Oral Session # 72: GIS and Remote Sensing I.
Presiding: J Drake
Thursday, August 7. 8:00 AM to 11:30 AM, SITCC Meeting Room 203.

Evaluating the effects of landcover classification parameters on landscape indices used to investigate wildlife-habitat relationships.

Laurent, Edward*,1, LeBouton, Joseph2, Walters, Michael2, Liu, Jianguo1, 1 Department of Fisheries & Wildlife, East Lansing, MI, USA2 Department of Forestry, East Lansing, MI, USA

ABSTRACT- Many studies of wildlife distribution patterns are based on land cover maps created from satellite images. While research into the relationships between wildlife distributions and land cover classes have been instrumental for increasing knowledge of species occurrence patterns, descriptions of landscape heterogeneity vary with the grain of analysis, the classification system in use, and the variability of spectral information employed. Further, land cover classes often do not accurately classify the landscape in ways that many species respond. For these reasons, an examination of classification parameters on map accuracy is badly needed. To address these concerns we used a grain representative assessment and inventory protocol (GRAIN) to survey sample locations (N = 198) for forest bird species in the Upper Peninsula of Michigan. Pixels overlapping survey plots were used as seed pixels in a habitat analysis by iterative classification procedure (HABICLASS) to create spectral signatures that described survey plots. These signatures were then used in supervised image classification to classify maps of species occurrence. Parameter values employed in supervised classification were incrementally modified to assess their influence on map accuracy. Map accuracy was affected by parameter values in different ways for each species. Parameter values resulting in the highest map accuracy for species occurrence were used to classify maps of land cover specific to each bird species. Landscape indices previously documented as having influence on the study species occurrence or life history were calculated for each land cover map and compared.

Key words: land cover, habitat analysis, remote sensing, landscape heterogeneity