Fig 1.
General simulation and analytical scheme to test the effect of collinearity among environmental variables on species distribution modeling results.
Fig 2.
Factor loadings for the first two principal components of the Bioclim environmental variables for South America.
Identifiers of the Bioclim variables: b1: annual mean temperature; b2: mean diurnal range; b3: isothermality; b4: temperature seasonality; b5: max temperature of warmest month; b6: min temperature of coldest month; b7: temperature annual range; b8: mean temperature of wettest quarter; b9: mean temperature of driest quarter; b10: mean temperature of warmest quarter; b11: mean temperature of coldest quarter; b12: annual precipitation; b13: precipitation of wettest month; b14: precipitation of driest month; b15: precipitation seasonality; b16: precipitation of wettest quarter; b17: precipitation of driest quarter; b18: precipitation of warmest quarter; b19: precipitation of coldest quarter.
Table 1.
Eigenvalues and proportion of total variance explained by each axis derived from a principal components analysis of climate data for South America.
Fig 3.
Virtual species used in simulations.
In each cell, the two maps represent species with the same niche centroid but with small and large tolerance in ecological space, respectively. The numbers will be used to identify each species for the rest of this paper.
Table 2.
Results of the factorial ANCOVA for the Overprediction (OP), Underprediction (UP) rates and TSS measure as dependent variables calculated under the balance threshold.
DF is degrees of freedom.
Fig 4.
Interaction plots for the effect of environmental variables, modeling algorithm, and species tolerance in the environmental space on the Overprediction (A) and Underprediction (B) rate of the distribution models using balance threshold. Bars represent confidence intervals of 95%.
Fig 5.
Interaction plots for the effect of environmental variables, modeling algorithm and species tolerance in the environmental space on the TSS of the distribution models using balance threshold.
Bars represent confidence intervals of 95%.
Fig 6.
Residual plot of the ANCOVA results of the TSS response variable in relation to modelled species, modeling algorithm, and tolerance in environmental space.
Bars represent confidence intervals of 95%.
Fig 7.
Variance of the TSS measure for different SDM algorithms with raw environmental variables and the first four PCA axis (4 PCA) and six PCA axis (6 PCA).
Bars represent the 95 confidence intervals.
Fig 8.
Linear regression of predicted on real range size for modeled species using Balance thresholds.
Red lines represent the linear regression of the data; blue line show the prediction of the relationship if the predicted range-size were equal to the real range-size.
Table 3.
Relationship between real and predicted range sizes based on a linear regression model.
a is the intercept, b is the slope of the regression. R2 is the coefficient of determination. R2 values higher than 0.8 are highlighted. Higher R2 associated with a≈0 and b≈1 denotes best models for range-size prediction.