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

Abstract

Abstract. Most methods for modeling species distributions from occurrence records require additional data representing the range of environmental conditions in the modeled region. These data, called background or pseudo-absence data, are usually drawn at random from the entire region, whereas occurrence collection is often spatially biased toward easily accessed areas. Since the spatial bias generally results in environmental bias, the difference between occurrence collection and background sampling may lead to inaccurate models. To correct the estimation, we propose choosing background data with the same bias as occurrence data. We investigate theoretical and practical implications of this approach. Accurate information about spatial bias is usually lacking, so explicit biased sampling of background sites may not be possible. However, it is likely that an entire target group of species observed by similar methods will share similar bias. We therefore explore the use of all occurrences within a target group as biased background data. We compare model performance using target-group background and randomly sampled background on a comprehensive collection of data for 226 species from diverse regions of the world. We find that target-group background improves average performance for all the modeling methods we consider, with the choice of background data having as large an effect on predictive performance as the choice of modeling method. The performance improvement due to target-group background is greatest when there is strong bias in the target-group presence records. Our approach applies to regression-based modeling methods that have been adapted for use with occurrence data, such as generalized linear or additive models and boosted regression trees, and to Maxent, a probability density estimation method. We argue that increased awareness of the implications of spatial bias in surveys, and possible modeling remedies, will substantially improve predictions of species distributions.

Differential effects of climate and species interactions on range limits at a hybrid zone: potential direct and indirect impacts of climate change

“…Because presence‐only SDM methods assume unbiased, random sampling (Phillips et al. 2009; Yackulic et al. 2013), we took multiple measures to correct for any sampling bias in our occurrence dataset.…”

Designing protected area networks that translate international conservation commitments into national action

“…Our inputs to MaxEnt were species locality, and topographic and meteorological data (Table A3). Survey data are usually spatially biased and can cause inaccuracies if this bias is not properly accounted for when inputted into distribution modelling (Phillips et al, 2009). Therefore, to improve model predictions and reduce errors associated with survey effort bias (Syfert et al, 2013), we constructed species-specific bias grids in R (R Core Team, 2015).…”

Optimizing resiliency of reserve networks to climate change: multispecies conservation planning in the Pacific Northwest, USA

“…To create spotted owl distribution models, we used a dataset from Carroll & Johnson (2008) containing locations of owl nest sites or activity centers derived from digital databases of surveys of owl occupancy and reproductive status from the late 1980s through 2000 for Oregon and Washington and primarily (93%) from 1987 to 2006 for California. Spatially biased survey effort typical of found data presents a major challenge to distribution modeling (Phillips et al, 2009). To reduce this problem, we thinned data to achieve a minimum separation for each species of 1 km between locations, using a geographic information system routine that identified clusters of adjacent records and then reduced the set of such records to one record randomly selected from that set.…”