Refining Uniform Discrimination Metrics: Towards a Case-by-Case Weighting Evaluation in Species Distribution Models With Presence-Absence Data.

Species distribution models are widely used in ecological research, but their validation remains challenging due to the representative effect. This effect, which reflects the strong dependence of discrimination performance on the distribution of suitability values, hampers the comparison and generalization of discrimination statistics across datasets. This study aims to address this issue by refining uniform discrimination metrics (e.g., uAUC and uSe*) to better harmonize evaluation scores and allow for biological interpretation of model performance differences. I propose an alternative method for calculating uniform discrimination scores that directly incorporates weights, eliminating the need for the resampling procedure in the original formulation. Through simulations, I demonstrate that this approach reduces bias and improves the coverage of 95% confidence intervals. Furthermore, the method provides a pathway to account for uncertainty associated with presence-absence data during model validation, offering a more robust evaluation framework. By addressing the representative effect and refining evaluation metrics, this study enhances the reliability of species distribution model validation. These improvements facilitate the meaningful comparison of model performance across datasets and support more accurate ecological interpretations.