Comparing Aster and Landsat 7 Etm+ for Change Detection

The limited radiometric resolution and expiring life expectancy of the Landsat sensors have prompted this study to investigate the potential of using ASTER imagery in conjunction with Landsat imagery. This project developed methods to calibrate processed ASTER imagery and Landsat 7 ETM+ imagery to equivalent measurements for change detection studies. Landsat 7 ETM+ and ASTER imagery of southwestern Bangladesh were acquired on the same date for this study, with a thirty three minute temporal difference between the acquisition times. The methods were developed on a training site and verified on a test site. The methods developed for calibration were regression with Discrete Fourier Transform and cross-calibration approach using digital number ratios. The results, using both regression with Discrete Fourier Transform and cross-calibration approach show that more than sixty-five percent pixels of the test site are within 3 DN difference range for green, red and SWIR I band. However, only approximately forty-five percent of the pixels in the near infrared band are within the 3 DN difference range for test site using both of these methods. The variances of the DN differences for the green, red and SWIR I bands are low, whereas the variances of DN differences are high for the near infrared band in the test site. The variance of DN differences using the regression with Discrete Fourier Transform method is lower than the variance using crosscalibration approach for each spectral band in the test site. The lower variance indicates that regression with Discrete Fourier Transform more precisely calibrated both sensors rather than the cross-calibration approach using DN ratio. The results of this study suggest that the ASTER can be used for change detection in conjunction with Landsat 7 ETM+ using developed calibration methods.