Revista Brasileira de Cartografia (Jun 2017)

LAND-USE AND LAND-COVER MAPPING OF THE BRAZILIAN CERRADO BASED MAINLY ON LANDSAT-8 SATELLITE IMAGES

  • Carlos Aberto de Mattos Scaramuzza,
  • Edson Eyji Sano,
  • Marcos Adami,
  • Edson Luis Bolfe,
  • Alexandre Camargo Coutinho,
  • Júlio César Dalla Mora Esquerdo,
  • Luís Eduardo Pinheiro Maurano,
  • Igor Silva Narvaes,
  • Francisco José Barbosa Oliveira Filho,
  • Roberto Rosa,
  • Elaine Barbosa Silva,
  • Dalton Morrison Valeriano,
  • Daniel Castro Victoria,
  • Adriana Panhol Bayma,
  • Gustavo Henrique Oliveira,
  • Bayma-Silva Gustavo

Journal volume & issue
Vol. 69, no. 6

Abstract

Read online

The Brazilian Cerrado is one of the world´s biodiversity hotspot and hosts some of the most intensive agricultural activities for food production in the world. The objective of this study was to produce a land-use and land-cover (LULC) map of the Cerrado based on Landsat-8 Operational Land Imager (OLI) images. A set of 121 scenes from 2013 was processed using the image segmentation technique. The segments were exported in the shapefi le format and interpreted visually in a geographical information system software using RGB/564 color composites. The following LULC classes were considered: annual croplands, perennial croplands, cultivated pasturelands, reforestation, mosaic of occupation, urban areas, mining areas, bare soil, forestlands, non-forestlands, water bodies, and non-identifi ed (clouds and burned areas). The overall accuracy was estimated by an independent scientist with large experience in Cerrado´s image interpretation. The results showed that 43.4% of the study area (88.5 million hectares) were already converted into agricultural, urban and mining areas, 54.6% (111 million hectares) were still natural areas, and 1.9% (3.9 million hectares) was classified as non-identified. Cultivated pasturelands were the most representative land-use type (29.5%), followed by annual croplands (8.5%) and perennial croplands (3.1%). The overall accuracy of the final map was 80.2%.

Keywords