Effects of Different Methods of Radiometric Calibration on the Use of Training Data for Supervised Classification of Landsat5/TM Images from other Dates

In studies that involves supervised classification of several temporal images, the use of specific samples extracted from each image may require field work or image interpretation and is often expensive. The cost could be reduced with the use of reference data from a different time. However, there may appear differences in the spectral behavior of land cover classes across time due to imaging issues, which can prevent the proper reuse of this type of training data. This paper assesses the influence of image calibration on the classification of Landsat5/Thematic Mapper (TM) images using Maximum Likelihood classifier and the use of land cover training samples collected in images obtained at different times. Results show that, although the calibration method may affect the classification results, it had a small impact on classification global accuracy.