Challenges in Identifying Sites Climatically Matched to the Native Ranges of Animal Invaders
Open Access
- 9 February 2011
- journal article
- research article
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 6 (2), e14670
- https://doi.org/10.1371/journal.pone.0014670
Abstract
Species distribution models are often used to characterize a species' native range climate, so as to identify sites elsewhere in the world that may be climatically similar and therefore at risk of invasion by the species. This endeavor provoked intense public controversy over recent attempts to model areas at risk of invasion by the Indian Python (Python molurus). We evaluated a number of MaxEnt models on this species to assess MaxEnt's utility for vertebrate climate matching. Overall, we found MaxEnt models to be very sensitive to modeling choices and selection of input localities and background regions. As used, MaxEnt invoked minimal protections against data dredging, multi-collinearity of explanatory axes, and overfitting. As used, MaxEnt endeavored to identify a single ideal climate, whereas different climatic considerations may determine range boundaries in different parts of the native range. MaxEnt was extremely sensitive to both the choice of background locations for the python, and to selection of presence points: inclusion of just four erroneous localities was responsible for Pyron et al.'s conclusion that no additional portions of the U.S. mainland were at risk of python invasion. When used with default settings, MaxEnt overfit the realized climate space, identifying models with about 60 parameters, about five times the number of parameters justifiable when optimized on the basis of Akaike's Information Criterion. When used with default settings, MaxEnt may not be an appropriate vehicle for identifying all sites at risk of colonization. Model instability and dearth of protections against overfitting, multi-collinearity, and data dredging may combine with a failure to distinguish fundamental from realized climate envelopes to produce models of limited utility. A priori identification of biologically realistic model structure, combined with computational protections against these statistical problems, may produce more robust models of invasion risk.Keywords
This publication has 55 references indexed in Scilit:
- Niches and distributional areas: Concepts, methods, and assumptionsProceedings of the National Academy of Sciences, 2009
- Do climate envelope models transfer? A manipulative test using dung beetle introductionsProceedings Of The Royal Society B-Biological Sciences, 2009
- Selecting pseudo-absence data for presence-only distribution modeling: How far should you stray from what you know?Ecological Modelling, 2009
- Do pseudo-absence selection strategies influence species distribution models and their predictions? An information-theoretic approach based on simulated dataBMC Ecology, 2009
- Claims of Potential Expansion throughout the U.S. by Invasive Python Species Are Contradicted by Ecological Niche ModelsPLOS ONE, 2008
- Predicting current and future biological invasions: both native and invaded ranges matterBiology Letters, 2008
- What parts of the US mainland are climatically suitable for invasive alien pythons spreading from Everglades National Park?Biological Invasions, 2008
- Predicting the potential distribution of the alien invasive American bullfrog (Lithobates catesbeianus) in BrazilBiological Invasions, 2007
- Habitat, environment and niche: what are we modelling?Oikos, 2006
- Generating surfaces of daily meteorological variables over large regions of complex terrainJournal of Hydrology, 1997