546
6 - VALIDATION OF THE ATSR-IK ALGORITHM
To validate this algorithm, it has been applied on another set of 35 ATSR images taken over the same region
during the same period of the year ( summer 1992). The water vapour amount estimated by the new algorithm
defined in Section 4 can be written :
Wv = 8.229 - 7.705/?ii.i2 (4)
This amount has been calculated in the same way over the ATSR-M integration spots and compared to the
microwave radiometer measurement. After elimination of the cloudy areas, the number of ATSR/MW spots
kept is 41. The results of this comparison is shown on Figure 5.
i*
3
©
a.
w
a>
«
£
£
<
Figure 5 : Scatter plot of retrieved water vapour from ATSR/IR versus the retrieval from ATSR/MW
The correlation is very good with a standard deviation of 0.38 g/cm 2 relatively to the ATSR/MW instrument.
But a complete study of the accuracy of the algorithm can only be done by comparing the retrievals with
coincident radiosonde measurements, to eliminate the errors due to the ATSR/MW instrument. Then, the error
will probably be lower.
Nevertheless, this accuracy is sufficient for our purpose which is the classification of the atmospheres
for the choice of the Split-Window algorithms for surface temperature estimations (the 8 water vapour classes
corresponding to the following mean values : 0.26, 0.72, 1.26, 1.71, 2.21, 2.76, 3.44, 5.01 g/cm 2 ).
These preliminary results have to be confirmed on a larger database involving situations corresponding to all
climatic conditions in order to test the sensitivity of the algorithm to the atmospheric temperature particularly.
The next step will be the comparison with another methodology developed to obtain the columnar
water vapour content which is the variance ratio method proposed by Jedlovec (1990). This method has several
advantages and among others, the effects of surface emissivity differences are minimized and generally
negligible as shown by Sobrino et al.(1994). Then, these methods will be tested over continental areas where
radiosonde data may be available.
