Insights into the area under the receiver operating characteristic curve (AUC) as a discrimination measure in species distribution modelling

By A. Jiménez‐Valverde. 2012

788 citations recorded by the source corpus.

Subfields: Ecological Modeling, Ecology, Nature and Landscape Conservation

View Paper: https://www.semanticscholar.org/paper/9a6b47bc19dba04a860c8367061d5cd4cf2b63d5

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecological Modeling

In Ecology

In Nature and Landscape Conservation

Citations in the corpus

A framework for species distribution modelling with improved pseudo-absence generation (2015, methodology): “However, since the317 potential distribution of the species is uncertain, we see no reason to pe-318 nalize the model based on the extent of the area predicted as suitable (see319 e.g. Jiménez-Valverde, 2012).”
Spatial bias in the GBIF database and its effect on modeling species' geographic distributions (2014, methodology): “…by modelling extent, for being not consistent with other evaluation criteria, and (if applied to presence-only data) for not representing “true” AUC (Lobo et al., 2008; Peterson et al., 2008; Jiménez-Valverde, 2011; Barve et al., 2011; L. Ballesteros-Mejia, I.J. Kitching & J. Beck, unpubl.).”
Minimum required number of specimen records to develop accurate species distribution models (2016, methodology): “2: red horizontal line), which is caused by our sampling strategy favoring locations with a higher given habitat suitability (Jiménez-Valverde 2012, Smith 2013).”
Using species distribution modelling to disentangle realised versus potential distributions for rare species conservation (2013, background): “Both presences and absences are necessary when modelling realised distributions (Jiménez-Valverde et al., 2008; Soberón and Nakamura, 2009; Jiménez-Valverde, 2012).”
High uncertainty in the effects of data characteristics on the performance of species distribution models (2021, background): “…discrimination metric most commonly used to evaluate SDM performance, it is also well-known to be unable to provide fair comparisons of model results across species with different data and eco-geographic characteristics (Hijmans, 2012; Jiménez-Valverde, 2012; Lobo et al., 2008; Phillips, 2008).”
Discrimination capacity in species distribution models depends on the representativeness of the environmental domain (2013, methodology): “However, without absence data for model testing, the application of discrimination Global Ecology and Biogeography, 22, 508–516, © 2012 Blackwell Publishing Ltd 509 measures such as the AUC is questionable (Jiménez-Valverde, 2012).”
The Matthews correlation coefficient (MCC) should replace the ROC AUC as the standard metric for assessing binary classification (2023, methodology): “The ROC AUC, however, suffers from multiple flaws and pitfalls [33, 34, 37–43] and does not inform us about positive predictive value and negative predictive value.”
MaxEnt versus MaxLike: empirical comparisons with ant species distributions (2013, methodology): “However, given the documented issues with AUC and pseudoabsence data (Lobo et al. 2008, Peterson et al. 2008, Jiménez-Valverde 2012), the interpretation of AUC is problematic.”
Improving species distribution model predictive accuracy using species abundance: Application with boosted regression trees (2020, result): “However, AUC has been criticized in several studies for giving misleading results for imbalanced data (Lobo et al., 2008; Peterson et al., 2008; Jimenez-Valverde, 2012).”
How to assess the prediction accuracy of species presence–absence models without absence data? (2013, methodology): “In addition, when background data are used in place of absence data, AUC is not truly informative (Jiménez-Valverde 2011) and new interpretations are needed (Peterson et al.”