From descriptive to predictive distribution models: a working example with Iberian amphibians and reptiles

Area of Science:

• Ecology
• Biogeography
• Conservation Biology

Background:

• Investigated the utility of spatial-environmental models for understanding species distributions.
• Focused on predictive performance as a key criterion for model evaluation.
• Utilized atlas data and ecological variables for amphibian and reptile species in Portugal.

Purpose of the Study:

• To determine conditions for effective spatial-environmental modeling of species distributions.
• To assess model predictive performance using explicit criteria.
• To evaluate model extrapolation capabilities across geographical regions.

Main Methods:

• Constructed distribution models using stepwise logistic regression and GIS.
• Included 17 amphibian and 21 reptile species from Portugal.
• Extrapolated models to Spain and validated against range maps and atlas data.

Main Results:

• Model precision varied: 'fair' to 'very good' for edge species (kappa 0.35-0.89), 'moderate' for non-edge species (kappa 0.03-0.54).
• Prediction accuracy in Spain correlated with descriptive model precision for edge species.
• Search-effort data outperformed environmental data for non-edge species.

Conclusions:

• Presence-absence atlas data are insufficient for modeling species distributions without range borders.
• High-precision models for edge species aid in identifying distribution drivers.
• Informed conservation measures benefit from accurate species distribution modeling.