Cross-validation of species distribution models: removing spatial sorting bias and calibration with a null model.

By Robert J. Hijmans. Ecology, 2012

641 citations recorded by the source corpus.

Subfields: Ecological Modeling, Ecology, Nature and Landscape Conservation

View Paper: https://www.semanticscholar.org/paper/ab13ff954417a4e2fcc29d18428aebb4ac7fe985

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

Incorporating model complexity and spatial sampling bias into ecological niche models of climate change risks faced by 90 California vertebrate species of concern (2014, methodology): “removing the effects of bias may reduce AUC scores on both training and test data when presence-only data are used for model validation (Phillips et al., 2009; Hijmans, 2012) even”
Evaluating presence‐only species distribution models with discrimination accuracy is uninformative for many applications (2019, methodology): “…the underlying biology and suitability of habitat boundaries or poorly justified assumptions about dispersal), and can affect both the inferred model (Acevedo et al., 2012) and the per‐ formance of metrics used to evaluate models (Acevedo et al., 2012; Hijmans, 2012; Jimenez‐Valverde et al., 2013).”
Protected areas network is not adequate to protect a critically endangered East Africa Chelonian: Modelling distribution of pancake tortoise, Malacochersus tornieri under current and future climates (2020, methodology): “Disadvantages of using cross validation, are that only part of the data can be used for model fitting and that it is challenging to obtain test data that is statistically (spatially) independent of training data (but see Hijmans 2012, Wenger and Olden 2012).”
Uncertainty associated with survey design in Species Distribution Models (2014, methodology): “In general, complex techniques generate models with better performance within the training data (Elith et al., 2006; Tsoar et al., 2007; but see Lobo, 2008; Hijmans, 2012).”
Mechanistic and Correlative Models of Ecological Niches (2015, background): “, spatial bias to accessible areas, incomplete sampling across an area, incomplete detectability of individuals) will have considerable potential to translate into new biases and problems in model outcomes (Hijmans 2012; Kramer-Schadt et al. 2013; Boria et al. 2014).”
ENMeval: An R package for conducting spatially independent evaluations and estimating optimal model complexity for Maxent ecological niche models (2014, background): “These tools, however, are subject to a number of methodological issues including the challenge of balancing goodness-of-fit with model complexity (Warren & Seifert 2011), and the need to evaluate model performance with independent data (Veloz 2009; Hijmans 2012).”
Comparing sample bias correction methods for species distribution modeling using virtual species (2021, background): “…space are sampled at higher intensities than others (Bean et al. 2012, Gelfand et al. 2012), thereby causing model coefficients to become more representative of areas that have been heavily sampled than of a species’ entire range (Hortal et al. 2008, Hijmans 2012, GuilleraArroita et al. 2015).”
Selecting predictors to maximize the transferability of species distribution models: lessons from cross‐continental plant invasions (2017, background): “…to minimize the spatial autocorrelation problem, increasing the ratio of 374 independent data in the split-sampling evaluation, including a spatial autocorrelation term or 375 disaggregating the calibration dataset based on a minimum distance can be alternatives (Dormann, 376 2007, Hijmans, 2012).”
KnowBR: An application to map the geographical variation of survey effort and identify well-surveyed areas from biodiversity databases (2018, background): “Hence, complex SDM algorithms calibrated with data containing background absences yield poor and inconsistent predictions, a fact that often passes unnoticed due to the use of inadequate evaluation methods (Hijmans, 2012).”
Habitat modelling of tracking data from multiple marine predators identifies important areas in the Southern Indian Ocean (2018, methodology): “AUC scores calculated from random cross- validation are usually overoptimistic due to spatial autocorrelation (Hijmans, 2012), but our cross- validation scheme should result in AUC scores which reflect the generalized model performance more accurately.”
What do we gain from simplicity versus complexity in species distribution models (2014, background): “Recent studies have emphasized that care should also be taken in how data is partitioned into training, evaluation and test data, in particular to control for spatial autocorrelation (Latimer et al. 2006, Dormann et al. 2007, Veloz 2009, Hijmans 2012; see below for more details).”
Crop wild relatives of pigeonpea [Cajanus cajan (L.) Millsp.]: Distributions, ex situ conservation status, and potential genetic resources for abiotic stress tolerance (2015, methodology): “15.1 (R Core Team, 2013), utilizing packages maptools (Bivand and Lewin-Koh, 2014), rgdal (Bivand et al., 2014), SDMTools (van der Wal et al., 2014), raster (Hijmans, 2014), sp (Bivand et al., 2013; Pebesma and Bivand, 2005), dismo (Hijmans et al., 2013), and plsdepot (Sanchez, 2012).”
Paintings predict the distribution of species, or the challenge of selecting environmental predictors and evaluation statistics (2018, background): “Many alternative metrics have been proposed to evaluate SDMs (see for example Allouche, Tsoar, & Kadmon, 2006; Hijmans, 2012; Phillips & Elith, 2010).”
High uncertainty in the effects of data characteristics on the performance of species distribution models (2021, methodology): “Variations in SDM performance have been attributed to the differences in the spatial resolutions of the analyses, the particular SDM technique used and the evaluation metric used to assess model performance (Hijmans, 2012; Marmion et al., 2009; Santika, 2011; Wogan, 2016).”