Table 1.
Season specific MaxEnt AUC scores for each study species.
Figure 1.
Barplot of annual cycle timing for study species.
Figure 2.
Map of the simulated equilibrium water table depth for the contiguous US [18].
The values illustrate the depth in meters below the surface where the simulated water table is located.
Figure 3.
Plot of variable importance measure from Random Forest.
Variable importance is measured in mean decrease in accuracy, which is the decrease in accuracy of a classification after the variable has been randomly permuted. A higher mean decrease in accuracy means the variable contributes more to the accuracy of the classification. The abbreviations are as follows: ABDU (American black duck), BWTE (blue-winged teal), CAGO (Canada goose), MALL (mallard), NOPI (northern pintail), WODU (wood duck), temp (temperature), ppt (precipitation), elev (elevation), dwt (depth to water table), and pw (NLCD's percent wetland).
Figure 4.
Maps of predicted probability of occurrence for all study species' winter habitat.
Predictions were created using MaxEnt with 100% of known presence locations to increase accuracy of the visual representation. Temperature, precipitation, elevation, and water table depth were the predicted variables used to construct the probability surfaces.
Figure 5.
Plot of relationship between water table depth (m) and occurrence probability for species in winter.
The plots were constructed by selecting 1,000 random points from the predicted probability of occurrence surface. The red curve is a smoothing spline fit to the mean of the data points, and is meant only to visually illustrate the trend of the data and the upper threshold of DWT.