A standard protocol for reporting species distribution models

Ecography, Journal Year: 2020, Volume and Issue: 43(9), P. 1261 - 1277

Published: June 1, 2020

Species distribution models (SDMs) constitute the most common class of across ecology, evolution and conservation. The advent ready‐to‐use software packages increasing availability digital geoinformation have considerably assisted application SDMs in past decade, greatly enabling their broader use for informing conservation management, quantifying impacts from global change. However, must be fit purpose, with all important aspects development applications properly considered. Despite widespread SDMs, standardisation documentation modelling protocols remain limited, which makes it hard to assess whether steps are appropriate end use. To address these issues, we propose a standard protocol reporting an emphasis on describing how study's objective is achieved through series modeling decisions. We call this ODMAP (Overview, Data, Model, Assessment Prediction) protocol, as its components reflect main involved building other empirically‐based biodiversity models. serves two purposes. First, provides checklist authors, detailing key model analyses, thus represents quick guide generic workflow modern SDMs. Second, introduces structured format documenting communicating models, ensuring transparency reproducibility, facilitating peer review expert evaluation quality, well meta‐analyses. detail elements ODMAP, explain can used different objectives applications, complements efforts store associated metadata define standards. illustrate utility by revisiting nine previously published case studies, provide interactive web‐based facilitate plan advance encouraging further refinement adoption scientific community.

Language: Английский

---

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

Ecology and Evolution, Journal Year: 2015, Volume and Issue: 6(1), P. 337 - 348

Published: Dec. 29, 2015

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 (max SSS and max F pb ) examine the effectiveness of threshold‐based prevalence estimation approach. Six virtual with varying were simulated within a real landscape in southeastern Australia. models built DOMAIN , generalized linear model, Maxent, Random Forest. Thresholds selected four presence‐only datasets different ratios number known presences random points ( KP – RP ratio ). Sensitivity, specificity, true skill statistic, measure used evaluate performance results. Species was estimated as predicted total evaluation dataset. varied changed. Datasets around 1 generally produced better than scores distant from 1. Results by We conclude that maxF had specificity too low very common using Forest Maxent models. contrast, consistent whichever dataset used. The almost always biased, bias large predictions. is affected datasets, but unaffected ratio. Unbiased estimations are difficult be determined

Language: Английский

---

Predictive performance of presence‐only species distribution models: a benchmark study with reproducible code

Ecological Monographs, Journal Year: 2021, Volume and Issue: 92(1)

Published: Oct. 8, 2021

Abstract Species distribution modeling (SDM) is widely used in ecology and conservation. Currently, the most available data for SDM are species presence‐only records (available through digital databases). There have been many studies comparing performance of alternative algorithms data. Among these, a 2006 paper from Elith colleagues has particularly influential field, partly because they several novel methods (at time) on global set that included independent presence–absence model evaluation. Since its publication, some further developed new ones emerged. In this paper, we explore patterns predictive across methods, by reanalyzing same (225 six different regions) using updated knowledge practices. We apply well‐established such as generalized additive models MaxEnt, alongside others received attention more recently, including regularized regressions, point‐process weighted random forests, XGBoost, support vector machines, ensemble framework biomod. All use include background samples (a sample environments landscape) fitting. impacts weights presence points introduce ways evaluating fitted to these data, area under precision‐recall gain curve, focusing rank results. find way matters. The top method was an tuned individual models. contrast, ensembles built biomod with default parameters performed no better than single moderate performing Similarly, second forest parameterized deal (contrasted relatively few records), which substantially outperformed other implementations. that, general, nonparametric techniques capability controlling complexity traditional regression MaxEnt boosted trees still among code working examples provided make study fully reproducible.

Language: Английский

---

blockCV: An r package for generating spatially or environmentally separated folds for k‐fold cross‐validation of species distribution models

Methods in Ecology and Evolution, Journal Year: 2018, Volume and Issue: 10(2), P. 225 - 232

Published: Oct. 13, 2018

Abstract When applied to structured data, conventional random cross‐validation techniques can lead underestimation of prediction error, and may result in inappropriate model selection. We present the r package block CV , a new toolbox for species distribution modelling. Although it has been developed with modelling mind, be used any spatial The generate spatially or environmentally separated folds. It includes tools measure autocorrelation ranges candidate covariates, providing user insights into structure these data. also offers interactive graphical capabilities creating blocks exploring data Package enables modellers more easily implement range evaluation approaches. will help community learn about impacts approaches on our understanding predictive performance models.

Language: Английский

---

A review of evidence about use and performance of species distribution modelling ensembles like BIOMOD

Diversity and Distributions, Journal Year: 2019, Volume and Issue: 25(5), P. 839 - 852

Published: Jan. 22, 2019

Abstract Aim The idea of combining predictions from different models into an ensemble has gained considerable popularity in species distribution modelling, partly due to free and comprehensive software such as the R package BIOMOD. However, despite proliferation models, we lack oversight how where they are used for modelling distributions, well perform. Here, present overview. Location Global. Methods Since BIOMOD is freely available widely by modellers, focused on articles that apply BIOMOD, filtering initial 852 papers identified our structured literature search a relevant final subset 224 eligible peer‐reviewed journal articles. Results BIOMOD‐based ensembles across many taxa locations, with terrestrial plants being most represented group ( n = 72) Europe continent 106). These studies often focus forecasting distributions future 109), commonly use presence‐only data 139) climatic environmental predictors 219). An average six ensembles, approximately half weight contributions their cross‐validation performance. discussion about choices made process unambiguous information performance versus individual limited. independent validate model particularly uncommon. Main conclusions We document breadth applications, but could not draw strong quantitative predictive reported. Understanding best when important enabling applications. To enable this objective be achieved, provide recommendations thorough reporting practices workflow.

Language: Английский

---

Climate change vulnerability assessment of species

Wiley Interdisciplinary Reviews Climate Change, Journal Year: 2018, Volume and Issue: 10(1)

Published: Oct. 11, 2018

Assessing species' vulnerability to climate change is a prerequisite for developing effective strategies conserve them. The last three decades have seen exponential growth in the number of studies evaluating how, how much, why, when, and where species will be impacted by change. We provide an overview rapidly field assessment (CCVA) describe key concepts, terms, steps considerations. stress importance identifying full range pressures, impacts their associated mechanisms that face using this as basis selecting appropriate approaches quantifying vulnerability. outline four CCVA approaches, namely trait‐based, correlative, mechanistic combined discuss use. Since any can deliver unreliable or even misleading results when incorrect data parameters are applied, we finding, selecting, applying input examples open‐access resources. Because rare, small‐range, declining‐range often particular conservation concern while also posing significant challenges CCVA, alternative ways assess CCVAs used inform IUCN Red List assessments extinction risk. Finally, suggest future directions propose areas research efforts may particularly valuable. This article categorized under: Climate, Ecology, Conservation > Extinction Risk

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

---

Testing whether ensemble modelling is advantageous for maximising predictive performance of species distribution models

Ecography, Journal Year: 2020, Volume and Issue: 43(4), P. 549 - 558

Published: Jan. 27, 2020

Predictive performance is important to many applications of species distribution models (SDMs). The SDM ‘ensemble’ approach, which combines predictions across different modelling methods, believed improve predictive performance, and used in recent studies. Here, we aim compare the ensemble that individual models, using a large presence–absence dataset eucalypt tree species. To test model divided our into calibration evaluation folds two spatial blocking strategies (checkerboard‐pattern latitudinal slicing). We calibrated cross‐validated all within folds, both repeated random division data (a common approach) blocking. Ensembles were built software package ‘biomod2’, with standard (‘untuned’) settings. Boosted regression (BRT) also fitted same data, tuned according published procedures. then ensembles against their component untuned BRTs. area under receiver‐operating characteristic curve (AUC) log‐likelihood for assessing performance. In tests, performed well, but not consistently better than or BRTs tests. Moreover, choosing best cross‐validation yielded good external blocked proving suited this choice, study, cross‐validation. slice was only possible four species; showed some particularly one, performing ensembles. This study shows no particular benefit over models. It suggests further robust testing required situations where are predict distant places environments.

Language: Английский

---

Without quality presence–absence data, discrimination metrics such as TSS can be misleading measures of model performance

Journal of Biogeography, Journal Year: 2018, Volume and Issue: 45(9), P. 1994 - 2002

Published: July 2, 2018

Abstract The discriminating capacity (i.e. ability to correctly classify presences and absences) of species distribution models ( SDM s) is commonly evaluated with metrics such as the area under receiving operating characteristic curve AUC ), Kappa statistic true skill TSS ). have been repeatedly criticized, but has fared relatively well since its introduction, mainly because it considered independent prevalence. In addition, discrimination contested they should be calculated on presence–absence data, are often used presence‐only or presence‐background data. Here, we investigate an alternative set metrics—similarity indices, also known F ‐measures. We first show that even in ideal conditions perfectly random sampling), can misleading dependence prevalence, whereas similarity/ ‐measures provide adequate estimations model capacity. Second, real‐world situations where sample prevalence different from biased sampling presence‐pseudoabsence), no metric provides estimation capacity, including specifically designed for modelling presence‐pseudoabsence Our conclusions twofold. First, unequivocally impel users understand potential shortcomings when quality data lacking, recommend obtaining specific case virtual species, which increasingly develop test methodologies, strongly use ‐measures, were not by contrary .

Language: Английский

---

Data Integration for Large-Scale Models of Species Distributions

Trends in Ecology & Evolution, Journal Year: 2019, Volume and Issue: 35(1), P. 56 - 67

Published: Nov. 2, 2019

With the expansion in quantity and types of biodiversity data being collected, there is a need to find ways combine these different sources provide cohesive summaries species' potential realized distributions space time. Recently, model-based integration has emerged as means achieve this by combining datasets that retain strengths each. We describe flexible approach using point process models, which convenient way translate across ecological currencies. highlight recent examples large-scale models based on outline conceptual technical challenges opportunities arise.

Language: Английский

---