Fig 1.
Predictive accuracies (AUC, Kappa, and TSS) of Larix principis-rupprechtii.
Nine different symbol types (dark circles) indicate nine random split-sample bouts (original data were randomly divided into two sets: a calibration set and a validation set) and the same symbols are linked by the same straight lines.
Fig 2.
Box-whisker plot of differences in model performance (AUC, Kappa, and TSS) among model classes when data were pooled for all species and split-sample bouts.
Dots show the mean predictive accuracy across species and split-sample bouts.
Table 1.
Repeated-measures ANOVA assessing changes in model predictive accuracy (AUC, Kappa and TSS) between modeling approaches and tree species.
Table 2.
Linear regression modeling of the effects of specie traits on niche model performance.
Table 3.
Significance (P-value) of difference in the changes in species’ range (relative to baseline) predicted by three different consensual approaches.
Fig 3.
Overlap maps of current and future potential presence-absence distributions predicted using three different consensual approaches for Pinus yunnanensis (left column) and Pinus tabulaeformis (right column).
Good (green) indicates species predicted to be present by all three consensus approaches. Moderate (blue) indicates species predicted to be present by any two of the three consensus approaches. Poor (read) indicates species predicted to be present by any one of the three consensus approaches.
Fig 4.
Pairwise correlation among predictions produced by three different consensual approaches (average, frequency, and median (PCA)).
Data are presented as mean ± SE. Means in the same time slice followed by the same letter are not significantly different at P ≤ 0.05 according to LSD.
Table 4.
Species traits, model accuracy, and map correlation.