| dc.description.abstract | Modeling suitable hábitats for biodiversity maintenance under changing environments may serve as an important tool for biological conservation purposes. Trindade and Martim Vaz is an oceanic archipelago located ~1.200km from the coast of Brazil, and has been under ecological succession since feral goat eradication, which have devastated its original vegetation. We modeled nesting hábitat suitability of a surfacenesting seabird - the Masked Booby Sula dactylatra - at Trindade Is., predicted its current nesting distribution, and proposed the future potential distribution under four possible vegetation scenarios on the island. We mapped 87 active nests of S. dactylatra during the breeding seasons of 2017 and 2019, from October to November at Trindade Is. We fitted nest occurrence points in response to five topographical (Elevation, Slope, Insulation, Aspect and Flow Length) and two vegetation (Succession and Remnant) variables, using an Ensemble Species Distribution Model (ESDM) of seven different algorithms: Boosted Regression Trees – GBM, Random Forests – RF, Generalized Linear Models – GLM, Generalized Additive Model – GAM, Artificial Neural Network – ANN, Multiple Adaptive Regression Spline – MARS, and Maximum Entropy - Maxent. We evaluated model accuracy based on two metrics: ROC and AUC, and tested for multicollinearity between variables using VIF test. Analyses were performed using the “biomod2” package in software R 3.6.1. Our models showed accuracy above 0.8 in both used metrics (ROC and TSS). Our results indicate areas of very high nesting hábitat suitability along the Southwest and Northwest faces of Trindade Is., mostly in elevations varying from 150-450 m. Elevation and Aspect were the variables that better explained S. dactylatra selection for nesting hábitats. Based on our predictions of distribution in response to possible vegetation scenarios, we found that the ecological succession will not affect the nesting distribution of S. dactylatra on the island substantially. Since S. dactylatra select areas of high elevations near cliff edges to nest, it is deductible that the predicted increase in the vegetation coverage does not reach these high areas because of geomorphological characteristics, such as elevation, soil composition and depth. We encourage further studies to address topographical variables of the finest spatial resolution spatial resolution as possible, as well biotic variables in order to model present and future species’ distribution. | en |
