Landsat 8 image classification through segmentation in free software to analyze the land use/land cover surrounding the Raso da Catarina ecological station, Bahia/Brazil
ISBN 978-85-88783-11-9
Authors
1Rios Oliveira, U.; 2Lustosa Brito, P.; 3César Pedrassoli, J.; 4Antunes Zaloti, F.; 5Mara Hadlich, G.
1MEAU/ESCOLA POLITÉCNICA DA UFBA Email: uldericovarzeano@hotmail.com
2ESCOLA POLITÉCNICA DA UFBA Email: britopatricia@hotmail.com
3ESCOLA POLITÉCNICA DA UFBA Email: pedrassoli.julio@gmail.com
4INSTITUTO DE GEOCIÊNCIAS DA UFBA Email: fabia.zaloti@gmail.com
5INSTITUTO DE GEOCIÊNCIAS DA UFBA Email: gisele@ufba.br
Abstract
The satellite imagery are widely used for the study of the Earth's surface and its dynamic by having several advantages associated to the power of coverage, time of collection and larger processing possibilities when compared to other classic methods of data collection. The present study aims to map the coverage and land use surrounding the Ecologic Station Raso da Catarina (EERC) using satellite imagery. The EERC is a Federal Conservation Unit under integral protection, located between the Parallels 9°33'13"S and 9°54'30"S and meridians 38°26'50"W to 44°38'00"W, in the municipalities of Jeremoabo, Paulo Afonso and Rodelas, in the State of Bahia, Brazil. The region have a significant biological importance, its predominant natural vegetation is the Caatinga, where there are several species of endemic occurrence and in danger of extinction. This work is part of a larger study that intends to identify the licuri tree (Syagrus coronata) and corn crops in order to promote the balance of the ecological and economic dynamics that involves Brazilian famous blue bird, the lear's macaw blue (Anodorhynchus leari), and local communities. The mapping of land cover surroundings of the EERC was conducted processing LANDSAT 8 image (5/10/2014, paths 66, 67 and row 216), provided by the United States Geological Survey-USGS, through the composition of the bands 4, 5 and 6. These images are part of the Landsat mission continuity program, released in February 2013. Image processing was carried out in SPRING (Geographic Information Processing System-INPE) which features image-processing modules, segmentation algorithms and integration of vector and raster formats in the same environment. The images passed through the following steps: segmentation (similarity 50 and area of 500); regions extraction and final classification (type of classifier: Bhattacharya). Supervised classification was carried out using samples collected visually to 5 unique classes of land use: urban use, agriculture, pasture, vegetation and water bodies, with subsequent classification results manual edition in order to eliminate potential errors and omissions aiming the refinement of results, considering checking and visual interpretation. In the surrounding area of the EERC considered in this paper, 8831.5 km², classes of land use computed the following results: 29 km² of urban area, 32.4 km² of agriculture use, 4552 km² of pasture use, 4050.8 km² of natural vegetation and 166.9 km² of water bodies. Analyzing the data for each of the three municipalities present in the study area we can observe that Jeremoabo (4627.7 km² area) have the following land-use classes results: 3 km² of urban area; 4.6 km² of agriculture; 1796.7 km² of pasture and 2823.4 km² of natural vegetation. In the Paulo Afonso area (1480 km²) the results compute: 24.5 km² of urban area; 843.1 km² of pasture use; 574 km² of natural vegetation and 38.4 km² of water bodies. To Rodelas area (2723.4 km²) the calculated uses are as following: 1.5 km² of urban area; 27.8 km² of agriculture use, 1912.2 km² of pasture use; 653.4 km² of natural vegetation and 128.5 km² of water bodies. Landsat 8 images and region segmentation algorithm allowed to reconstitute and to quantify land cover/use scenarios. The analysis of their main classes of use as vegetation, agriculture and urban area allows the modeling of future scenarios of this environmentally important area, witch is also a ecosystem under pressure. Nevertheless, future recommendations for this initial approach include the need for temporal analysis of land use and human occupation in the study area, considering the possibility presented by the continuity of Landsat mission. The integration with higher spatial resolution images and with additional data for object-oriented analysis is planned as the next step towards the identification of corn crops and licuri trees.
Keywords
Remote Sensing; Landsat image ; cover and land use