Species Distribution Modelling Methods Research Articles

Do aposematic species have larger range sizes? A case study with neotropical poison frogs

AbstractAimAposematic animals, i.e., those that are defended and warn potential predators through signals, are suggested to have resource‐gathering advantages against non‐aposematic ones. We here explore this in a biogeographic framework expecting that aposematic species are better dispersers, which translates into larger geographic range size.LocationSouth America.TaxonPoison frogs (Amphibia; Aromobatidae and Dendrobatidae).MethodsWe use 43 toxic and 26 non‐toxic poison frog species from the lowlands only as representatives of aposematic and non‐aposematic study organisms, respectively. Realised and potential geographic ranges are calculated using minimum convex polygon and species distribution modelling methods, respectively. Accounting for species body size and phylogeny, we test if both range and aposematism are correlated using linear mixed‐effects models.ResultsAposematic and non‐aposematic species neither differ in realised nor in potential geographic range size. There was no effect on body size.Main ConclusionsThe role of aposematism is not yet as clear as suggested and determinants of poison frog range sizes are multifaceted. A more integrative approach is needed using the information on behaviour, predation risk, and reproductive biology to assess the role of aposematism on observed species distributions. Such data are not yet available for most species, neither poison frogs nor other aposematic animals.

Thinning occurrence points does not improve species distribution model performance

AbstractSpatial biases are an intrinsic feature of occurrence data used in species distribution models (SDMs). Thinning species occurrences, where records close in the geographic or environmental space are removed from the modeling procedure, is an approach often used to address these biases. However, thinning occurrence data can also negatively affect SDM performance, given that the benefits of removing spatial biases might be outweighed by the detrimental effects of data loss caused by this approach. We used real and virtual species to evaluate how spatial and environmental thinning affected different performance metrics of four SDM methods. The occurrence data of virtual species were sampled randomly, evenly spaced, and clustered in the geographic space to simulate different types of spatial biases, and several spatial and environmental thinning distances were used to thin the occurrence data. Null datasets were also generated for each thinning distance where we randomly removed the same number of occurrences by a thinning distance and compared the results of the thinned and null datasets. We found that spatially or environmentally thinned occurrence data is no better than randomly removing them, given that thinned datasets performed similarly to null datasets. Specifically, spatial and environmental thinning led to a general decrease in model performances across all SDM methods. These results were observed for real and virtual species, were positively associated with thinning distance, and were consistent across the different types of spatial biases. Our results suggest that thinning occurrence data usually fails to improve SDM performance and that the use of thinning approaches when modeling species distributions should be considered carefully.

Environmental drivers and the distribution of cold-water corals in the global ocean

Species distribution models (SDMs) are useful tools for describing and predicting the distribution of marine species in data-limited environments. Outputs from SDMs have been used to identify areas for spatial management, analyzing trawl closures, quantitatively measuring the risk of bottom trawling, and evaluating protected areas for improving conservation and management. Cold-water corals are globally distributed habitat-forming organisms that are vulnerable to anthropogenic impacts and climate change, but data deficiency remains an ongoing issue for the effective spatial management of these important ecosystem engineers. In this study, we constructed 11 environmental seabed variables at 500 m resolution based on the latest multi-depth global datasets and high-resolution bathymetry. An ensemble species distribution modeling method was used to predict the global habitat suitability for 10 widespread cold-water coral species, namely, 6 Scleractinian framework-forming species and 4 large gorgonian species. Temperature, depth, salinity, terrain ruggedness index, carbonate saturation state, and chlorophyll were the most important factors in determining the global distributions of these species. The Scleractinian Madrepora oculata showed the widest niche breadth, while most other species demonstrated somewhat limited niche breadth. The shallowest study species, Oculina varicosa, had the most distinctive niche of the group. The model outputs from this study represent the highest-resolution global predictions for these species to date and are valuable in aiding the management, conservation, and continued research into cold-water coral species.

Mapping current and future risk of scorpion sting from a species with low medical concern, Mesobuthus phillipsii (Scorpiones: Buthidae) in Iran.

Scorpion stings are one of the most important health challenges and high priority research topic in public health. In this study, we aimed to model habitat suitability of the Mesobuthus phillipsii (Pocock 1889), a species with low medical concern, under current and future climatic conditions in Iran. We also identified vulnerable populations to scorpion stings in the country. Scorpion sting risk modeling was done using an ensemble approach by considering two species distribution modeling methods: MaxEnt and Random Forest methods. Distribution modeling was performed using the sdm R package. The results showed that due to climate change in 2070, the high-risk areas will increase from 20,839 to 79,212 km2. Habitats with a moderate risk of scorpion stings will also increase from 139,347 to 222,833 km2. Consequently, the number of villages in high-risk areas of scorpion stings will increase from 2,870 to 7,017, while this number will increase from 12,759 to 20,104 in the case of medium-risk villages. The results of this study can be used for scorpion stings management in Iran. This study can be used as an example for similar studies on scorpions with high medical emergency.

Trends in mosquito species distribution modeling: insights for vector surveillance and disease control

Species distribution modeling (SDM) has become an increasingly common approach to explore questions about ecology, geography, outbreak risk, and global change as they relate to infectious disease vectors. Here, we conducted a systematic review of the scientific literature, screening 563 abstracts and identifying 204 studies that used SDMs to produce distribution estimates for mosquito species. While the number of studies employing SDM methods has increased markedly over the past decade, the overwhelming majority used a single method (maximum entropy modeling; MaxEnt) and focused on human infectious disease vectors or their close relatives. The majority of regional models were developed for areas in Africa and Asia, while more localized modeling efforts were most common for North America and Europe. Findings from this study highlight gaps in taxonomic, geographic, and methodological foci of current SDM literature for mosquitoes that can guide future efforts to study the geography of mosquito-borne disease risk.Graphical

A Modeling Framework to Frame a Biological Invasion: Impatiens glandulifera in North America.

Biological invasions are a major component of global environmental change with severe ecological and economic consequences. Since eradicating biological invaders is costly and even futile in many cases, predicting the areas under risk to take preventive measures is crucial. Impatiens glandulifera is a very aggressive and prolific invasive species and has been expanding its invasive range all across the Northern hemisphere, primarily in Europe. Although it is currently spread in the east and west of North America (in Canada and USA), studies on its fate under climate change are quite limited compared to the vast literature in Europe. Hybrid models, which integrate multiple modeling approaches, are promising tools for making projections to identify the areas under invasion risk. We developed a hybrid and spatially explicit framework by utilizing MaxEnt, one of the most preferred species distribution modeling (SDM) methods, and we developed an agent-based model (ABM) with the statistical language R. We projected the I. glandulifera invasion in North America, for the 2020-2050 period, under the RCP 4.5 scenario. Our results showed a predominant northward progression of the invasive range alongside an aggressive expansion in both currently invaded areas and interior regions. Our projections will provide valuable insights for risk assessment before the potentially irreversible outcomes emerge, considering the severity of the current state of the invasion in Europe.

VoluModel: Modelling species distributions in three‐dimensional space

Abstract Ecological niche modelling (ENM), species distribution modelling and related spatial analytical methods were first developed in two‐dimensional (2‐D) terrestrial systems; many common ENM workflows organize and analyse geographically structured occurrence and environmental data based on 2‐D latitude and longitude coordinates. This may be suitable for most terrestrial organisms, but pelagic marine species are distributed not only horizontally but also vertically. Extracting environmental data for marine species based only on latitude and longitude coordinates may result in poorly trained ENMs and inaccurate prediction of species' geographical distributions, as water conditions may vary strikingly with depth. We developed the voluModel R package to efficiently extract three‐dimensional (3‐D) environmental data for training ENMs (i.e. presences and absences/pseudoabsences/background). voluModel also provides tools for 3‐D ENM projection visualization and estimation of model extrapolation risk. We present the main features of the voluModel R package and provide a simple modelling workflow for Luminous Hake, Steindachneria argentea, as an example. We also compare results from 2‐D and 3‐D spatial models to demonstrate differences in how the modelling methods perform. The use of 3‐D environmental data generates more precise estimates of environmental conditions for training ENMs. This method also improves inference of species' suitable abiotic ecological niches and potential geographic ranges. 3‐D niche modelling is important step forward for marine macroecology and biogeography, as it will yield more accurate estimates of ocean species richness and potential past and future changes in the horizontal and vertical dimensions of species' geographic ranges. The latter is particularly relevant considering ongoing climate change that may cause redistribution of species in environmental space (both in latitude and depth) over time.

Flexible species distribution modelling methods perform well on spatially separated testing data

AbstractAimTo assess whether flexible species distribution models that perform well at nearby testing locations still perform strongly when evaluated on spatially separated testing data.LocationAustralian Wet Tropics (AWT), Ontario, Canada (CAN), north‐east New South Wales, Australia (NSW), New Zealand (NZ), five countries of South America (SA), and Switzerland (SWI).Time periodMost species data were collected between 1950 and 2000.Major taxa studiedBirds, mammals, plants and reptiles.MethodsWe compared 10 species distribution modelling methods with varying flexibility in terms of the allowed complexity of their fitted functions [boosted regression trees (BRT), generalized additive model (GAM), multivariate adaptive regression splines (MARS), maximum entropy (MaxEnt), support vector machine (SVM), variants of generalized linear model (GLM) and random forest (RF), and an Ensemble model]. We used established practices for model selection to avoid overfitting, including parameter tuning in learning methods. Models were trained on presence–background data for 171 species and tested on presence–absence data. Training and testing data were separated using both random and spatial partitioning, the latter based on 75‐km blocks. We calculated the average performance and mean rank of the methods (focussing on the area under the receiver operating characteristic and precision‐recall gain curves, and correlation) and assessed the statistical significance of the differences between them.ResultsThe ranking of methods did not change when evaluated on spatially separated testing data. Methods with the strongest predictive performance were nonparametric methods known to be flexible. An ensemble formed by averaging predictions of five pre‐selected modelling methods was the best model in both random and spatial partitioning, followed by MaxEnt and a variant of random forest.Main conclusionsWhilst some modellers expect methods limited to simple smooth functions to predict better spatially separated data, we found no evidence of that using blocks of 75 km. We conclude that flexible models that are tuned well enough to avoid overfitting are effective at predicting to spatially distinct areas.

Открытые источники данных для моделирования ареалов: информационные системы о биоразнообразии и наборы пространственных данных условий среды

Открытые источники данных для моделирования ареалов: информационные системы о биоразнообразии и наборы пространственных данных условий среды

Habitat suitability model and range shift analysis for American Chestnut (Castanea dentata) in the United States

American chestnut (Castanea dentata) was a multifunctional hardwood tree species that provide both economic and ecological benefits in the United States. It was a major source of timber before the chestnut blight (Cryphonectria parasitica) invasion wiped out the stands of this species from the forest between the early 1900s and 1950s. Our study focuses on the estimation of suitable habitat in the eastern US, range shift predictions, and identification of habitat drivers for American chestnut (AMC) in the eastern US to help corroborate the restoration efforts of C. dentata using an ensemble model from five species distribution modeling (SDMs) methods while also projecting the prediction until 2100 in response to the changing climate. 1806 occurrence records were obtained from four sources including the United States Department of Agriculture (USDA) Forest Service. The result from the SDM techniques shows a model accuracy of the true skill statistics (TSS) and Area-Under-the-Curve (AUC) of greater than 0.8 and 0.90 respectively, which were ensembled. The historic suitable habitat of AMC compared to the present habitat suitability prediction in the US shows a significant reduction. The mean temperature of the warmest quarter, precipitation seasonality, and temperature seasonality are the top three variables dictating the distribution of the American chestnut in the US. The future projection also shows up to 100% less suitable habitats by the end of the century under high emissions and high precipitation with a northeast range shift up to 476.9 km. Modeling the American chestnut's habitat suitability is imperative for a successful restoration program and these results highlight the potential need for greater conservation and use of southern genotypes.

Finding what you don’t know: Testing SDM methods for poorly known species

AbstractAimA limitation of species distribution models (SDMs) is that species with low sample sizes are difficult to model. Yet, it is often important to know the habitat associations of poorly known species to guide conservation efforts. Techniques have been proposed for modelling species’ distributions from a few records, but their performance relative to one another has not been compared. Because these models are built and evaluated with small data sets, sampling error could cause severely biased sampling in environmental space. As a result, SDMs are likely to underpredict geographic distributions given small sample sizes. We perform the first comparison of methods explicitly promoted or developed for predicting the geographic ranges of species with very low sample sizes.LocationNorth Carolina, USA.TaxonSouth Mountains Grey‐cheeked Salamander (Plethodon meridianus).MethodsUsing the sparse, existing georeferenced records of P. meridianus, we built SDMs using a range of methods that previous researchers have argued should work for low sample sizes. We then tested each SDM’s ability to accurately predict independent survey data that were not georeferenced prior to our study. We compared SDMs using omission error and AUC.ResultsRoughly half of the models successfully predicted survey records in the range centre, and all models had high omission error rates in the range exterior. In the range of interior or exterior, the ‘ensemble of small models’ technique produced SDMs with high omission error rates. Spatial filtering had a negligible impact on model performance. Most, but not all, models outperformed predictions using distance from known populations. Using one of the best‐performing methods, we developed an improved range map of P. meridianus.Main ConclusionsGeographically peripheral populations were difficult to predict for all SDMs, though some methods were clearly inferior for our data set. We recommend that when sample sizes are low, researchers use Maxent with species‐specific model settings.

Influence of species interaction on species distribution simulation and modeling methods.

The species distribution models (SDMs) simulate and predict the potential distribution of species in geographical space by quantifying the relationships between species distribution and environmental variables, and extrapolating these relationships to unknown landscape units, which makes them important tools in ecology, biogeo-graphy, and conservation biology. Current SDMs mainly take abiotic factors as prediction variables, whereas biotic factors, especially species interactions, are often ignored due to the difficulties in data quantification and modeling. Incorporating species interactions into SDMs is considered as the main challenge of SDMs. We reviewed the influence of species interactions on species distribution simulations, clarified the necessity of incorporating species interactions into SDMs, summarized four main ways to incorporate species interactions into SDMs, analyzed their strengths and limitations, and discussed the future development direction of incorporating species interactions into SDMs. The study showed that incorporating species interaction into SDMs was based on the premise that the spatial scale of species distribution simulation was consistent with that of species interactions, and that the training data should be collected from large environmental heterogeneous space to ensure the diversity of species interactions in heterogeneous habitats. In order to eliminate the influence of multicollinearity on the prediction of SDMs, all abiotic and biotic factors should be fully considered and accurately quantified. Modeling the complex population/community dynamics would be an important development direction of incorporating species interactions into SDMs.

Predicting Current Potential Distribution and the Range Dynamics of Pomacea canaliculata in China under Global Climate Change.

Simple SummaryPomacea canaliculata is one of the 100 worst invasive alien species in the world, causing significant effects and harm to native species, ecological environment, human health, and social economy. In this study, we used species distribution modeling (SDM) methods to predict the potential distribution of P. canaliculata in China and found that with climate change, there would be a trend of expanding and moving northward in the future.Pomacea canaliculata is one of the 100 worst invasive alien species in the world, which has significant effects and harm to native species, ecological environment, human health, and social economy. Climate change is one of the major causes of species range shifts. With recent climate change, the distribution of P. canaliculata has shifted northward. Understanding the potential distribution under current and future climate conditions will aid in the management of the risk of its invasion and spread. Here, we used species distribution modeling (SDM) methods to predict the potential distribution of P. canaliculata in China, and the jackknife test was used to assess the importance of environmental variables for modeling. Our study found that precipitation of the warmest quarter and maximum temperature in the coldest months played important roles in the distribution of P. canaliculata. With global warming, there will be a trend of expansion and northward movement in the future. This study could provide recommendations for the management and prevention of snail invasion and expansion.

Population decline at distribution margins: Assessing extinction risk in the last glacial relictual but still functional metapopulation of a European butterfly

AbstractAimTo determine the interplay between climate and land use changes in driving population dynamics in a butterfly species, Coenonympha hero, at the southern limit of its distribution.LocationFrench Jura massif and Europe.MethodsWe analysed patterns of genetic diversity distribution at 817 loci in 136 butterflies from 31 sites using NGS to infer the genetic structure and population size changes over time, using two methods of demographic inference (SNP frequency spectrum analyses and coalescent ABC inferences). We then characterized the climate and land use descriptors of C. hero geographic distribution using species distribution modelling (SDM) and ordination method and compared demographic changes to changes in climatically suitable areas.ResultsCoenonympha hero persists in the Jura massif as three core populations that share a common history of decline in two steps: an old decline and a more recent decline that resulted in population fragmentation, the southernmost sites being the most threatened (lowest genetic diversity). Climate change during the Würm glaciation is presumably the main factor explaining the first demographic decline. The second decline started 2000 years ago possibly under increased human pressure as suggested by recent extinctions in several sites nowadays characterized by urban and agricultural surfaces. Both climate and land use variables are important descriptors of C. hero distribution, as SDM predictions were improved by adding aridity index, altitude and land use to bioclimatic predictors. Its habitats include forests in north‐eastern Europe and grasslands in the Jura massif.Main conclusionsUsing SDM and genetic demographic inferences, we identified a persistent glacial refuge for the species in Europe. We show that although this relictual population has declined and fragmented under the combined effects of climate warming and anthropization, the metapopulation is still functional but requires particular conservation attention to maintain its connectivity, and to favour the local persistence of this highly endangered butterfly species.

Solving sampling bias problems in presence–absence or presence‐only species data using zero‐inflated models

AbstractAimLarge 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.LocationAll locationsTaxonAll taxaMethodsWe 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.ResultsZI 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 ConclusionsOur 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.

Terrestrial or marine species distribution model: Why not both? A case study with seabirds.

Species reliant on both the terrestrial and marine realms present a challenge for conventional species distribution models (SDMs). For such species, standard single‐realm SDMs may omit key information that could result in decreased model accuracy and performance. Existing approaches to habitat suitability modeling typically do not effectively combine information from multiple realms; this methodological gap can ultimately hamper management efforts for groups such as seabirds, seals, and turtles. This study, for the first time, jointly incorporates both terrestrial information and marine information into a single species distribution model framework. We do this by sampling nearby marine conditions for a given terrestrial point and vice versa using parameters set by each species’ mean maximum foraging distance and then use standard SDM methods to generate habitat suitability predictions; therefore, our method does not rely on post hoc combination of several different models. Using three seabird species with very different ecologies, we investigate whether this new multi‐realm approach can improve our ability to identify suitable habitats for these species. Results show that incorporating terrestrial information into marine SDMs, or vice versa, generally improves model performance, sometimes drastically. However, there is considerable variability between species in the level of improvement as well as in the particular method that produces the most improvement. Our approach provides a repeatable and transparent method to combine information from multiple ecological realms in a single SDM framework. Important advantages over existing solutions include the opportunity to, firstly, easily combine terrestrial and marine information for species that forage large distances inland or out to sea and, secondly, consider interactions between terrestrial and marine variables.

USING SDM FOR MODELING THE POTENTIAL RANGE OF <i>COCHLODINA LAMINATA</i> (GASTROPODA, PULMONATA, CLAUSILIIDAE) IN EUROPE

The study presents a model of the potential distribution of the land snail Cochlodina laminata (Clausiliidae), created using the SDM (species distribution modeling) methods. The forecast was made on the basis C. laminata records from GBIF repository. This species has the wide distribution in of this and wide ecological valence in comparison with the most of the european clausilids. Climatic parameters and altitude from the open database WorldClim.org were selected as predictors. For the forecast, the models with the best statistical indicators were selected, and the ensemble forecast procedure was carried out to optimize the results. The result of the ensemble forecast showed that the most suitable for habitat of C. laminata was the territory of Western and Southern Europe. Obviously, the influence of the GBIF records on the result obtained. In addition to the main territory, the potential range also included a number of the British Isles and part of the Jutland Peninsula. To the east of the main part of the potential range, the southern and eastern Black Sea regions were found in the zone with a high suitability for the species. In addition to the main territory of the potential range, which is located in Europe, local areas have been identified in the territory of North and South America, as well as Australia and New Zealand. The points with a high probability of finding the species on other continents are confined to the sea coasts and are extremely few in number. The forecast results are determined not only by the optimal combination of the values of climatic predictors, but also by spatial reference of the initial data. Thus, due to the absence of finds from Eastern Europe in the initial data, the territory of Eastern Europe did not fall into a potentially suitable zone. Thus, the forecast results are determined not only by the optimal combination of the climatic predictors, but also by the initial data. In particular, due to the absence of records on the territory of Eastern Europe in the initial data, this territory was not included in a potentially suitable zone.

Impact of climate change on the spatial distribution of endemic legume species of the Guineo-Congolian forest, Africa

Legumes (family Leguminosae) are an economically important and widely distributed species found in several regions of the world, including the Guineo-Congolian (GC) forest region. Recent studies have reported the impact of climate change on plant biodiversity. Thus, understanding the distribution pattern of the endemic legume species of the GC region under current and future climate scenarios are imperative for long-term conservation management plans. We used 762 occurrence data of the 237 endemic legume species to elucidate the habitat suitability and impact of climate change on the species. Analyses used an ecological niche modeling ensemble method to map the current and future (2050–2070) distribution patterns of the endemic legumes under the RCP 4.5 scenario. Further, we generated the ensemble of four different species distribution modeling methods to predict areas where climatically suitable regions is lost or gained under future climate change. The model predicted the current suitable areas for endemic legumes to be Southern Cameroon, Equatorial Guinea, Gabon, Republic of Congo (COG), and Democratic Republic of Congo (DRC). Our findings indicate precipitation of Warmest Quarter and seasonality as the significant climatic variables affecting the habitat suitability of endemic legumes in the GC regions. Results showed areas such as the coastal region of Cote d' Ivoire, southwest province of Cameroon, Central African Republic, COG (Ngounnie), central DRC, and southern Nigeria will experience >40% loss of climatically suitable areas by 2050. Also, species such as Baphia maxima, and Psophocarpus scandens were projected to experience a 70% reduction in their climatically suitable areas under future climate scenarios. Our study concludes that the projected distribution of the endemic legume species in the GC region will consequently decline in the future due to the gross effect of climate change. The anticipated changes in the distribution patterns of the endemic legumes could have significant ecological and socio‐economic significance. Therefore, we recommend improved conservation to protect the endemic legume species and its associated habitats in GC forest.

The potential for using rare, native species in reforestation– A case study of yews (Taxaceae) in China

Ecosystem restoration is regarded as one of the most cost-effective ways of mitigating the effects of the ongoing climate- and anthropogenic changes and slow or revert the loss of biodiversity. Forest restoration has much potential to conserve forest specialist species and ecosystem services, by using multiple, native tree species to promote a high local diversity of trees, with likely positive effects on overall biodiversity and on ecosystem resilience. In this study, we assessed the potential of using two rare, native species of the yew family (Taxaceae) in forest restoration in China. Species of this family are only rarely used in reforestation despite their potential contribution to tree functional diversity as long-lived, shade-tolerant, evergreen understory trees with fleshy seed cones of value to frugivorous animals. By using species distribution modelling methods, we analysed national and local scale occurrence data for Taxus cuspidata and Torreya grandis to determine the climate-based potential ranges as well as important factors for growth on a local scale. The analyses showed that both species have large potential ranges driven mainly by precipitation and by comparing these ranges with the areas that have potential for sustaining forests, we found large areas available for forest restoration where these species could be included. On the local scale, we found that low light levels and low competition from co-occurring trees are more important for the growth of seedlings compared to the adult individuals of both species. If the ecological requirements for seedlings are ensured, i.e. by creating moderately shaded environments in which seedlings can escape competition, both Taxaceae species have high potential for reforestation in China and will increase the ecological qualities of a restored forest, and at the same time, the conservation of rare tree-species in their native ranges is ensured. Conclusively, both Ta. cuspidata and To. grandis are shade-tolerant, slow-growing trees that, by creating an evergreen scrubby layer, add to forest structural complexity and stability, thereby helping support ecosystem services and biodiversity, e.g. microhabitat and resources for birds and other animals.

Data quantity is more important than its spatial bias for predictive species distribution modelling.

Biological records are often the data of choice for training predictive species distribution models (SDMs), but spatial sampling bias is pervasive in biological records data at multiple spatial scales and is thought to impair the performance of SDMs. We simulated presences and absences of virtual species as well as the process of recording these species to evaluate the effect on species distribution model prediction performance of (1) spatial bias in training data, (2) sample size (the average number of observations per species), and (3) the choice of species distribution modelling method. Our approach is novel in quantifying and applying real-world spatial sampling biases to simulated data. Spatial bias in training data decreased species distribution model prediction performance, but sample size and the choice of modelling method were more important than spatial bias in determining the prediction performance of species distribution models.