Good news! The PRISM website is available for submissions. The planned data migration to the Scholaris server has been successfully completed. We’d love to hear your feedback at openservices@ucalgary.libanswers.com
 

Vegetation classification using Landsat TM and SPOT-HRV imagery in mountainous terrain, Kananaskis Country, southwestern Alberta

Date

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Landsat-5 Thematic Mapper (TM) 7-band data and SPOT-HRV 3-band data, and topographic measures derived from a digital elevation model (DEM) were used to determine the best set of classification variables for mapping vegetation communities in a mountainous region in southwestern Alberta. Ground and topographic data were gathered for 765 pixels sampled in the field. Discriminant analysis of an integrated TM+DEM data set resulted in overall accuracy of 72.77% for twenty-eight landcover classes, an improvement_ of 10.81 % over TM data alone. Discriminant analysis of a SPOT +DEM data set resulted in a similar accuracy of 73.44% for twenty-three classes, an improvement of 10.23% over SPOT data alone. Individual classes showed considerable variability in accuracy, which was reduced significantly with the addition of elevation, slope and incidence from the DEM. Most accuracy improvement occurred in poplar slope (by 60.7%), herb avalanche slope (40.9%), fen (by 28.5%) and Cassiope alpine meadow heath (by 25%) classes. Most spectrally distinct vegetated classes included mixedwood forest, herb- and pine-dominated rock outcrop complexes, several shrub-dominated communities, and burned and logged areas. Herb-dominated wetlands (fen, marsh, floating bog) were least distinct. Higher spatial resolution data (SPOT) is important in the separation of non­forested wetlands, particularly herb-dominated communities. In turn, higher spectral resolution data (TM) is important in the separation of non-forested upland classes. Integration of best classification accuracies from the two satellite data sets improved overall accuracy to 88.31 % and shows the need for further research in this area. A maximum likelihood classification and mapping of the TM data set illustrates the spatial distribution of the classes, and provides further support for statistical evidence that topographic data is necessary for landcover classification of high relief environments.

Description

Bibliography: p. 126-135.

Keywords

Citation

Williams, J. A. (1992). Vegetation classification using Landsat TM and SPOT-HRV imagery in mountainous terrain, Kananaskis Country, southwestern Alberta (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/15410

Collections