Table 2.
Importance of ecogeographic predictor variables used in species distribution models (SDMs) of Australian humpback dolphins in northern Ningaloo Marine Park over the entire survey period (May 2013—October 2015).
| SDM period | Model | Ecogeographic predictor variables | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Habitat type | Water depth | Slope | Seabed complexity | Distance to coast | Distance to boat ramp | Distance to reef crest | Distance to sanctuary zone | ||
| Entire | GAM | 0.179 | 0.412 | 0.23 | 0.116 | 0.412 | 0.051 | 0.094 | 0.019 |
| GBM | 0.037 | 0.499 | 0.129 | 0.085 | 0.266 | 0.02 | 0.045 | 0.004 | |
| CTA | 0.085 | 0.556 | 0.328 | 0.211 | 0.566 | 0.074 | 0.141 | 0.054 | |
| FDA | 0.051 | 0.737 | 0.108 | 0.057 | 0.184 | 0.000 | 0.046 | 0.003 | |
| RF | 0.046 | 0.239 | 0.139 | 0.091 | 0.176 | 0.041 | 0.055 | 0.022 | |
| MAXENT | 0.069 | 0.326 | 0.205 | 0.108 | 0.428 | 0.093 | 0.169 | 0.097 | |
| Mean of means | 0.078 | 0.462 | 0.190 | 0.111 | 0.339 | 0.047 | 0.092 | 0.033 | |
Variable importance is presented as the mean over 10 cross-validation runs of each modelling algorithm, and as the mean of means amongst them.
GAM generalised additive model, GBM generalised boosted model, CTA classification tree analysis, FDA flexible discriminant analysis, RF random forest, MAXENT maximum entropy. Environmental variables of greatest influence based on the randomisation procedure in biomod2 are highlighted in bold. For variable definitions see Supplementary Table S1.1 in Appendix S1.