Methodologies for mapping the spatial extent and fragmentation of grassland using optical remote sensing

Thumbnail Image
Roy, Gairik
Journal Title
Journal ISSN
Volume Title
Lethbridge, Alta. : University of Lethbridge, Dept. of Physics and Astronomy, c2012
Grassland is an important part of the ecosystem in the Canadian prairies and its loss and fragmentation affect biodiversity, as well as water and carbon fluxes at local and regional levels. Over the years, native grasslands have been lost to agricultural activities, urban development and oil and gas exploration. This research reports on new methodologies developed for mapping the spatial extent of native grasslands to an unprecedented level of detail and assessing how the grasslands are fragmented. The test site is in the Newell County region of Alberta (NCRA). 72 Landsat and 34 SPOT images from 1985 to 2008 were considered for the analysis. With an airport runway used as a pseudo-invariant feature (PIF), relative radiometric correction was applied to 17 Landsat and 8 SPOT images that included the same airport runway. All the images were classified using the Support Vector Machine (SVM) classification algorithm into grassland, crop, water and road infrastructure classes. The classification results showed an average of 98.2 % overall accuracy for Landsat images and SPOT images. Spatial extents and their temporal change were estimated for all the land cover classes after classifying the images. Fragmentation statistics were obtained using FRAGSTATS 3.3 software that calculated land cover pattern metrics (patch, class and landscape). Based on the available satellite image data, it is found that in Newell County there is almost no significant change found in the grassland and road infrastructure land cover in over two decades. Also, the fragmentation results suggest that fragmentation of grassland was not due to the result of road infrastructure.
x, 105 leaves : ill., ; 29 cm
Grasslands -- Alberta , Grassland ecology -- Alberta , Fragmented landscapes -- Alberta , Remote sensing , Dissertations, Academic