Sample selection bias and presence-only distribution models: implications for background and pseudo-absence data.

By Steven J. Phillips, Miroslav Dud'ik, Jane Elith, Catherine H. Graham, Anthony Lehmann, J. Leathwick, Simon Ferrier. Ecological Applications, 2009

2,924 citations recorded by the source corpus.

Subfields: Ecological Modeling, Ecology, Nature and Landscape Conservation

View Paper: https://www.semanticscholar.org/paper/858f56b9861c0051a39ffe302cbbf88c93c7d3a0

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

Presence‐only modelling using MAXENT: when can we trust the inferences? (2013, methodology): “Phillips et al. (2009) suggest two ways of introducing sample selection bias into the background points.”
Correction note: Poisson point process models solve the “pseudo-absence problem” for presence-only data in ecology (2010, methodology): “…Leathwick (2007); Zarnetske, Edwards and Moisen (2007); Phillips et al. (2009)], and what spatial resolution to use in model-ﬁtting [Guisan et al. (2007); Elith and Leathwick (2009)], all issues that have natural solutions given a point process model speciﬁcation of the problem, as in Section 2.”
Use of Maximum Entropy Modeling in Wildlife Research (2009, background): “1), if the sampling distribution is expected to be strongly biased, background locations can be targeted from heavily sampled areas to provide unbiased results [18].”
Mapping the Global Potential Geographical Distribution of Black Locust (Robinia Pseudoacacia L.) Using Herbarium Data and a Maximum Entropy Model (2014, methodology): “Maximum entropy modeling uses only presence data as the basis of its predictions, unlike presence/absence models, and is therefore more valuable in regions where collecting absence points has been problematic or completely neglected [25,26].”
Predicting distributions of invasive species (2013, methodology): “…in unexpected environments rather than randomly), although in the equilibrium SDM literature, research on quantifying biases and methods for dealing with them in models is gradually emerging (e.g. Dorazio, 2014; Fithian et al., 2014; Hortal et al., 2008; Phillips et al., 2009; Warton et al., 2013).”
Without quality presence–absence data, discrimination metrics such as TSS can be misleading measures of model performance (2018, background): “Indeed, samplings of species presences are generally spatially biased (Phillips et al., 2009; Varela, Anderson, Garc í a ‐ Vald é s, & Fern á ndez ‐ Gonz á lez, 2014).”
Species distribution models of the Spotted Wing Drosophila ( Drosophila suzukii , Diptera: Drosophilidae) in its native and invasive range reveal an ecological niche shift (2018, background): “…Maxent only to draw background samples from regions (counties) with known occurrences in each specific model, restricting background sample selection to local areas that are assumed to have been surveyed for the species (Jarnevich & Young, 2015; Merow et al., 2013; Phillips et al., 2009).”
A strict maximum likelihood explanation of MaxEnt, and some implications for distribution modelling (2013, methodology): “…a taxonomic or ecological entity to which the targeted species belongs is used as background data, rather than the set of all uninformed background cells or a random selection of these cells (Elith & Leathwick 2007, Phillips et al. 2009, Williams et al. 2009, Mateo et al. 2010, Yates et al. 2010).”
Spatial filtering to reduce sampling bias can improve the performance of ecological niche models (2014, methodology): “, via a target group), it can be used in model calibration to correct for sampling bias (Anderson, 2012; Phillips et al., 2009).”
Uses and misuses of bioclimatic envelope modeling. (2012, background): “2011), the decision as to whether background or pseudo-absence data are warranted and how to select them (e.g., Phillips et al. 2009, VanDerWal et al. 2009a, Wisz and Guisan 2009), choice of methods to convert probabilities of occurrence or suitability scores into projections of species presence…”
Comparing sample bias correction methods for species distribution modeling using virtual species (2021, background): “…that sampling effort is uniform across the landscape such that environmental conditions are sampled proportionally to their relative availability (Hastie and Fithian 2013) and that a species’ niche is sampled over the full range of environmental conditions in which it occurs (Phillips et al. 2009).”
The area under the precision‐recall curve as a performance metric for rare binary events (2019, background): “…from our low prevalence virtual species based on occurrence locations of red brome Bromus rubens, and used a target- group background approach (Phillips et al., 2009) based on occurrence records of grass species exotic to the U.S. (see Supporting Information Figure S2) to mimic sampling…”
Benthos distribution modelling and its relevance for marine ecosystem management (2015, background): “As a consequence, sampling effort of marine biological data generally has a relatively low spatial resolution and is often biased towards shallow sites close to the coast (Phillips et al., 2009; Robinson et al., 2011) and driven towards politically, socially, and economically interesting areas (e.”