Remote sensing image processing to identify spatial units of human occupation along Trans-Amazonian Highway (BR-230), Brazil

To investigate the urban phenomenon in the Amazon is necessary to observe the cities and communities. Identifying these population nuclei can provide information about where the population is concentrated and how it relates to the space and environment, therefore, how Amazonian urban is structured. This study identified spatial units of human occupation along the Trans-Amazonian Highway (BR-230) by applying remote sensing image processing techniques. The study site is located in Pará state, Brazil, in the municipalities of Altamira, Brasil Novo, Medicilândia and Uruará, inside a 15 km buffered from the Highway. Four Landsat-5 Thematic Mapper orthorrectfied scenes from 2011 were processed using software SPRING. The processing steps consisted in mosaicking the scenes, the application of dilation filter, segmentation and maximum likelihood classification. The validation was based on manual classification of middle resolution RapidEye images (5 metres) and ancillary data from Brazilian Institute of Geography and Statistics (IBGE). Twenty three spatial units of human occupation were mapped and the validation showed a Kappa coefficient of 0.6785. The application of dilation filter during the processing was able to identify spatial units of human occupation in the study site, although some misclassified pixels occurred mainly in small patches.