Optical characteristics of synthetic aerosol and the information composition of pixel in remotely sensed data are analyzed. The information of every pixel can be generalized as the sum of soil, vegetation, and water or moisture with nonlinear relationship. Then a group of equations based of multi channel remote sensing data can be built to decompose the pixel information and quantitatively retrieve the synthetic aerosol turbidity using Landsat TM data. The results of experiment show that the accuracy of air pollution information retrieved satisfies the requirement. Applying this model, air pollution circumstances over the Changjiang River Delta area in 1984 and in 1997 were detected, respectively. The results indicated that the area and the intensity of air pollution over this region are increasing quickly.