Land-use history or land-use fantasy: Scale dependant errors in vegetation classification and their implications for environmental monitoring

Jones, S 2003, 'Land-use history or land-use fantasy: Scale dependant errors in vegetation classification and their implications for environmental monitoring', Environmental Informatics Archives, vol. 1, pp. 467-474.


Document type: Journal Article
Collection: Journal Articles

Title Land-use history or land-use fantasy: Scale dependant errors in vegetation classification and their implications for environmental monitoring
Author(s) Jones, S
Year 2003
Journal name Environmental Informatics Archives
Volume number 1
Start page 467
End page 474
Total pages 7
Publisher International Society for Environmental Information Sciences
Abstract As environmental scientists we are used to ordering or classifying landscapes into convenient land-cover/-use units. These may be based on ecological survey data, modeled predictions and/or remotely sensed imagery. If we wish to reconstruct a land-use history we are often forced to use 'second-hand' or 'recycled' data that has frequently been collected at a different spatial scale and/or for a different purpose. This paper seeks to ask the questions: Can the classification of vegetation data be carried out without regard for spatial scale? Can different classifications really be compared? Further, are there any protocols that can be proposed to aide researchers when they seek to compare classifications from different lineages? In order to evaluate these questions a series of test grids was generated: varying from a completely random distribution of scene elements to a highly clustered or spatially auto-correlated pattern. Each of these grids was then classified and degraded to four levels of generalisation. In parallel, the grids were also degraded and then classified. The resultant classifications were then compared. In all cases the integrity of the data was best preserved when the image was degraded first and then classified. Changing the spatial scale (degrading) of classifications resulted in a rapid decline in information content. In order to understand the impacts of these errors on land-use change studies, images (of differing spatial resolutions) of an Australian National Park, were classified. These results were then compared to a variety of land-use classifications used by the government agency managing the Park. Substantial differences in classification accuracy were revealed with significant implications for land management in the area.
Subject Photogrammetry and Remote Sensing
Copyright notice © International Society for Environmental Information Sciences
ISSN 1811-0231
Versions
Version Filter Type
Access Statistics: 101 Abstract Views  -  Detailed Statistics
Created: Wed, 04 Aug 2010, 13:48:05 EST by Catalyst Administrator
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us