Ecological niche modeling in Maxent: the importance of model complexity and the performance of model selection criteria.

By Dan L. Warren, Stephanie N. Seifert. Ecological Applications, 2011

2,209 citations recorded by the source corpus.

Subfields: Ecological Modeling, Ecology, Nature and Landscape Conservation

View Paper: https://www.semanticscholar.org/paper/364950f1eb1c3078c55bc978d11791f80b8ba622

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

ENMeval 2.0: Redesigned for customizable and reproducible modeling of species’ niches and distributions (2021, methodology): “When users prefer to select models with AICc, which comes with a statistical caveat for Maxent (Warren & Seifert, 2011), ENMeval 2.0 can additionally quantify how well such models predict withheld data compared with null models (Kass et al., 2020).”
Predicting range shifts of Asian elephants under global change (2019, methodology): “…based on testing data of k‐fold cross‐validation bins; AUCDIFF, the difference between training and testing AUC, with high values for overfit models (Warren & Seifert, 2011); and two omission rates that quantify model overfitting with 0 and 10% omission rate of the training localities under which…”
ENVIREM: An expanded set of bioclimatic and topographic variables increases flexibility and improves performance of ecological niche modeling (2016, methodology): “AUCDIFF is the difference between the AUC calculated from training localities and AUCTEST, and is a measure of model overfitting, with higher values of AUCDIFF representing more overfit models (Warren and Seifert 2011).”
kuenm: an R package for detailed development of ecological niche models using Maxent (2019, methodology): “Finally, to evaluate model complexity, AICc, delta AICc, and AICc weights, are calculated; AICc values indicate how well models fit to the data while penalizing complexity to favor simple models ( Warren & Seifert, 2011 ).”
ENMeval: An R package for conducting spatially independent evaluations and estimating optimal model complexity for Maxent ecological niche models (2014, methodology): “The function also calculates the AICc value, DAICc and AICc weight for each full model, providing information on the relative quality of a model given the data (Burnham & Anderson 2004; Warren & Seifert 2011).”
Species-specific tuning increases robustness to sampling bias in models of species distributions: An implementation with Maxent (2011, methodology): “This process s similar in intent to selecting optimal models using the Akaike nformation Criterion (AICc; Warren and Seifert, 2011).”
Effect of the Maxent model’s complexity on the prediction of species potential distributions (2016, neutral): “…在物种分布点>10时使用, hinge feature在分布点>15 时使用, threshold和product feature在分布点>80时使用(Elith et al, 2010)。ENMeval数据包通过测试不同参数条件下Maxent模型修正的AIC值(即AICc)来评价模型的复杂度。AIC信息量准则(Akaike information criterion correction, AIC)是衡量统计模型拟合优良性的一种标准, 可以权衡所估计模型的复杂度和此模型拟合数据的优良性, AIC信息量准则优先考虑AIC值最小的模型(Akaike,…”
The challenge of modeling niches and distributions for data‐poor species: a comprehensive approach to model complexity (2018, methodology): “We first identified the optimal model based on AICc (corrected for small sample sizes; Warren and Seifert 2011), which scores models based on balancing complexity and goodness of fit.”
Prevalence, thresholds and the performance of presence–absence models (2014, background): “A recent simulation study (Warren & Seifert 2011) found that AIC was superior to AUC at uncovering the relationships underlying species distributions – a result that supports our emphasis on the importance of calibration assessment (AICmeasures calibration, but AUC does not).”