Scholarlink

Context <p>Estimating the distribution of rare species is an important component of conservation. The occurrence of such species is often only known from few presence-only observations and the spatial scales at which they select resources is poorly understood. Scientists often select arbitrary discrete spatial scales for estimating distribution models; however, resource selection follows a hierarchical multiscale process and selection of an incorrect scale may bias model inference and lead to misallocation of resources.</p> Objectives <p>Our objectives were to develop a Bayesian latent indicator scale selection (BLISS) model capable of estimating spatial scales of resource selection using small presence-only datasets and apply our model to empirical data characteristic of species of conservation concern.</p> Methods <p>We developed a BLISS model using the resource selection function (BLISS<sub>RSF</sub>) and conducted a simulation study to evaluate its ability to predict spatial scales, predictor coefficients, and proportional probability of use under constraints typical of environmental presence-only data, including small occurrence dataset, variable pseudo-absence sample size, pseudo-absence contamination, and spatial scale autocorrelation. We applied BLISS<sub>RSF</sub> to estimate winter resource selection for little brown (<i>Myotis lucifugus</i>), northern long-eared (<i>M. septentrionalis</i>), and tricolored (<i>Perimyotis subflavus</i>) bats in the Midwest, USA.</p> Results <p>Simulations demonstrated that BLISS<sub>RSF</sub> accurately estimated model parameters when using 10,000 pseudo-absences under constraints of scale autocorrelation and contamination. BLISS<sub>RSF</sub> yielded predictive models of winter resource selection for each bat species and high fit with cross-validation datasets.</p> Conclusions <p>As broad-scale population declines precipitate, the ability to predict species distributions may be more reliant on presence-only data, particularly for species that are difficult to sample or those which lack adequate support for rigorous field study. BLISS<sub>RSF</sub> is an adaptable and computationally efficient method for estimating RSF parameters using small presence-only datasets under a used–available study design.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Bayesian method for estimating multiscale resource selection using presence-only data: a case study predicting winter distributions for little brown, northern long-eared, and tricolored bats

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年3月24日

查看全文
  • Dan J. Kaminski,
  • Kelly E. Poole,
  • Tyler M. Harms,
  • Amber J. Andress

摘要

Context

Estimating the distribution of rare species is an important component of conservation. The occurrence of such species is often only known from few presence-only observations and the spatial scales at which they select resources is poorly understood. Scientists often select arbitrary discrete spatial scales for estimating distribution models; however, resource selection follows a hierarchical multiscale process and selection of an incorrect scale may bias model inference and lead to misallocation of resources.

Objectives

Our objectives were to develop a Bayesian latent indicator scale selection (BLISS) model capable of estimating spatial scales of resource selection using small presence-only datasets and apply our model to empirical data characteristic of species of conservation concern.

Methods

We developed a BLISS model using the resource selection function (BLISSRSF) and conducted a simulation study to evaluate its ability to predict spatial scales, predictor coefficients, and proportional probability of use under constraints typical of environmental presence-only data, including small occurrence dataset, variable pseudo-absence sample size, pseudo-absence contamination, and spatial scale autocorrelation. We applied BLISSRSF to estimate winter resource selection for little brown (Myotis lucifugus), northern long-eared (M. septentrionalis), and tricolored (Perimyotis subflavus) bats in the Midwest, USA.

Results

Simulations demonstrated that BLISSRSF accurately estimated model parameters when using 10,000 pseudo-absences under constraints of scale autocorrelation and contamination. BLISSRSF yielded predictive models of winter resource selection for each bat species and high fit with cross-validation datasets.

Conclusions

As broad-scale population declines precipitate, the ability to predict species distributions may be more reliant on presence-only data, particularly for species that are difficult to sample or those which lack adequate support for rigorous field study. BLISSRSF is an adaptable and computationally efficient method for estimating RSF parameters using small presence-only datasets under a used–available study design.