PARENT SESSION
Oral Session 7: GIS / Remote Sensing I.
Presiding: E Ellis and P Valko
Monday, August 2, 8:00 AM to 11:30 AM, Meeting Room B 116.
Using the spatial and spectral precision of satellite imagery to predict wildlife occurrence patterns.
Laurent, Edward*,1, Shi, Haijin1, LeBouton, Joseph2, Walters, Michael2, Liu, Jianguo1, 1 Dept. of Fisheries & Wildlife, East Lansing, MI, USA2 Dept. of Forestry, East Lansing, MI, USA
ABSTRACT- Many studies of wildlife distribution patterns, such as Gap analysis, use land cover maps derived from satellite images. However, land cover classes describing overstory composition often do not accurately or precisely classify the landscape in ways that many species perceive or respond. An alternate approach to mapping species occurrence is to bypass a land cover map altogether and directly classify maps of species distributions using spectral descriptions of locations where species are known to occur. To investigate the potential of this approach we used point count surveys for birds and Landsat imagery to classify occurrence maps of three warbler species over a 385,000 ha region of the Upper Peninsula of Michigan. These species, black-throated green warblers, Nashville warblers, and ovenbirds, were known to select for aspects of forest understory during breeding. Spectral values of pixels from multiple season imagery overlapping the center of point count survey plots were used to create spectral signatures describing each plot. These signatures were used in supervised image classification to create maps of species occurrence independent of a land cover map. Accuracy and Kappa values of species occurrence maps were quantified using cross-validation, a separate validation data set, and comparisons with contemporary Gap analysis maps. Spectral values predicted species occurrence much better than both Gap analysis and chance assignment of pixel values. Results also indicate that care must be taken in interpreting outcomes of similar efforts, as map accuracy (percent correctly classified) is rarely a practical measurement of predictive ability.
Key words: remote sensing, birds, Kappa, GAP