Integrating satellite imagery and digital elevation data for geobotanical mapping in Quetico Provincial Park, Ontario, Canada

Timothy Andrew Warner, Purdue University

Abstract

This geobotanical remote sensing study of a granite - greenstone terrain in eastern Quetico Provincial Park, north-western Ontario, Canada, is based on a topographic analysis of digital elevation data. The greenstone units consist of metamorphosed mafic and occasional ultramafic lithologies, surrounded by granites and granodiorites. Glacial till is thin and capped by loess in places. The soils are mainly Inceptisols, Histosols and Boralfs. The vegetation is dominantly boreal in nature. The digital elevation data analysis identified four micro-climate and drainage classes. Landsat Thematic Mapper (TM) data were transformed into the nPDF Deciduous Forest Index and then density sliced in six informational classes of various deciduous and coniferous cover. The distribution of forest classes on each of the two lithologies shows that stands growing on mafic lithologies are enriched in deciduous species, compared to those growing on granites. Of the cover classes found on mafics, the highest coniferous component was on north facing slopes, and the highest deciduous component on south facing slopes. By contrast, granites showed no appreciable variation between site classes. These associations were then applied in an integrated, rule-based classification that utilizes average community vegetation, topographic landscape units and aeromagnetic anomalies to separate greenstone belts from surrounding granitic areas. The procedure, called Topoveg, is not overtly hampered by scattered clouds in the area, since areas that are obscured or have no forests on adjacent south facing slopes are assumed to be similar to their surrounding areas. Topoveg achieved an 85.0 percent accuracy in a classification of the 10 by 13 kilometer primary test site. A standard maximum likelihood classification had an accuracy of only 71.3 percent, and produces output that has a distinctive noisy texture, compared to the large, homogenous classes of Topoveg. In a neighboring test site to the north, Topoveg accuracy was similarly high (86.3 percent). The southern test site, however, had an increased percentage of deciduous species on south facing granitic sites, and consequently a lower accuracy (78.6 percent). Thus these particular classification rules may be used with confidence in areas to the north, east and west. New rules should be developed for classification of areas to the south.

Degree

Ph.D.

Advisors

Levandowski, Purdue University.

Subject Area

Geology|Geography|Ecology|Remote sensing|Geography

COinS