Do stacked species distribution models reflect altitudinal diversity patterns?

Area of Science:

• Ecology and Biodiversity
• Computational Biology
• Plant Sciences

Background:

• Understanding species diversity patterns along elevation gradients is crucial for biodiversity conservation.
• Species distribution models (SDMs) are valuable tools for predicting biodiversity, but their performance can vary.
• Ensemble modeling approaches can improve the accuracy and robustness of SDM predictions.

Purpose of the Study:

• To evaluate stacked species distribution models for predicting alpha and gamma diversity of Anthurium and Bromeliaceae in the Andes.
• To compare the performance of different ensemble stacking methods and prediction types (original vs. binary).

Main Methods:

• Modeled distribution of 53 Anthurium and 89 Bromeliaceae species using six modeling techniques.
• Created ensemble models using average predictions from all or best techniques, and stacked original or binary predictions.
• Computed alpha and gamma diversity per 500m elevation band and compared model predictions against specialist-compiled data and literature.

Main Results:

• Stacked ensemble models showed good correlation with observed alpha diversity and literature-derived gamma diversity patterns along elevation.
• Models generally tended to overpredict species richness; stacking binary predictions reduced this overprediction.
• Randomization tests proved promising for evaluating stacked model performance.

Conclusions:

• Stacked ensemble species distribution models are a promising approach for modeling species assemblages and predicting diversity patterns.
• Further research is needed to refine stacking methods, particularly for binary predictions, to mitigate overprediction.
• The developed methods offer a robust framework for ecological forecasting and biodiversity assessments in mountainous regions.