Landscape indicators derived from Landsat-TM data: A case study in Brazil

Laboratory of Geo-Information Science and Remote Sensing
  Education
  Research
  Publications
  Models
  News & Calendar
  News
  Calendar
  Archive
  News
  Calendar
  2011
  2010
  2009
  2008
  2007
  2006
  2005
  2012
  Staff
  Equipment
  Contact details
  Workshops

29 Aug 2006 14:30 - 29 Aug 2006 15:00
Unit: Laboratory of Geo-Information Science and Remote Sensing
Organisation: Wageningen University

By Silvia Weel     

Abstract:  
Human pressures on earth’s ecosystems, through the continued exploitation of natural resources and induced changes in the landscape composition, results in the configuration of unsustainable ecological systems and impacts the persistence of natural dynamics. This scenario degrades ecological integrity and limits the provision of the ecosystem services that support human well-being. The depiction of landscape indicators, through the assessment of landscape patterns and processes, provides spatial information to assist a more balanced approach in the planning and sustainable management of natural resources. Campinas is a municipality in south-east Brazil requiring a suitable and strategic environmental policy, due to intense anthropogenic influence in the landscape composition. Remote sensing and GIS provide different techniques for the assessment of these parameters. Classes of land cover and plant functional types were depicted from Landsat TM 5 images from 1988 and 2004, through maximum likelihood (ML) and decision tree (DT) algorithms, in order to derive the landscape pattern. Different band composition was considered for each classifier, where the ML used original bands and the DT took several vegetation indices (NDVI, MVI and TCT). A conversion analysis was based on the post-classification comparison of the classified images; deriving indicators of human induced changes and natural processes, such as expansion of human activities, degradation and regeneration of natural vegetation. The results of the classification procedure indicated that the ML and DT had similar performance, considering the bands input used in each model. However, some classes had a low accuracy due to the algorithms functions, biophysical variations and sensor’s qualities. The conversion matrix proved to be very useful in delivering information for understanding the overall processes occurring within the landscape as well as providing an insight on the best technique and variables to be used for deriving more specific processes. The planning and management perspectives of this landscape analysis are dependent on the goal and scale of the approach of each administrative institution acting at the regional, municipal, and local level.

  

Print this activity