A guide to Bayesian model selection for ecologists

Ecological Monographs, Journal Year: 2014, Volume and Issue: 85(1), P. 3 - 28

Published: May 21, 2014

The steady upward trend in the use of model selection and Bayesian methods ecological research has made it clear that both approaches to inference are important for modern analysis models data. However, teaching working with our colleagues, we have noticed a general dissatisfaction available literature on multimodel inference. Students researchers new quickly find published advice is often preferential its treatment options analysis, frequently advocating one particular method above others. recent appearance many articles textbooks modeling provided welcome background relevant framework, but most these either very narrowly focused scope or inaccessible ecologists. Moreover, methodological details spread thinly throughout literature, appearing journals from different fields. Our aim this guide condense large body present specifically quantitative ecologists as neutrally possible. We also bring light few fundamental concepts relating directly seem gone unnoticed literature. Throughout, provide only minimal discussion philosophy, preferring instead examine breadth well their practical advantages disadvantages. This serves reference using methods, so they can better understand make an informed choice best aligned goals

Language: Английский

---

Modelling of species distributions, range dynamics and communities under imperfect detection: advances, challenges and opportunities

Ecography, Journal Year: 2016, Volume and Issue: 40(2), P. 281 - 295

Published: June 20, 2016

Building useful models of species distributions requires attention to several important issues, one being imperfect detection species. Data sets detections are likely suffer from false absence records. Depending on the type survey, positive records can also be a problem. Disregarding these observation errors may lead biases in model estimation as well overconfidence about precision. The severity problem depends intensity and how they correlate with environmental characteristics (e.g. where detectability strongly habitat features). A powerful modelling framework that accounts for has developed last 10–15 yr. Fundamental this is data must collected way informative process. For instance, such form multiple detection/non‐detection obtained visits/observers/detection methods at (at least) some sites, or times within survey visit. extend studying species’ range dynamics communities, approaches analysing abundance occupancy states (rather than binary presence/absence). This paper summarizes advances, discusses evidence effects difficulties working it, concludes current outlook future research application methods.

Language: Английский

---

Integrating multiple data sources in species distribution modeling: a framework for data fusion

Ecology, Journal Year: 2016, Volume and Issue: 98(3), P. 840 - 850

Published: Dec. 28, 2016

The last decade has seen a dramatic increase in the use of species distribution models (SDMs) to characterize patterns species' occurrence and abundance. Efforts parameterize SDMs often create tension between quality quantity data available fit models. Estimation methods that integrate both standardized non-standardized types offer potential solution tradeoff quantity. Recently several authors have developed approaches for jointly modeling two sources (one high one lesser quality). We extend their work by allowing explicit spatial autocorrelation detection error using Multivariate Conditional Autoregressive (MVCAR) model develop three share information less direct manner resulting more robust performance when auxiliary is quality. describe these new ("Shared," "Correlation," "Covariates") combining show case study Brown-headed Nuthatch Southeastern U.S. through simulations. All which used second source improved out-of-sample predictions relative single ("Single"). When quality, Shared performs best, but Correlation Covariates also perform well. performed better suggesting they are alternatives little known about collected opportunistically or citizen scientists. Methods allow be will maximize useful estimating distributions.

Language: Английский

---

The recent past and promising future for data integration methods to estimate species’ distributions

Methods in Ecology and Evolution, Journal Year: 2019, Volume and Issue: 10(1), P. 22 - 37

Published: Jan. 1, 2019

Abstract With the advance of methods for estimating species distribution models has come an interest in how to best combine datasets improve estimates distributions. This spurred development data integration that simultaneously harness information from multiple while dealing with specific strengths and weaknesses each dataset. We outline general principles have guided review recent developments field. then key areas allow a more framework integrating provide suggestions improving sampling design validation integrated models. Key advances been using point‐process thinking estimators developed different types. Extending this new types will further our inferences, as well relaxing assumptions about parameters are jointly estimated. These along better use regarding effort spatial autocorrelation inferences. Recent form strong foundation implementation Wider adoption can inferences distributions dynamic processes lead distributional shifts.

Language: Английский

---

A multispecies occupancy model for two or more interacting species

Methods in Ecology and Evolution, Journal Year: 2016, Volume and Issue: 7(10), P. 1164 - 1173

Published: June 29, 2016

Summary Species occurrence is influenced by environmental conditions and the presence of other species. Current approaches for multispecies occupancy modelling are practically limited to two interacting species often require assumption asymmetric interactions. We propose a model that can accommodate or more generalize single‐species assuming latent state multivariate Bernoulli random variable. probability each potential with both multinomial logit probit present details Gibbs sampler latter. As an example, we co‐occurrence probabilities bobcat ( Lynx rufus ), coyote Canis latrans grey fox Urocyon cinereoargenteus ) red Vulpes vulpes as function human disturbance variables throughout 6 Mid‐Atlantic states in eastern United States. found evidence pairwise interactions among most species, some pairs occupying same site varied along gradients; were independent at sites little disturbance, but these likely occur together high disturbance. Ecological communities composed multiple Our proposed method improves our ability draw inference from such permitting detection/non‐detection data arbitrary number without Additionally, permits variables. These advancements represent important improvement community‐level subject imperfect detection.

Language: Английский

---

Spatial autoregressive models for statistical inference from ecological data

Ecological Monographs, Journal Year: 2017, Volume and Issue: 88(1), P. 36 - 59

Published: Nov. 13, 2017

Abstract Ecological data often exhibit spatial pattern, which can be modeled as autocorrelation. Conditional autoregressive (CAR) and simultaneous (SAR) models are network‐based (also known graphical models) specifically designed to model spatially autocorrelated based on neighborhood relationships. We identify discuss six different types of practical ecological inference using CAR SAR models, including: (1) selection, (2) regression, (3) estimation autocorrelation, (4) other connectivity parameters, (5) prediction, (6) smoothing. compare showing their development connection partial correlations. Special cases, such the intrinsic (IAR), described. depend weight matrices, whose uses definition row‐standardization. Weight matrices also include covariates structures, we emphasize, but have been rarely used. Trends in harbor seals ( Phoca vitulina ) southeastern Alaska from 463 polygons, some with missing data, used illustrate types. develop a variety regression fit maximum likelihood Bayesian methods. Profile graphs for covariance parameters. The same set is both prediction smoothing, relative merits each discussed. show nonstationary variances correlations demonstrate effect several take‐home messages including choosing between IAR modeling effects matrix, appeal how handle isolated neighbors. highlight reasons why ecologists will want make use directly hierarchical not only explicit settings, more general models.

Language: Английский

---

Outstanding challenges and future directions for biodiversity monitoring using citizen science data

Methods in Ecology and Evolution, Journal Year: 2022, Volume and Issue: 14(1), P. 103 - 116

Published: Feb. 20, 2022

Abstract There is increasing availability and use of unstructured semi‐structured citizen science data in biodiversity research conservation. This expansion a rich source ‘big data’ has sparked numerous directions, driving the development analytical approaches that account for complex observation processes these datasets. We review outstanding challenges analysis monitoring. For many challenges, potential impact on ecological inference unknown. Further can document explore ways to address it. In addition outlining describing may be useful considering design future projects or additions existing projects. outline monitoring using four partially overlapping categories: arise as result (a) observer behaviour; (b) structures; (c) statistical models; (d) communication. Potential solutions are combinations of: collecting additional metadata; analytically combining different datasets; developing refining models. While there been important progress develop methods tackle most remain substantial gains subsequent conservation actions we believe will possible by further areas. The degree challenge opportunity each presents varies substantially across datasets, taxa questions. some cases, route forward clear, while other cases more scope exploration creativity.

Language: Английский

---

spOccupancy: An R package for single‐species, multi‐species, and integrated spatial occupancy models

Methods in Ecology and Evolution, Journal Year: 2022, Volume and Issue: 13(8), P. 1670 - 1678

Published: May 16, 2022

Abstract Occupancy modelling is a common approach to assess species distribution patterns, while explicitly accounting for false absences in detection–nondetection data. Numerous extensions of the basic single‐species occupancy model exist multiple species, spatial autocorrelation and integrate data types. However, development specialized computationally efficient software incorporate such extensions, especially large datasets, scarce or absent. We introduce spOccupancy R package designed fit multi‐species spatially explicit models. all models within Bayesian framework using Pólya‐Gamma augmentation, which results fast inference. provides functionality integration datasets via joint likelihood framework. The leverages Nearest Neighbour Gaussian Processes account autocorrelation, enables potentially massive (e.g. 1,000s–100,000s sites). user‐friendly functions simulation, fitting, validation (by posterior predictive checks), comparison (using information criteria k‐fold cross‐validation) out‐of‐sample prediction. illustrate package's vignette, simulated analysis two bird case studies. platform variety single models, making it straightforward address detection biases even datasets.

Language: Английский

---

Synergistic effects of climate and land‐use change influence broad‐scale avian population declines

Global Change Biology, Journal Year: 2019, Volume and Issue: 25(5), P. 1561 - 1575

Published: Feb. 27, 2019

Climate and land-use changes are expected to be the primary drivers of future global biodiversity loss. Although theory suggests that these factors impact species synergistically, past studies have either focused on only one in isolation or substituted space for time, which often results confounding between drivers. Tests synergistic effects require congruent time series animal populations, climate change replicated across landscapes span gradient correlations change. Using a unique high-resolution (measured as temperature precipitation) forest change) data, we show act synergistically influence bird population declines over 29 years Pacific Northwest United States. Nearly half examined had declined this time. Populations most response loss early seral mature forest, with responses amplified warmed In addition, birds more areas dried did not appear populations limited habitat loss, except when those were initially warmer than average landscape. Our provide some first empirical evidence dynamics, suggesting accelerated under pressure from multiple Furthermore, our findings suggest strong spatial variability impacts highlight need evaluate simultaneously avoid potential misattribution effects.

Language: Английский

---

Occupancy models for citizen‐science data

Methods in Ecology and Evolution, Journal Year: 2019, Volume and Issue: 10(1), P. 8 - 21

Published: Jan. 1, 2019

Abstract Large‐scale citizen‐science projects, such as atlases of species distribution, are an important source data for macroecological research, understanding the effects climate change and other drivers on biodiversity, more applied conservation tasks, early‐warning systems biodiversity loss. However, challenging to analyse because observation process has be taken into account. Typically, leads heterogeneous non‐random sampling, false absences, detections, spatial correlations in data. Increasingly, occupancy models being used atlas We advocate a dual approach strengthen inference from citizen science questions programme is intended address: (a) survey design should chosen with particular set associated analysis strategy mind (b) statistical methods tailored not only those but also specific characteristics review consequences choices that typically need made atlas‐style projects. These include resolution sampling units, allocation effort space, collection information about process. On side, we extensions basic frequently necessary data, including dealing heterogeneity, non‐independent violation closure assumption. New technologies, cell‐phone apps fixed remote detection devices, revolutionizing There opportunity maximize usefulness resulting datasets if protocols rooted robust designs issues considered. Our provides guidelines designing new projects overview current can

Language: Английский

---