Presence-only Data Research Articles

Mapping suitability for rice production in inland valley landscapes in Benin and Togo using environmental niche modeling

Inland valleys (IVs) in Africa are important landscapes for rice cultivation and are targeted by national governments to attain self-sufficiency. Yet, there is limited information on the spatial distribution of IVs suitability at the national scale. In the present study, we developed an ensemble model approach to characterize the IVs suitability for rainfed lowland rice using 4 machine learning algorithms based on environmental niche modeling (ENM) with presence-only data and background sample, namely Boosted Regression Tree (BRT), Generalized Linear Model (GLM), Maximum Entropy (MAXNT) and Random Forest (RF). We used a set of predictors that were grouped under climatic variables, agricultural water productivity and soil water content, soil chemical properties, soil physical properties, vegetation cover, and socio-economic variables. The Area Under the Curves (AUC) evaluation metrics for both training and testing were respectively 0.999 and 0.873 for BRT, 0.866 and 0.816 for GLM, 0.948 and 0.861 for MAXENT and 0.911 and 0.878 for RF. Results showed that proximity of inland valleys to roads and urban centers, elevation, soil water holding capacity, bulk density, vegetation index, gross biomass water productivity, precipitation of the wettest quarter, isothermality, annual precipitation, and total phosphorus among others were major predictors of IVs suitability for rainfed lowland rice. Suitable IVs areas were estimated at 155,000–225,000 Ha in Togo and 351,000–406,000 Ha in Benin. We estimated that 53.8% of the suitable IVs area is needed in Togo to attain self-sufficiency in rice while 60.1% of the suitable IVs area is needed in Benin to attain self-sufficiency in rice. These results demonstrated the effectiveness of an ensemble environmental niche modeling approach that combines the strengths of several models.

Using MaxEnt algorithm to assess habitat suitability of a potential Iberian lynx population in central Iberian Peninsula

Iberian lynx distribution is currently restricted to the south of the Iberian Peninsula. Nevertheless, there is evidence of the presence of several small groups in the peninsular centre that have been forgotten by management and conservation actions. In this research, we gathered evidences of Iberian lynx presence along 21 transects located in the southwest of the Madrid province. In these transects lynx DNA was identified in 47 scats, which scientifically proves the presence of the species in that location. Using these locations (presence-only data) we built a maximum entropy model (MaxEnt) to estimate the suitability of the study area for the species. Our results show the existence of an almost continuous area that is approximately 744 km2 that is suitable for the Iberian lynx. Seventy-eight percent of this area is within the Natura 2000 network and, therefore, it falls under regulations to preserve and restore habitat types, flora and fauna. This study shows the suitability of this territory has for the Iberian lynx.

Land use and cover effects on an ecosystem engineer

Human induced land-use change in the form of urbanization and agriculture are rapidly transforming our planet’s terrestrial landscapes and causing the loss of species at unprecedented rates. Studies documenting the impacts of development are extensive; however, less is known about the scale at which species respond to development and specific features of altered landscapes that make them vulnerable. This shortcoming may be due to the difficulty of collecting data across large spatial extents, but filling these information gaps are critical, particularly for species that have a disproportionate influence on ecosystems, such as ecosystem engineers. One such species, the southeastern pocket gopher (Geomys pinetis), has been declining throughout its range (southeastern United States) and continued loss will likely have strong effects on the imperiled longleaf pine ecosystem in which it is closely associated. Using range-wide data from presence-absence surveys and publicly available presence-only data, we used recent advances in species distribution modeling to understand this ecosystem engineer’s associations with extensive land use and cover changes. Our study showed that pocket gophers inhabit a variety of land-use types, including regions with low levels of anthropogenic development, but are largely absent from intensely urbanized areas and closed-canopy forests. Conservation planning to reduce the spread of extensive development and forest management strategies to reduce closed-canopy systems will be instrumental in reducing the decline of the species and ultimately protection of the longleaf pine ecosystem. The approach used in this study will be increasingly beneficial for understanding species response to land-use change.

Using imperfect data in predictive mapping of vectors: a regional example of Ixodes ricinus distribution

BackgroundKnowledge of Ixodes ricinus tick distribution is critical for surveillance and risk management of transmissible tick-borne diseases such as Lyme borreliosis. However, as the ecology of I. ricinus is complex, and robust long-term geographically extensive distribution tick data are limited, mapping often relies on datasets collected for other purposes. We compared the modelled distributions derived from three datasets with information on I. ricinus distribution (quantitative I. ricinus count data from scientific surveys; I. ricinus presence-only data from public submissions; and a combined I. ricinus dataset from multiple sources) to assess which could be reliably used to inform Public Health strategy. The outputs also illustrate the strengths and limitations of these three types of data, which are commonly used in mapping tick distributions.MethodsUsing the Integrated Nested Laplace algorithm we predicted I. ricinus abundance and presence–absence in Scotland and tested the robustness of the predictions, accounting for errors and uncertainty.ResultsAll models fitted the data well and the covariate predictors for I. ricinus distribution, i.e. deer presence, temperature, habitat, index of vegetation, were as expected. Differences in the spatial trend of I. ricinus distribution were evident between the three predictive maps. Uncertainties in the spatial models resulted from inherent characteristics of the datasets, particularly the number of data points, and coverage over the covariate range used in making the predictions.ConclusionsQuantitative I. ricinus data from scientific surveys are usually considered to be gold standard data and we recommend their use where high data coverage can be achieved. However in this study their value was limited by poor data coverage. Combined datasets with I. ricinus distribution data from multiple sources are valuable in addressing issues of low coverage and this dataset produced the most appropriate map for national scale decision-making in Scotland. When mapping vector distributions for public-health decision making, model uncertainties and limitations of extrapolation need to be considered; these are often not included in published vector distribution maps. Further development of tools to better assess uncertainties in the models and predictions are necessary to allow more informed interpretation of distribution maps.

Sampling bias correction in species distribution models by quasi-linear Poisson point process

Species distribution modeling has an essential role in ecology to investigate habitat suitability based on the relationship between species occurrences and environmental conditions. The presence-only data for organisms is usually assumed to be obtained randomly from the region of interest; however, it is often the case that it is biased toward the areas easily accessed and adversely affects prediction accuracy. To address this sampling bias problem relevant to the prediction accuracy and model fitting of habitat distributions, we propose a new Poisson point process (PPP) model named as a quasi PPP, where environmental effect and sampling bias are explicitly modeled in separate clusters in a framework of quasi-linear modeling. The quasi-linear modeling is designed for capturing homogeneity within clusters and heterogeneity between clusters to improve the estimation accuracy of species distribution. The proposed model includes conventional models such as thinned and superposed PPPs as special cases. We have found that the quasi PPP outperforms the other existing methods in terms of goodness of model fitting. A statistical index based on the quasi-linear modeling is proposed to measure how the presence-only data used for the estimation of the species habitat distribution is affected by the sampling bias. The utility of the quasi PPP has been illustrated using simulation studies as well as the comprehensive vascular plant data in Japan. Our proposed model flexibly incorporates the effect of sampling biases to improve the estimation accuracy of species distributions. The results of data analysis are easily reproducible and applications to other data sets are also easily implementable by a package of qPPP of a statistical software R.

Modeling current and potential distributions of mammal species using presence-only data: A case study on British deer.

AimDecisions on wildlife conservation, management, and epidemiological risk are best based on robust evidence. The continual improvement of species distributions, such that they can be relied upon in decision‐making, is important. Here we seek to refine aspects of a generic modelling approach and improve the utility of species distribution maps.LocationGreat Britain (GB).MethodsWe applied a modeling framework based on hierarchical Bayesian species distribution models exploiting opportunistic occurrence records from citizen science datasets to predict both current and potential distributions for each of the six deer species known to be present in GB. Using the resulting maps, we performed a simple analysis of the overlap between species to illustrate possible contact, which we interpret as the relative risk of potential disease spread given an introduction.ResultsPredicted distribution maps showed good agreement with the broader scale occurrence reported by a recent national deer survey with an average True Skill Statistics and AUC of 0.69 and 0.89, respectively. Aggregation of the maps for all species highlighted regions of central and eastern England as well as parts of Scotland where extensive areas of range overlap could result in interspecific contact with consequences for risk assessments for diseases of deer. However, if populations are allowed to expand to their predicted potential, then areas of overlap, and therefore disease interspecific transmission risk, will become extensive and widespread across all of mainland Britain.Main conclusionsThe generic modeling approach outlined performed well across all of the deer species tested, offering a robust and reliable tool through which current and potential animal distributions can be estimated and presented. Our application, intended to inform quantitative risk assessments, demonstrates the practical use of such outputs to generate the valuable evidence required to inform policy decisions on issues such as management strategy.

A Sensitivity Analysis of the Application of Integrated Species Distribution Models to Mobile Species: A Case Study with the Endangered Baird’s Tapir

SummarySpecies distribution models (SDMs) are statistical tools used to develop continuous predictions of species occurrence. ‘Integrated SDMs’ (ISDMs) are an elaboration of this approach with potential advantages that allow for the dual use of opportunistically collected presence-only data and site-occupancy data from planned surveys. These models also account for survey bias and imperfect detection through the use of a hierarchical modelling framework that separately estimates the species–environment response and detection process. This is particularly helpful for conservation applications and predictions for rare species, where data are often limited and prediction errors may have significant management consequences. Despite this potential importance, ISDMs remain largely untested under a variety of scenarios. We performed an exploration of key modelling decisions and assumptions on an ISDM using the endangered Baird’s tapir (Tapirus bairdii) as a test species. We found that site area had the strongest effect on the magnitude of population estimates and underlying intensity surface and was driven by estimates of model intercepts. Selecting a site area that accounted for the individual movements of the species within an average home range led to population estimates that coincided with expert estimates. ISDMs that do not account for the individual movements of species will likely lead to less accurate estimates of species intensity (number of individuals per unit area) and thus overall population estimates. This bias could be severe and highly detrimental to conservation actions if uninformed ISDMs are used to estimate global populations of threatened and data-deficient species, particularly those that lack natural history and movement information. However, the ISDM was consistently the most accurate model compared to other approaches, which demonstrates the importance of this new modelling framework and the ability to combine opportunistic data with systematic survey data. Thus, we recommend researchers use ISDMs with conservative movement information when estimating population sizes of rare and data-deficient species. ISDMs could be improved by using a similar parameterization to spatial capture–recapture models that explicitly incorporate animal movement as a model parameter, which would further remove the need for spatial subsampling prior to implementation.

Estimating the annual distribution of monarch butterflies in Canada over 16 years using citizen science data

Monarch butterflies ( Danaus plexippus, Linnaeus, 1758) are comprised of two migratory populations separated by the Rocky Mountains and are renowned for their long-distance movements among the United States, Canada, and Mexico. Both populations have declined over several decades across North America prompting all three countries to evaluate conservation efforts. Monitoring monarch distribution and abundance is a necessary aspect of ongoing management in Canada where they are a species at risk. We used presence-only data from two citizen science data sets to estimate the annual breeding distribution of monarch butterflies in Canada between 2000 and 2015. Monarch breeding distribution in Canada varied widely among years owing to natural variation, and when considering the upper 95% of the probability of occurrence, the annual mean breeding distribution in Canada was 484 943 km2 (min: 173 449 km2; max: 1 425 835 km2). The area of occurrence was approximately an order of magnitude larger in eastern Canada than in western Canada. Habitat restoration for monarch butterflies in Canada should prioritize productive habitats in southern Ontario where monarchs occur annually and, therefore, likely contribute most to the long-term viability of monarchs in eastern North America. Overall, our assessment sets the geographic context to develop successful management strategies for monarchs in Canada.

Modelling common dolphin (Delphinus delphis) coastal distribution and habitat use: Insights for conservation

The world's ecosystems are altered to different extents by anthropogenic activities. Marine habitats, especially coastal areas, are subjected to an increasing pressure derived from human activities on both land and ocean. Information about species distribution is fundamental to develop effective conservation and management measures and counteract negative anthropogenic impacts. The present work explores the use of species distribution models by using the Environmental Niche Factor Analysis (ENFA) to assess the habitat suitability of common dolphins (Delphinus delphis) in Northwest Spain, and its application to the development of effective conservation and management measures. The relationship between presence-only data and ecogeographical variables (EGV) was used to assess the potential distribution of the species. Data was collected during 273 days at sea, covering a total distance of 9 417 km between March 2014 and October 2017 with a total of 91 common dolphin encounters. This study shows that tide level and sea surface salinity are the main EGVs driving the distribution of the species in coastal areas especially in waters above the continental shelf. Additionally, this study reveals the most suitable habitats for common dolphin and outlines the need to develop conservation measures and management plans to promote the protection of this species. Findings of the study contribute to a more accurate and comprehensive understanding of the common dolphin distribution and emphasize the importance of species distribution models in the development of effective conservation and management strategies.

Classification and regression with random forests as a standard method for presence-only data SDMs: A future conservation example using China tree species

The random forests (RF) algorithm is a superb learner and classifier in machine learning applications. This ensemble model is also one of the most popular species distribution model algorithms (SDMs) available to date. RF by default can produce categorical and numerical species distribution maps based on its classification tree (CT) and regression tree (RT) algorithms, respectively. Statistically, CT can also produce numerical predictions (class probability). Many real-world applications (e.g. conservation planning) employ binary presence–absence outputs that use classification thresholds to make these conversions. However, there is little available information regarding the difference in model performance between CT and RT for inference settings. Here, under an ensemble modeling framework, 52 forest tree species with presence-only data for all of China were selected for comparison of the performance of CT and RT algorithms in projecting the distribution and potential range shifts of these species under current and future climates. Five climatic variables were used to develop CT and RT models. Eight threshold-setting approaches were employed to convert numerical predictions into binary predictions. With regard to probabilistic predictions, the relative performance of CT and RT depended on the choice of the evaluation criteria. For both RT and CT, threshold-setting methods significantly altered the determination of thresholds, model performance, and subsequently projections of species range shifts under climate change. The four threshold selection methods (MaxKappa, MaxOA, MaxTSS, and MinROCdist) based on the composite model accuracy measures most often achieved significantly higher model performance than CT default threshold method and other threshold methods. They consistently projected that species' geographical ranges changed in response to climate change with the same direction and magnitude. We argue for choosing RT rather than CT as the SDM if model discrimination capacity (the ability to differentiate between occurrences of presence and absence) is viewed as more important than model reliability (the agreement between predicted relative indexes of occurrence and observed proportions of occurrence), and vice versa. In line with gradient theory, we can recommend the use of numerical predictions for species distribution modeling since they help to convey more information than binary predictions. Binary conversion of model outputs should only be carried out when it is clearly justified by the application's objective. The four aforementioned threshold methods are promising objective methods for binary conversions of continuous predictions when presence-only data are available. This study proposes guidelines on how machine learning can be used for specific applied and theoretical applications in a SDM context.

Assessing potential environmental influences on killer whale (Orcinus orca) distribution patterns in the Bremer Canyon, south-west Australia

ABSTRACTSubmarine canyons play an important role in the regional distribution, abundance and dispersal of marine biota and are increasingly being recognised as geomorphic features of high conservation significance along Australia’s continental margin. Certain canyons have been described as foraging ‘hotspots’ attributable to the high abundance of apex cetacean species aggregating in these areas. Anecdotal evidence of large seasonal aggregations of killer whales in the Bremer Canyon, south-west Australia, has attracted significant research attention in the last decade. To identify important environmental drivers influencing aggregation patterns, a predictive spatial habitat model using the Maxent model was developed based on presence-only whale sighting data. In addition, remotely sensed sea surface temperature and chlorophyll-a concentrations were assessed to investigate the spatio-temporal variation in sea surface conditions. Habitat preference was predicted in areas between canyon heads, with the most influential predictor variables being depth and distance from the continental shelf break. Analysis of remote-sensing data highlighted low localised variability in surface waters and illustrated the seasonal trends of the Leeuwin Current. This study demonstrates the influence of bathymetry and submarine geomorphology on enhanced cetacean abundance and highlights the need for recognition of this potential foraging area in marine reserve planning.

Prediction of the Species Distribution of Cryptococcus neoformans Throughout Turkey

Cryptococcus neoformans is a human pathogenic yeast that causes life-threatening infections especially in immunosuppressed patients. The environmental isolation of C.neoformans from Turkey was reported as early as 2004, although this was mostly from Eucalyptus camaldulensis colonization. Successful isolations were also reported from pomegranate (Punica granatum), oriental plane (Platanus orientalis), pine tree (Pinaceae), chestnut (Castanea sativa) and salt cedar (Tamarix hispida). The investigation of the relationship between the bioclimatic factors affecting the environmental isolation sites and the colonization of pathogens is a frequently used method. With this method, detailed risk maps can be generated in which environmental colonization can be estimated. The aim of this study was to use the high-resolution bioclimatic and previously-isolated yeasts' coordinates to create a valid model for the occurrence of C.neoformans in Turkey and provide insight into ecological processes. A machine learning approach using presence-only data software, maximum entropy (MaxEnt), was used to for the prediction of C.neoformans distribution. Climatic data and environmental bioclimatic variables from WorldClim were downloaded as 30 seconds spatial resolutions. The correlation between different Turkey bioclimatic layers were analyzed with ENMTools and similar layers were discarded. Forty-one different coordinates representing C.neoformans isolation points were used to generate a predictive map. The area under the curve and the omission rate were used to validate the model. Meanwhile, Jackknife tests were applied to enumerate the contribution of different environmental variables, and then to predict the final model. Maps were created using QGIS mapping software. In this study, we have shown that the coastal region of Anatolia, which is geographically located in the Northeastern Mediterranean Basin, as well as the entire Aegean region, carry an extremely high risk for the colonization of C.neoformans. Other areas which have not previously been reported for the isolation of C.neoformans were predicted to be potential colonization hotspots, including the western part of Ataturk Dam, the Amik Plain and the Bakırçay and Gediz valleys. The maximum temperature of the warmest month, the mean temperature of the warmest quarter and the precipitation of the coldest quarter were the most important factors influencing the model's predictions. It was determined that the humidity in the environment affected the colonization especially in November. In conclusion, we produced a C.neoformans colonization risk map of Turkey for the first time. Obtaining more regional data will facilitate the identification of the regions having similar risk. This approach is useful for the clinical prediagnosis of cryptococcosis cases, which may be more common in places with environmental niches.

Assessing the Vulnerability of Aquatic Macroinvertebrates to Climate Warming in a Mountainous Watershed: Supplementing Presence-Only Data with Species Traits

Mountainous running water ecosystems are vulnerable to climate change with major changes coming from warming temperatures. Species distribution will be affected and some species are anticipated to be winners (increasing their range) or losers (at risk of extinction). Climate change vulnerability is seldom integrated when assessing threat status for lists of species at risk (Red Lists), even though this might appear an important addition in the current context. The main objective of our study was to assess the potential vulnerability of Ephemeroptera (E), Plecoptera (P) and Trichoptera (T) species to global warming in a Swiss mountainous region by supplementing Species Distribution Models (SDMs) with a trait-based approach, using available historical occurrence and environmental data and to compare our outcomes with the Swiss National Red List. First, we used nine different modelling techniques and topographic, land use, climatic and hydrological variables as predictors of EPT species distribution. The shape of the response curves of the species for the environmental variables in the nine modelling techniques, together with three biological and ecological traits were used to assess the potential vulnerability of each species to climate change. The joint use of SDMs and trait approach appeared complementary and even though discrepancies were highlighted between SDMs and trait analyses, groups of potential “winners” and “losers” were raised out. Plecoptera appeared as the most vulnerable group to global warming. Divergences between current threat status of species and our results pointed out the need to integrate climate change vulnerability in Red List assessments.

Bayesian mixture models and their Big Data implementations with application to invasive species presence-only data

Due to their conceptual simplicity and flexibility, non-parametric mixture models are widely used to identify latent clusters in data. However, when it comes to Big Data, such as Landsat imagery, such model fitting is computationally prohibitive. To overcome this issue, we fit Bayesian non-parametric models to pre-smoothed data, thereby reducing the computational time from days to minutes, while disregarding little of the useful information. Tree based clustering is used to partition the clusters into smaller and smaller clusters in order to identify clusters of high, medium and low interest. The tree-based clustering method is applied to Landsat images from the Brisbane region, which were the actual sources of motivation for development of the method. The images are taken as a part of the red imported fire-ant eradication program that was launched in September 2001 and which is funded by all Australian states and territories, along with the federal government. To satisfy budgetary constraints, modelling is performed to estimate the risk of fire-ant incursion in each cluster so that the eradication program focuses on high risk clusters. The likelihood of containment is successfully derived by combining the fieldwork survey data with the results obtained from the proposed method.

Ecological Niche Modeling of Three Species of Stenella Dolphins in the Caribbean Basin, With Application to the Seaflower Biosphere Reserve

Dolphins of the genus Stenella occur in pelagic waters of both tropical and warm-temperate oceans. Three species, the Atlantic spotted dolphin (Stenella frontalis), the pantropical spotted dolphin (S. attenuata), and the spinner dolphin (S. longirostris) are abundant worldwide, but in the Caribbean Basin they have been poorly studied and information on their distribution patterns is scarce. Specifically, in Colombia’s remote Seaflower Biosphere Reserve (SFBR) S. attenuata has been reported occasionally, but S. frontalis and S. longirostris have never been recorded before. To address this information gap, an ecological niche modeling approach was used to determine the potential distribution patterns of these three dolphin species in the region. Records of these species for the Caribbean Basin were compiled, including both published and unpublished data. Environmental information, including bathymetry, bathymetric slope, distance to shore, sea surface temperature, sea surface salinity, and chlorophyll-a concentration was gathered from public databases (MARSPEC and Bio-ORACLE) in raster format. The maximum entropy algorithm (Maxent) for modeling species’ geographic distributions with presence-only data was used. After filtering the data, 210 records of S. attenuata, 204 of S. frontalis, and 80 of S. longirostris were used to run models. The best configuration for each model was chosen based on the ΔAICc criterion. For all three species, the final ecological niche models returned AUC test values higher than 0.8, indicating satisfactory model performance. The resulting potential distribution maps suggested that areas closest to continental shorelines of the Caribbean Basin and surrounding islands had the highest environmental suitability for all species (>70%). All models reported high environmental suitability for S. attenuata and S. longirostris in the SFBR, mainly in the southernmost part surrounding San Andres and Providence Archipelago. Assessment of niche overlap from the predictions of species distributions using the similarity statistic and pairwise map overlap indicated that S. frontalis and S. longirostris had niches slightly more similar in comparison to S. attenuata. As this was a first effort to fill a gap in our understanding of the distribution of species in the genus Stenella in the Caribbean Basin, further studies are necessary using both niche modeling and biological/ecological approaches.

Modelling potential habitat for snow leopards (Panthera uncia) in Ladakh, India.

The snow leopard Panthera uncia is an elusive species inhabiting some of the most remote and inaccessible tracts of Central and South Asia. It is difficult to determine its distribution and density pattern, which are crucial for developing conservation strategies. Several techniques for species detection combining camera traps with remote sensing and geographic information systems have been developed to model the habitat of such cryptic and low-density species in challenging terrains. Utilising presence-only data from camera traps and direct observations, alongside six environmental variables (elevation, aspect, ruggedness, distance to water, land cover, and prey habitat suitability), we assessed snow leopard habitat suitability across Ladakh in northern India. This is the first study to model snow leopard distribution both in India and utilising direct observation data. Results suggested that elevation and ruggedness are the two most influential environmental variables for snow leopard habitat suitability, with highly suitable habitat having an elevation range of 2,800 m to 4,600 m and ruggedness of 450 m to 1,800 m. Our habitat suitability map estimated approximately 12% of Ladakh's geographical area (c. 90,000 km2) as highly suitable and 18% as medium suitability. We found that 62.5% of recorded livestock depredation along with over half of all livestock corrals (54%) and homestays (58%) occurred within highly suitable snow leopard habitat. Our habitat suitability model can be used to assist in allocation of conservation resources by targeting construction of livestock corrals to areas of high habitat suitability and promoting ecotourism programs in villages in highly suitable snow leopard habitat.

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

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

The use of classification and regression algorithms using the random forests method with presence-only data to model species’ distribution

Random forests (RF) is a powerful species distribution model (SDM) algorithm. This ensemble model by default can produce categorical and numerical species distribution maps based on its classification tree (CT) and regression tree (RT) algorithms, respectively. The CT algorithm can also produce numerical predictions (class probability). Here, we present a detailed procedure involving the use of the CT and RT algorithms using the RF method with presence-only data to model the distribution of species. CT and RT are used to generate numerical prediction maps, and then numerical predictions are converted to binary predictions through objective threshold-setting methods. We also applied simple methods to deal with collinearity of predictor variables and spatial autocorrelation of species occurrence data. A geographically stratified sampling method was employed for generating pseudo-absences. The detailed procedural framework is meant to be a generic method to be applied to virtually any SDM prediction question using presence-only data.•How to use RF as a standard method for generic species distributions with presence-only data•How to choose RF (CT or RT) methods for the distribution modeling of species•A general and detailed procedure for any SDM prediction question.

Landscape Modeling of the Potential Natural Vegetation of Santa Catalina Island, California

The vegetation of Santa Catalina Island has been significantly transformed through a history of introduction of exotic plant species and disturbance by large introduced herbivores. Many of these disturbances have been reduced in recent decades, using measures such as carefully controlling the number of bison and removing cattle, sheep, feral pigs, and goats. The success of subsequent vegetation restoration actions depends on the choice of the right plant community for a location, which may not be obvious for an island with extensive areas dominated by exotic species. Environmental niche modeling is an approach to re-create the spatial distribution of habitat types for such a purpose. Such models, however, often require both presence and absence data to be meaningful, while in this scenario absence is misleading because it may reflect a long history of disturbance. Maximum entropy modeling is a technique to model species distributions with presence-only data that has been shown to produce accurate results. We used this modeling tool to model the environmental niche for distinct vegetation types, conceptualized as potential natural vegetation, on Catalina Island as a means to predict locations where restoration actions would be most successful and to predict potential natural vegetation prior to anthropogenic disturbance. Using an existing vegetation map, we extracted random points from within the polygons defining each native vegetation type. We then modeled the habitat suitability for each habitat using high-resolution environmental data that included elevation, aspect, hillshade, northeastness, slope, solar radiation, and topographic wetness index. The resulting models were combined to produce a map of potential natural vegetation. A 1977 map of potential natural vegetation included 4 vegetation types (woodland, chaparral, scrub, and grassland) to which we compared our results. Our new model of potential natural vegetation has high spatial complexity and high resolution. It also shows naturalistic responses to topography that are consistent with the broad patterns mapped in 1977 while providing fine-scale resolution to inform restoration efforts.

The potential for citizen science to produce reliable and useful information in ecology

We examined features of citizen science that influence data quality, inferential power, and usefulness in ecology. As background context for our examination, we considered topics such as ecological sampling (probability based, purposive, opportunistic), linkage between sampling technique and statistical inference (design based, model based), and scientific paradigms (confirmatory, exploratory). We distinguished several types of citizen science investigations, from intensive research with rigorous protocols targeting clearly articulated questions to mass‐participation internet‐based projects with opportunistic data collection lacking sampling design, and examined overarching objectives, design, analysis, volunteer training, and performance. We identified key features that influence data quality: project objectives, design and analysis, and volunteer training and performance. Projects with good designs, trained volunteers, and professional oversight can meet statistical criteria to produce high‐quality data with strong inferential power and therefore are well suited for ecological research objectives. Projects with opportunistic data collection, little or no sampling design, and minimal volunteer training are better suited for general objectives related to public education or data exploration because reliable statistical estimation can be difficult or impossible. In some cases, statistically robust analytical methods, external data, or both may increase the inferential power of certain opportunistically collected data. Ecological management, especially by government agencies, frequently requires data suitable for reliable inference. With standardized protocols, state‐of‐the‐art analytical methods, and well‐supervised programs, citizen science can make valuable contributions to conservation by increasing the scope of species monitoring efforts. Data quality can be improved by adhering to basic principles of data collection and analysis, designing studies to provide the data quality required, and including suitable statistical expertise, thereby strengthening the science aspect of citizen science and enhancing acceptance by the scientific community and decision makers.