Search

Selecting thresholds to convert continuous predictions of species distribution models proves critical for many real‐world applications and model assessments. Prevalent threshold selection methods for presence‐only data require unproven pseudo‐absence data or subjective researchers' decisions. This study proposes a new method, Boyce‐Threshold Quantile Regression (BTQR), to determine thresholds objectively without pseudo‐absence data. We summarize that the mutation point is a typical shape feature of the predicted‐to‐expected (P/E) curve after reviewing relevant articles. Analysis based on source‐sink theory suggests that this mutation point may represent a transition in habitat types and serve as an appropriate threshold. Threshold regression is introduced to accurately locate the mutation point. To validate the effectiveness of BTQR, we used four virtual species of varying prevalence and a real species with reliable distribution data. Six different species distribution models were employed to generate continuous suitability predictions. BTQR and nine other traditional methods transformed these continuous outputs into binary results. Comparative experiments show that BTQR has advantages in terms of accuracy, applicability, and consistency over the existing methods. [ABSTRACT FROM AUTHOR]

Aim: Spatial sampling bias (SSB) is a feature of opportunistically sampled species records. Species distribution models (SDMs) built using these data (i.e. presence‐background models) can produce biased predictions of suitability across geographic space, confounding species occurrence with the distribution of sampling effort. A wide range of SSB correction methods have been developed but simulations suggest effects on predictive performance are highly variable. Here, we aim to identify the SSB correction methods that have the highest likelihood of improving true predictive performance and evaluation strategies that provide a reliable indicator of model performance when independent test data are unavailable. Location: Global, simulation. Time Period: Current, simulation. Methods: A meta‐analysis was used to evaluate the performance of SSB correction methods in studies where there were direct comparisons between corrected and uncorrected SDMs. A simulation model was then developed to test evaluation strategies against a known truth using four common SSB correction methods. Results: Effect sizes from published studies suggest some support for small positive effects of SSB correction on predictive performance when assessed using independent test data, but this was not evident using internal cross‐validation and no single method stood out as consistently effective. Simulations support these findings and show that evaluation using internal test data was generally a poor indicator of the true effect of SSB correction. Methods that adjust models relative to a known driver of SSB produced the largest performance gains, but were also the most inconsistent. Main Conclusions: Correcting SSB in presence‐background SDMs without independent test data to evaluate the effect on model performance requires careful implementation. We recommend clearer documentation of SSB correction effects on SDMs, presenting results from models with and without correction, evaluating effects of different assumptions of SSB implementation on predictions, as well as greater efforts to collect independent test datasets to validate model predictions. [ABSTRACT FROM AUTHOR]

Species distribution modeling is a widely used technique for estimating the potential habitats of target organisms based on their environmental preferences. These methods serve as valuable tools for resource managers and conservationists, and their utilization is increasing, particularly in marine environments where data limitations persist as a challenge. In this study, we employed the global distribution predictions of six cold-water coral species as a case study to investigate various factors influencing predictions, including modeling algorithms, background points sampling strategies and sizes, and the collinearity of environmental datasets, using both discriminative and functional performance metrics. The choice of background sampling method exhibits a stronger influence on model performance compared to the effects of modeling algorithms, background point sampling size, and the collinearity of the environmental dataset. Predictions that utilize kernel density backgrounds, maintain an equal number of presences and background points for algorithms of BRT, RF, and MARS, and employ a substantial number of background points for MAXENT, coupled with a collinearity-filtered environmental dataset in species distribution modeling, yield higher levels of discriminative and functional performance. Overall, BRT and RF outperformed MAXENT, a conclusion that is further substantiated by the analysis of smoothed residuals and the uncertainty associated with the predicted habitat suitability of Madrepora oculata. This study offers valuable insights for enhancing species distribution modeling in marine benthic environments, thereby benefiting resource management and conservation strategies for benthic species. [ABSTRACT FROM AUTHOR]

Abstract Selecting thresholds to convert continuous predictions of species distribution models proves critical for many real‐world applications and model assessments. Prevalent threshold selection methods for presence‐only data require unproven pseudo‐absence data or subjective researchers' decisions. This study proposes a new method, Boyce‐Threshold Quantile Regression (BTQR), to determine thresholds objectively without pseudo‐absence data. We summarize that the mutation point is a typical shape feature of the predicted‐to‐expected (P/E) curve after reviewing relevant articles. Analysis based on source‐sink theory suggests that this mutation point may represent a transition in habitat types and serve as an appropriate threshold. Threshold regression is introduced to accurately locate the mutation point. To validate the effectiveness of BTQR, we used four virtual species of varying prevalence and a real species with reliable distribution data. Six different species distribution models were employed to generate continuous suitability predictions. BTQR and nine other traditional methods transformed these continuous outputs into binary results. Comparative experiments show that BTQR has advantages in terms of accuracy, applicability, and consistency over the existing methods.

Biodiversity community science projects are growing rapidly in popularity. The enormous amounts of data generated by these programs are transforming how we conduct ecological research and conservation management. However, as with other biodiversity surveys, community science datasets suffer from biases in time and locations of observations. To better use these data, we modeled the spatial biases present in the popular community science platform, iNaturalist. iNaturalist uses crowdsourcing to collect georeferenced and time‐stamped observations of all taxa worldwide. With its wealth of biodiversity data, iNaturalist is now being used to answer a broad range of questions in ecology and conservation, but little is known about the platform's spatial biases. We focus on the more than 1.75 million iNaturalist observations available (as of December 2021) from British Columbia, Canada, a region with a strong community science presence and diversity of ecosystems. Using machine learning and species distribution modeling, we examined which landscape factors (e.g., protected areas, roads, human population density, habitat zones, elevation) were most important in determining where observations are taken, and we created a predicted probability map revealing how likely different regions are to be sampled by community scientists. We found strong road biases for observations in iNaturalist, with over 94% of observations within 1 km of roads. In addition, human population density and broad habitat ecosystem zones played a large role in predicting where iNaturalist observations occur across the landscape. These methods demonstrate tools for modeling the effects of spatial biases in large opportunistic datasets that can then be used to produce more accurate species distribution and biodiversity models from community science data. [ABSTRACT FROM AUTHOR]

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

Species distribution modeling (SDM) plays a crucial role in investigating habitat suitability and addressing various ecological issues. While likelihood analysis is commonly used to draw ecological conclusions, its statistical performance is not robust when confronted with slight deviations due to model misspecification in SDM. We proposed a new robust estimation method based on a novel divergence for the Poisson point process model. The method is characterized by weighting the log-likelihood equation to mitigate the impact of heterogeneous observations in the presence-only data, which can result from model misspecification. We demonstrated that our proposed method improved the predictive performance of maximum likelihood estimation in both simulation studies and the analysis of vascular plant data in Japan. • We introduced a robust method to address model misspecification in species distribution using thinned Poisson processes. • Simulation studies revealed bias in conventional estimates for mixed data, while our method provided appropriate estimates. • Our method improved prediction performance of the likelihood approach in analyzing presence-only data of vascular plants. [ABSTRACT FROM AUTHOR]

Aim: Large databases of species records such as those generated through citizen science projects, archives or museum collections are being used with increasing frequency in species distribution modelling (SDM) for conservation and land management. Despite the broad spatial and temporal coverage of the data, its application is often limited by the issue of sampling bias and consequently, zero inflation; there are more zeros (which are potentially 'false absences') in the data than expected. Here, we demonstrate how pooling species presence data into a 'pseudo‐abundance' count can allow identification and removal of sampling bias through the use of zero‐inflated (ZI) models, and thus solve a common SDM problem. Location: All locations Taxon: All taxa Methods: We present the results of a series of simulations based on hypothetical ecological scenarios of data collection using random and non‐random sampling strategies. Our simulations assume that the locations of occurrence records are known at a high spatial resolution, but that the absence of occurrence records may reflect under‐sampling. To simulate pooling of presence–absence or presence‐only data, we count occurrence records at intermediate and coarse spatial resolutions, and use ZI models to predict the counts (species abundance per grid cell) from environmental layers. Results: ZI models can successfully identify predictors of bias in species data and produce abundance prediction maps that are free from that bias. This phenomenon holds across multiple spatial scales, thereby presenting an advantage over presence‐only SDM methods such as binomial GLMs or MaxEnt, where information about species density is lost, and model performance declines at coarser scales. Main Conclusions: Our results highlight the value of converting presence–absence or presence‐only species data to 'pseudo‐abundance' and using ZI models to address the problem of sampling bias. This method has huge potential for ecological researchers when using large species datasets for research and conservation. [ABSTRACT FROM AUTHOR]

Aim: To evaluate the utility of opportunistic data from citizen science programmes for forecasting species distributions against forecasts with a model of individual‐based population dynamics. Location: Sweden. Methods: We evaluated whether alternative methods for building habitat suitability models (HSMs) based on opportunistic data from citizen science programmes produced forecasts that were consistent with forecasts from two benchmark models: (1) a HSM based on data from systematic monitoring and (2) an individual‐based model for spatially explicit population dynamics based on empirical demographic and movement data. We forecasted population numbers and habitat suitability for three realistic, future forest landscapes for a forest bird, the Siberian jay (Perisoreus infaustus). We ranked simulated forest landscapes with respect to their benefits to Siberian jays for each modelling method and compared the agreement of the rankings among methods. Results: Forecasts based on our two benchmark models were consistent with each other and with expectations based on the species' ecology. Forecasts from logistic regression models based on opportunistic data were consistent with the benchmark models if species detections were combined with high‐quality inferred absences derived via retrospective interviews with experienced "super‐reporters." In contrast, forecasts with three other widely used methods were inconsistent with the benchmark models, sometimes with misleading rankings of future scenarios. Main conclusions: Our critical evaluation of alternative HSMs against a spatially explicit IBM demonstrates that information on species absences critically improves forecasts of species distributions using opportunistic data from citizen science programmes. Moreover, high‐quality information on species absences can be retrospectively inferred from surveys of the consistency of reporting of individual species and the identification skills of participating reporters. We recommend that citizen science projects incorporate procedures to evaluate reporting behaviour. Inferred absences may be especially useful for improving forecasts for species and regions poorly covered by systematic monitoring schemes. [ABSTRACT FROM AUTHOR]

Climate-only species distribution models are commonly used to evaluate a species' vulnerability to future climate change. However, these models may overestimate a species' vulnerability when occurrence data comes from only a subset of the species' historical range due to anthropogenic range contractions. This study investigated the impacts of using climate-only species distribution models with anthropogenically truncated occurrence data on future predictions, using the little-studied leopard cat (Prionailurus bengalensis) of mainland Tropical Asia as a case study. Like most other species in mainland Tropical Asia, the leopard cat's current distribution represents a non-random subset of its historical range due to the disproportionate loss of lowland forests in the region. We found that climate-only species distribution models trained on current occurrence data significantly overestimated the leopard cat's vulnerability to future climate change, predicting considerable range contractions by 2081–2100 due to rising summer temperatures. A similar range contraction was observed using a virtual species whose occurrence was solely predicated on extant forest cover and not climate. This suggested the leopard cat's apparent vulnerability to future climate change was driven by a spurious relationship with summer temperatures, as modern occurrence data primarily came from refugial forest cover in the region's cooler uplands rather than the warmer, human-dominated lowlands. This bias was not apparent, however, from our current predictions, which had high predictive accuracy and were spatially consistent with current knowledge of the species' distribution; an observation highlighting the inadequacy of predictive accuracy alone as a metric for selecting models in studies where transferability to novel environments is critical. We next considered two potential strategies for detecting and mitigating the effects of anthropogenically truncated occurrence data: 1) the use of hybrid models that incorporate both climatic and non-climatic variables, and 2) the use of climate-only models with historical datasets. Despite using different data, both proposed strategies produced spatially similar predictions of present and future leopard cat occurrence with little-to-no range contractions predicted. We thus recommend all species distribution modeling studies interested in assessing a species' vulnerability to climate change compare future predictions derived from climate-only and hybrid models as well as compare predictions from current and historical datasets, as differences in predictions can highlight the presence of potential biases and other limitations. Finally, we discuss our findings as they apply to both our focal species and to the overall field of species distribution modeling. [ABSTRACT FROM AUTHOR]

Ensemble habitat selection modeling is becoming a popular approach among ecologists to answer different questions. Since we are still in the early stages of development and application of ensemble modeling, there remain many questions regarding performance and parameterization. One important gap, which this paper addresses, is how the number of background points used to train models influences the performance of the ensemble model. We used an empirical presence-only dataset and three different selections of background points to train scale-optimized habitat selection models using six modeling algorithms (GLM, GAM, MARS, ANN, Random Forest, and MaxEnt). We tested four ensemble models using different combinations of the component models: (a) equal numbers of background points and presences, (b) background points equaled ten times the number of presences, (c) 10,000 background points, and (d) optimized background points for each component model. Among regression-based approaches, MARS performed best when built with 10,000 background points. Among machine learning models, RF performed the best when built with equal presences and background points. Among the four ensemble models, AUC indicated that the best performing model was the ensemble with each component model including the optimized number of background points, while TSS increased as the number of background points models increased. We found that an ensemble of models, each trained with an optimal number of background points, outperformed ensembles of models trained with the same number of background points, although differences in performance were slight. When using a single modeling method, RF with equal number of presences and background points can perform better than an ensemble model, but the performance fluctuates when the number of background points is not properly selected. On the other hand, ensemble modeling provides consistently high accuracy regardless of background point sampling approach. Further, optimizing the number of background points for each component model within an ensemble model can provide the best model improvement. We suggest evaluating more models across multiple species to investigate how background point selection might affect ensemble models in different scenarios. [ABSTRACT FROM AUTHOR]