Fig 1.
Spatial extent and occurrence record locations of hoary bats (Lasiurus cinereus) in North America, 1950–2000.
Table 1.
Final set of predictor variables used in modeling patterns of season distribution of hoary bats (Lasiurus cinereus) in North America.
Fig 2.
Boxplot comparing Area Under the Curve (AUC), sensitivity, and specificity statistics using test data and the cross-validation mean for each season and each of 5 forms using the final set of predictor variables in modeling patterns of season distribution of hoary bats (Lasiurus cinereus) in North America.
G = Generalized Linear Models using logistic regression in a maximum likelihood framework; M = Multivariate Adaptive Regression Splines; B = Boosted Regression Trees; R = Random Forest; X = Maxent approach to maximum entropy modeling.
Fig 3.
Ensemble maps for each season and sex for North American hoary bats (Lasiurus cinereus) using ensemble binary prediction from 5 species distribution model forms.
Dots are occurrence record locations for hoary bats.
Table 2.
Comparison of Area Under the Curve (AUC) statistic using test data and the cross-validation mean for each season and each of 5 species distribution model forms using the final set of predictor variables in modeling patterns of seasonal distribution of hoary bats (Lasiurus cinereus) in North America.
Table 3.
Comparison of sensitivity and specificity using test data and the cross-validation mean for each season and each of 5 species distribution model forms using the final set of predictor variables in modeling patterns of season distribution of hoary bats (Lasiurus cinereus) in North America.
Table 4.
Variable importance by season using increase in Area Under the Curve (AUC) when each predictor variable is permuted using 5 SDMs to model seasonal distributions of hoary bats (Lasiurus cinereus) in North America.
Fig 4.
Ensemble maps for each season and sex for hoary bats (Lasiurus cinereus) in the United States using ensemble binary prediction from 5 species distribution model forms.
Locations of > 47,000 industrial wind turbines in the contiguous United States are indicated by black stars. The color ramp indicates the proportion of the 5 SDM model runs for each season that are considered suitable habitat. Darkest blue indicates that all of the models predict the area to be suitable habitat, while lightest blue indicates that all of the models predict the area to be unsuitable habitat. The histogram associated with each map shows the empirical distribution for mean habitat suitability for all wind turbines in the United States, estimated using the 5 SDM approaches. Negative (left) skewness in the histograms, or higher frequencies of values toward the right side (= 1) of the horizontal axis, indicate overlap between modeled suitable habitat and turbine locations.