Urban air pollution can be quantified in terms of atmospheric turbidity if satellite images of similar geometry acquired under clear atmosphere and pollution conditions are radiometrically compared. This comparison can be conveniently carried out by evaluating 'the apparent contrast reduction' due to aerosols over land. However, this method is subject to misclassification due to ground-reflectance temporal variations. This paper proposes the complementary application of 'the temperature attenuation' procedure, which allows one to decouple the results of the previous method from such variations and, thus, increase confidence in classification results. The respective image processing code developed has been tested successfully on Landsat-5/TM data of the urban area of Athens.