Path Radiance Based PM10 Concentration Estimation Using Artificial Neural Network and Support Vector Regression Methods on Landsat 8 Remote Sensing Imagery (Case Study: Tehran)

Remote sensing imagery is a rich source of information with applications in varied fields. Monitoring of environment pollution is one of them. The work presented in this paper is focused on estimation of the ambient concentration of pollutant using remote sensing. Particulate Matter with particle sizes less than 10 microns (PM10) is estimated for the study area of Tehran. Landsat 8 data of different wavelength has been processed and analyzed for the relationship with coincident ground station PM10 data. In first step, the path radiance was calculated for all bands and then the relationship between ground station PM10 data and path radiance of corresponding pixels was defined by using Support Vector Regression (SVR) and Artificial Neural Network (ANN) methods. In final step the best model was used predict the PM10 concentraion value for all pixels in Landsat 8 images. The result shows that ANN model achieved better correlation coefficient R2=0.73 on test data and better choice for predicting the PM10 concentration from Lansat 8 imagery.

کلیدواژه ها:

Path Radiance ، PM10 Concentration ، Support Vector Regression ، Artificial Neural Network ، Landsat 8 Imagery