Whooping Crane Migrational Habitat Use Draft (2)

Home Browse Search

been chosen from a"choice set" consisting of points in the available sample. For used points in the wetted channel, the choice set consisted of available points in the wetted channel. For used points outside of the wetted channel, the choice set consisted of available points outside of the wetted channel. The discrete choice model accounts for different land cover options within the two strata, wetted channel and upland, while modeling the underlying relationships between selection and land cover predictor variables. The model for all opportunistic locations had sample sizes of 320 used and 6420 available locations. The model for all systematic locations had sample sizes of 72 used and 6420 available locations. The 27 variables considered as candidates for each of these models were the land cover category, the distance to each land cover category, and the proportion of each of the 9 land cover categories within 0.25 miles of the location (Table 2). Study area in-channel selection of land cover In-channel selection of land cover was modeled with an exponential function (Manly et al. 2002). The model for in-channel systematic locations had sample sizes of 58 used and 189 available locations. The model including opportunistic locations had sample sizes of 150 used and 189 available locations. The same 27 variables were considered as candidates these two models. Local area selection of land cover Local area selection of land cover was modeled as a discrete choice in space (Manly et al. 2002). In this model, a used location was considered a choice, and each location was considered to have been chosen from a"choice set" consisting of points in the available sample within the local area (as described above). A discrete choice model will account for different land cover options within the local area around the crane observation, while modeling the underlying relationships between selection and land cover predictor variables. The model for all opportunistic locations had sample sizes of 321 used and approximately 300 available locations in each choice set. The model for systematic locations had sample sizes of 72 used and approximately 300 available locations in each choice set. The same 27 land cover variables were considered as candidates for each of these models, and forward model selection proceeded as described above. Local in-channel selection of land cover Local area selection of in-channel land cover was modeled as a discrete choice in space (Manly et al. 2002). In this model, a used location was considered a choice, and each location was considered to have been chosen from a"choice set" consisting of in-channel points in the available sample within the local area (as described above). A discrete choice model will account for different land cover options within the local area around the crane observation, while modeling the underlying relationships between selection and land cover predictor variables. The model for in-channel opportunistic locations had sample sizes of 151 used and approximately 15 available locations in each choice set. The model for in-channel systematic locations had sample sizes of 58 used and approximately 15 available locations in each choice set. The same 27 land cover variables were considered as candidates for each of these models, and forward model selection proceeded as described above. 8