Ensemble species distribution modelling with transformed suitability values. (February 2018)

• Record Type: Journal Article
• Authors: Kindt, R.
• Is Part Of: Environmental modelling & software. Volume 100(2018)
• Page Start: 136
• Page End: 145
• Publication Date: 2018-02
• Subjects: Species distribution model -- Ensemble model -- Spatial sorting bias -- R statistical language and environment -- Ecological software -- BiodiversityR package; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2017.11.009
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Species distribution modelling (SDM) was integrated in version 2.0 of the BiodiversityR package released in 2012. Ensemble habitat suitability is calculated as the weighted average of suitabilities predicted by different algorithms. Advanced options for SDM in the current version (2.8–4) of the package include tuning the best combination of the number and weights of models contributing to the ensemble suitability and calculating the absence-presence threshold as the average or minimum of recommended threshold values. Algorithm-specific suitability values can be transformed via generalized linear models with probit link so that they become more similar in range. Other options include reducing spatial sorting bias by selecting background locations in circular neighbourhoods and generating suitability maps that show the number of contributing models that predict species presence. The approaches are illustrated for two species with open-access point location data sets, Bradypus variegatus and Thryothorus ludovicianus . Highlights: Ensemble suitability in BiodiversityR calculated from 23 contributing algorithms. Ensemble suitability also calculated as count of algorithms predicting presence. Parameter settings such as formulae can easily be modified from default settings. Ensembles are among the best or the best algorithms for suitability modelling. Transformed suitability values are more similar in range and indicate overfitting.