Refining Uniform Discrimination Metrics: Towards a Case‐by‐Case Weighting Evaluation in Species Distribution Models With Presence‐Absence Data
Open Access
- 11 December 2025
- journal article
- Published by Wiley in Ecology and Evolution
- Vol. 15 (12) , e72573
Abstract
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.Keywords
Funding Information
- Consejo Superior de Investigaciones Científicas (2024ICT253)
This publication has 22 references indexed in Scilit:
- Sample size for the evaluation of presence-absence modelsEcological Indicators, 2020
- KnowBR: An application to map the geographical variation of survey effort and identify well-surveyed areas from biodiversity databasesEcological Indicators, 2018
- Habitat Suitability and Distribution Models2017
- What Teachers Should Know About the Bootstrap: Resampling in the Undergraduate Statistics CurriculumThe American Statistician, 2015
- Estimation of the area under the ROC curveStatistics in Medicine, 2002
- The GARP modelling system: problems and solutions to automated spatial predictionInternational Journal of Geographical Information Science, 1999
- A review of methods for the assessment of prediction errors in conservation presence/absence modelsEnvironmental Conservation, 1997
- Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicineClinical Chemistry, 1993
- An Introduction to the Bootstrap1993
- The area above the ordinal dominance graph and the area below the receiver operating characteristic graphJournal of Mathematical Psychology, 1975