula. The C values corresponding to cloudy 
areas. 
4- IMPROVING THE SPLIT-WINDOW METHOD 
Based on the above results, we propose in the present section another extension of the split-window technique for 
land surface temperature (LST) determination that incorporates the information on the atmospheric transmittance 
and lead to an improved algorithm for the atmospheric correction.