In this study, RapidEye (RE_ 5m multispectral-orthorectified) and Landsat 7 ETM+ imagery over the Ahipara region were utilized to classify the land cover from the study area. Various methods of image classification were implemented to produce the thematic maps of the land cover types. Twelve classified images from the L7 and RE images were generated by using different band (432, 543) and principle components 1, 2 with vegetation index layer combinations, as well as applying the supervised classification algorithms, including the maximum likelihood classifier (MLC) and combination of MLC with the parallelepiped algorithm. The error matrix and Kappa statistic of the classified images were estimated. The results of the classified images for both sensors (L7 and RE) identified all classified images by PPMLC had a higher accuracy and Kappa statistic than the classified images used by MLC approach. Furthermore, the Kappa statistics and the overall accuracies represented that the Red-Edge band from the RapidEye system combined with NIR and Red can improve the classification performance as it is sensitive to distinguish in vegetation cover.

This study revealed that one of the most accurate procedures for classifying the RapidEye image of the study area was a combination of principal components and vegetation index layers (PC12VI) while the degraded images (RE with 30m spatial resolution) and also the images with 432 band combination had lower accuracy assessment results. The results of the classification processes and comparative assessments between LCDB2 and RE data indicated the RE_543_PPMLC and RE_PC12VI_PPMLC images had higher classification performance in discriminating land cover types than the LCDB2 classification.

The RapidEye high spatial and spectral-resolution image represented more accurate classification performance with useful information in classifying the study area in Ahipara region whereas the Landsat 7 classified images had moderate classification accuracy.