Dataset Information

On the selection of thresholds for predicting species occurrence with presence-only data.

ABSTRACT: Presence-only data present challenges for selecting thresholds to transform species distribution modeling results into binary outputs. In this article, we compare two recently published threshold selection methods (maxSSS and maxF pb) and examine the effectiveness of the threshold-based prevalence estimation approach. Six virtual species with varying prevalence were simulated within a real landscape in southeastern Australia. Presence-only models were built with DOMAIN, generalized linear model, Maxent, and Random Forest. Thresholds were selected with two methods maxSSS and maxF pb with four presence-only datasets with different ratios of the number of known presences to the number of random points (KP-RP ratio). Sensitivity, specificity, true skill statistic, and F measure were used to evaluate the performance of the results. Species prevalence was estimated as the ratio of the number of predicted presences to the total number of points in the evaluation dataset. Thresholds selected with maxF pb varied as the KP-RP ratio of the threshold selection datasets changed. Datasets with the KP-RP ratio around 1 generally produced better results than scores distant from 1. Results produced by We conclude that maxFpb had specificity too low for very common species using Random Forest and Maxent models. In contrast, maxSSS produced consistent results whichever dataset was used. The estimation of prevalence was almost always biased, and the bias was very large for DOMAIN and Random Forest predictions. We conclude that maxF pb is affected by the KP-RP ratio of the threshold selection datasets, but maxSSS is almost unaffected by this ratio. Unbiased estimations of prevalence are difficult to be determined using the threshold-based approach.

SUBMITTER: Liu C

PROVIDER: S-EPMC4716501 | biostudies-literature | 2016 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

On the selection of thresholds for predicting species occurrence with presence-only data.

Liu Canran C Newell Graeme G White Matt M

Ecology and evolution 20151229 1

Presence-only data present challenges for selecting thresholds to transform species distribution modeling results into binary outputs. In this article, we compare two recently published threshold selection methods (maxSSS and maxF pb) and examine the effectiveness of the threshold-based prevalence estimation approach. Six virtual species with varying prevalence were simulated within a real landscape in southeastern Australia. Presence-only models were built with DOMAIN, generalized linear model, ...[more]

PMID: 26811797

Similar Datasets

Project description:Sensitivity to habitat fragmentation often has been examined in terms of thresholds in landscape composition at which a species is likely to occur. Observed thresholds often have been low or absent, however, leaving much unexplained about habitat selection beyond initial thresholds of occurrence, even for species with strong habitat preferences. We examined responses to varying amounts of tree cover, a widely influential measure of habitat loss, for 40 woodland bird species in a mixed woodland/grassland landscape in eastern North Dakota, USA. We used LOESS smoothing to describe incidence for each species at three scales: within 200, 400, and 1200 m around sample locations. For the 200-m scale, we also calculated the most-preferred range of tree cover (within which at least half of observations were predicted to occur) for each species. Only 10 of 40 species had occurrence thresholds greater than about 10% tree cover. After initial occurrence, species showed three general patterns: some increased monotonically with tree cover; some increased up to an asymptote; some peaked at intermediate amounts of tree cover and then declined. These patterns approximate selection for interior woodlands and for edge-rich environments, but incidence plots provide greater detail in landscape-scale selection than do those categories. For most species, patterns persisted at larger scales, but for some, larger scales had distinctly different patterns than local scales. Preferred ranges of tree cover varied from <20% tree cover (common grackle, Quiscalus quiscula) to >60% (veery, Catharus fuscescens). We conclude that incidence patterns provide more information on habitat selection than do threshold measures for most species: in particular, they differentiate species preferring concentrated woodlands from those preferring mixed landscapes, and they show contrasting degrees of selectiveness. [Correction added on 16 October 2012, after first online publication: the Abstract section has been reworded].

Dataset Information

On the selection of thresholds for predicting species occurrence with presence-only data.

Publications

On the selection of thresholds for predicting species occurrence with presence-only data.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets