Application of fusion techniques between Radarsat and Landsat images for landuse distribution at Cartagena, Colombia, using supervised classification

Abstract

Use of satellite images has been of great help for studies of landuse distribution in different places at different conditions. Landsat images have been one of the best ones due to their great spectral resolution (Landsat TM-7 with 7 bands), but their limited spatial resolution of 30m has been of the biggest disadvantages for detailed landuse studies. In the other hand Radarsat images have one of the best spatial resolution (6 m) but poor spectral resolution (pancromatic mode). Besides their good spatial resolution Radar images have other advantages over other images like their capacity to operate under any climatic conditions and at any time (day or night). Radar images are able to get information even in cloudy conditions and this characteristic make them perfect for studies of coastal zones and specially on tropical areas, where most part of the year are covered by clouds. The study area in this case was located at a tropical area (09°33’45” N and 75°23’45”W) so it was absolutely necessary to find a solution to perform the best detailed landuse distribution in order to actualize coastal zone management plans for the area. Nowadays techniques have been able to overcome these disadvantages. Applications of fusion techniques were able to solve these limitations, by the combination of the best characteristics of these two types of images. In this specific case the good multiespectral resolution of Landsat images and the good spatial resolution of Radarsat images were combined in order to obtain a completely new image, but only with 20 m of spatial resolution since Radarsat resolution was decreased due to the exaggerated difference in spatial resolution between the two images (6m and 30m). This is very much recommended to decrease errors created during the pixel to pixel fusion. With this new image obtained a good and detailed landuse distribution was performed by the application of supervised classifications in the study area. The results obtained were of good percentage of reliability and then used for new coastal zone plans.