Table 1.
Performance metrics of hypervolumes in the environmental space
| Environmental predictors | Model calibration region | Mean performance in E (SD) | Total volume | ||
|---|---|---|---|---|---|
| OC-SVM | Convex hulls | OC-SVM | Convex hulls | ||
|
PCs prop. = 95% |
Centroid based buffer | 0.70 (0.12) | 0.54 (0.17) | 80.25 | 87.16 |
|
PCs prop. = 95% |
South America | 0.68 (0.11) | 0.56 (0.18) | 50.55 | 61.04 |
|
PCs prop. = 96% |
Americas | 0.76 (0.15) | 0.45 (0.19) | 20.04 | 15.32 |
| Climates | Centroid based buffer | 0.74 (0.11) | 0.44 (0.17) | 8,426,855 | 10,131,022 |
| Climates | South America | 0.74 (0.13) | 0.48 (0.20) | 8,444,272 | 10,131,022 |
| Climates | Americas | 0.72 (0.13) | 0.51 (0.18) | 8,507,409 | 10,131,022 |
Performance metrics for one-class support vector machines (OC-SVM) and convex hulls hypervolumes were measured in the environmental (E) space using three different model calibration regions and two categories of environmental predictors. Best performing model in bold. SD Standard deviation, PCs Principal components, Prop. Cumulative proportion of the three principal components used for model calibration