BIOMOD – optimizing predictions of species distributions and projecting potential future shifts under global change

A new computation framework (BIOMOD: BIOdiversity MODelling) is presented, which aims to maximize the predictive accuracy of current species distributions and the reliability of future potential distributions using different types of statistical modelling methods. BIOMOD capitalizes on the different techniques used in static modelling to provide spatial predictions. It computes, for each species and in the same package, the four most widely used modelling techniques in species predictions, namely Generalized Linear Models (GLM), Generalized Additive Models (GAM), Classification and Regression Tree analysis (CART) and Artificial Neural Networks (ANN). BIOMOD was applied to 61 species of trees in Europe using climatic quantities as explanatory variables of current distributions. On average, all the different modelling methods yielded very good agreement between observed and predicted distributions. However, the relative performance of different techniques was idiosyncratic across species, suggesting that the most accurate model varies between species. The results of this evaluation also highlight that slight differences between current predictions from different modelling techniques are exacerbated in future projections. Therefore, it is difficult to assess the reliability of alternative projections without validation techniques or expert opinion. It is concluded that rather than using a single modelling technique to predict the distribution of several species, it would be more reliable to use a framework assessing different models for each species and selecting the most accurate one using both evaluation methods and expert knowledge.     

Different climatic envelopes among invasive populations may lead to underestimations of current and future biological invasions

Aim We explore the impact of calibrating ecological niche models (ENMs) using (1) native range (NR) data versus (2) entire range (ER) data (native and invasive) on projections of current and future distributions of three Hieracium species. Location H. aurantiacum, H. murorum and H. pilosella are native to Europe and invasive in Australia, New Zealand and North America. Methods Differences among the native and invasive realized climatic niches of each species were quantified. Eight ENMs in BIOMOD were calibrated with (1) NR and (2) ER data. Current European, North American and Australian distributions were projected. Future Australian distributions were modelled using four climate change scenarios for 2030. Results The invasive climatic niche of H. murorum is primarily a subset of that expressed in its native range. Invasive populations of H. aurantiacum and H. pilosella occupy different climatic niches to those realized in their native ranges. Furthermore, geographically separate invasive populations of these two species have distinct climatic niches. ENMs calibrated on the realized niche of native regions projected smaller distributions than models incorporating data from species’ entire ranges, and failed to correctly predict many known invasive populations. Under future climate scenarios, projected distributions decreased by similar percentages, regardless of the data used to calibrate ENMs; however, the overall sizes of projected distributions varied substantially. Main conclusions This study provides quantitative evidence that invasive populations of Hieracium species can occur in areas with different climatic conditions than experienced in their native ranges. For these, and similar species, calibration of ENMs based on NR data only will misrepresent their potential invasive distribution. These errors will propagate when estimating climate change impacts. Thus, incorporating data from species’ entire distributions may result in a more thorough assessment of current and future ranges, and provides a closer approximation of the elusive fundamental niche.     

Shifting ranges and conservation challenges for lemurs in the face of climate change

Geospatial modeling is one of the most powerful tools available to conservation biologists for estimating current species ranges of Earth's biodiversity. Now, with the advantage of predictive climate models, these methods can be deployed for understanding future impacts on threatened biota. Here, we employ predictive modeling under a conservative estimate of future climate change to examine impacts on the future abundance and geographic distributions of Malagasy lemurs. Using distribution data from the primary literature, we employed ensemble species distribution models and geospatial analyses to predict future changes in species distributions. Current species distribution models (SDMs) were created within the BIOMOD2 framework that capitalizes on ten widely used modeling techniques. Future and current SDMs were then subtracted from each other, and areas of contraction, expansion, and stability were calculated. Model overprediction is a common issue associated Malagasy taxa. Accordingly, we introduce novel methods for incorporating biological data on dispersal potential to better inform the selection of pseudo‐absence points. This study predicts that 60% of the 57 species examined will experience a considerable range of reductions in the next seventy years entirely due to future climate change. Of these species, range sizes are predicted to decrease by an average of 59.6%. Nine lemur species (16%) are predicted to expand their ranges, and 13 species (22.8%) distribution sizes were predicted to be stable through time. Species ranges will experience severe shifts, typically contractions, and for the majority of lemur species, geographic distributions will be considerably altered. We identify three areas in dire need of protection, concluding that strategically managed forest corridors must be a key component of lemur and other biodiversity conservation strategies. This recommendation is all the more urgent given that the results presented here do not take into account patterns of ongoing habitat destruction relating to human activities.     

Improving species distribution models using biotic interactions: a case study of parasites,pollinators and plants

Biotic interactions have been considered as an important factor to be included in species distribution modelling, but little is known about how different types of interaction or different strategies for modelling affect model performance. This study compares different methods for including interspecific interactions in distribution models for bees, their brood parasites, and the plants they pollinate. Host–parasite interactions among bumble bees (genus Bombus: generalist pollinators and brood parasites) and specialist plant–pollinator interactions between Centris bees and Krameria flowers were used as case studies. We used 7 different modelling algorithms available in the BIOMOD R package. For Bombus, the inclusion of interacting species distributions generally increased model predictive accuracy. The improvement was better when the interacting species was included with its raw distribution rather than with its modeled suitability. However, incorporating the distributions of non‐interacting species sometimes resulted in similarly increased model accuracy despite their being no significance of any interaction for the distribution. For the Centris‐Krameria system the best strategy for modelling biotic interactions was to include the interacting species model‐predicted values. However, the results were less consistent than those for Bombus species, and most models including biotic interactions showed no significant improvement over abiotic models. Our results are consistent with previous studies showing that biotic interactions can be important in structuring species distributions at regional scales. However, correlations between species distributions are not necessarily indicative of interactions. Therefore, choosing the correct biotic information, based on biological and ecological knowledge, is critical to improve the accuracy of species distribution models and forecast distribution change.     

Species Distribution Models of Tropical Deep-Sea Snappers

Deep-sea fisheries provide an important source of protein to Pacific Island countries and territories that are highly dependent on fish for food security. However, spatial management of these deep-sea habitats is hindered by insufficient data. We developed species distribution models using spatially limited presence data for the main harvested species in the Western Central Pacific Ocean. We used bathymetric and water temperature data to develop presence-only species distribution models for the commercially exploited deep-sea snappers Etelis Cuvier 1828, Pristipomoides Valenciennes 1830, and Aphareus Cuvier 1830. We evaluated the performance of four different algorithms (CTA, GLM, MARS, and MAXENT) within the BIOMOD framework to obtain an ensemble of predicted distributions. We projected these predictions across the Western Central Pacific Ocean to produce maps of potential deep-sea snapper distributions in 32 countries and territories. Depth was consistently the best predictor of presence for all species groups across all models. Bathymetric slope was consistently the poorest predictor. Temperature at depth was a good predictor of presence for GLM only. Model precision was highest for MAXENT and CTA. There were strong regional patterns in predicted distribution of suitable habitat, with the largest areas of suitable habitat (> 35% of the Exclusive Economic Zone) predicted in seven South Pacific countries and territories (Fiji, Matthew & Hunter, Nauru, New Caledonia, Tonga, Vanuatu and Wallis & Futuna). Predicted habitat also varied among species, with the proportion of predicted habitat highest for Aphareus and lowest for Etelis. Despite data paucity, the relationship between deep-sea snapper presence and their environments was sufficiently strong to predict their distribution across a large area of the Pacific Ocean. Our results therefore provide a strong baseline for designing monitoring programs that balance resource exploitation and conservation planning, and for predicting future distributions of deep-sea snappers.     

How interactions between animal movement and landscape processes modify local range dynamics and extinction risk

Forecasts of range dynamics now incorporate many of the mechanisms and interactions that drive species distributions. However, connectivity continues to be simulated using overly simple distance-based dispersal models with little consideration of how the individual behaviour of dispersing organisms interacts with landscape structure (functional connectivity). Here, we link an individual-based model to a niche-population model to test the implications of this omission. We apply this novel approach to a turtle species inhabiting wetlands which are patchily distributed across a tropical savannah, and whose persistence is threatened by two important synergistic drivers of global change: predation by invasive species and overexploitation. We show that projections of local range dynamics in this study system change substantially when functional connectivity is modelled explicitly. Accounting for functional connectivity in model simulations causes the estimate of extinction risk to increase, and predictions of range contraction to slow. We conclude that models of range dynamics that simulate functional connectivity can reduce an important source of bias in predictions of shifts in species distributions and abundances, especially for organisms whose dispersal behaviours are strongly affected by landscape structure.     

The range implications of lizard traits in changing environments

Aim Most predictions of species ranges are based on correlating current species localities to environmental conditions. These correlative models do not explicitly include a species' biology. In contrast, some mechanistic models link traits to energetics and population dynamics to predict species distributions. These models enable one to ask whether considering a species' biology is important for predicting its range. I implement mechanistic models to investigate how a species' morphology, physiology and life history influence its range. Location North America. Methods I compare the mechanistic model predictions with those of correlative models for eight species of North American lizards in both current environments and following a uniform 3 °C temperature warming. I then examine the implications of superimposing habitat and elevation requirements on constraints associated with environmental tolerances. Results In the mechanistic model, species with a narrower thermal range for activity are both predicted and observed to have more restricted distributions. Incorporating constraints on habitat and elevation further restricts species distributions beyond areas that are thermally suitable. While correlative models generally outperform mechanistic models at predicting current distributions, the performance of mechanistic models improves when incorporating additional factors. In response to a 3 °C temperature warming, the northward range shifts predicted by the mechanistic model vary between species according to trait differences and are of a greater extent than those predicted by correlative models. Main conclusions These findings highlight the importance of species traits for understanding the dynamics of species ranges in changing environments. The analysis demonstrates that mechanistic models may provide an important complement to correlative models for predicting range dynamics, which may underpredict climate‐induced range shifts.     

Biotic interactions affect the elevational ranges of high‐latitude plant species

Ecological theory suggests that positive plant–plant interactions can extend species distributions into areas that would otherwise be unfavourable. However, few studies have tested this hypothesis, and none have explicitly examined the associated prediction that inter‐specific interactions between plants may broaden species altitudinal distributions. Here we test this prediction, using fine‐scale species distribution data for 156 bryophytes, lichens and vascular plants spanning a 900 m elevational gradient in north‐western Finland and Norway, analysed with a niche modelling approach. Species altitudinal ranges of all three groups of plants were more accurately predicted when including the cover of any of the 24 most wide‐spread and abundant species (‘dominants’) than when using abiotic variables alone, emphasizing the importance of including relevant biotic predictors in species distribution models. Half of the models showed that species had very low probabilities of occurrence under high cover of dominants, suggesting a strong negative impact of dominant species. Similarly, for species that are predicted to occur irrespective of dominant species cover, 62% of models showed narrower species altitudinal distributions when occurring under high dominant cover, with contractions of species’ lower and upper elevational limits being common. Nonetheless, high cover of dominant species was associated with upslope range extension in 43 species, and a net range expansion in nearly 10% of all models. Species distributional responses to dominants were only weakly related to species traits, with larger range contractions associated with arctic‐alpine dominants. Therefore, dominant species appear to exert a strong influence on the elevational distribution of other species in high latitude environments.     

Do climate envelope models transfer? A manipulative test using dung beetle introductions

Climate envelope models (CEMs) are widely used to forecast future shifts in species ranges under climate change, but these models are rarely validated against independent data, and their fundamental assumption that climate limits species distributions is rarely tested. Here, we use the data on the introduction of five South African dung beetle species to Australia to test whether CEMs developed in the native range can predict distribution in the introduced range, where the confounding effects of dispersal limitation, resource limitation and the impact of natural enemies have been removed, leaving climate as the dominant constraint. For two of the five species, models developed in the native range predict distribution in the introduced range about as well as models developed in the introduced range where we know climate limits distribution. For the remaining three species, models developed in the native range perform poorly, implying that non-climatic factors limit the native distribution of these species and need to be accounted for in species distribution models. Quantifying relevant non-climatic factors and their likely interactions with climatic variables for forecasting range shifts under climate change remains a challenging task.     

Species distribution models for Peruvian plantcutter improve with consideration of biotic interactions

Biotic interactions have been controversial in distributional ecology, mainly in regards to whether they have effects over broad extents, with the negative view known as the Eltonian noise hypothesis (ENH). In this study, we evaluated the ENH for Phytotoma raimondii, a restricted‐range Peruvian endemic bird species: we developed models based on 1) only abiotic conditions, 2) only host plant distributions, and 3) both abiotic conditions and host plant distributions; models were evaluated with partial receiver operating characteristic test and Akaike information criteria metrics. We rejected the ENH for this case: biotic interactions improved the model. The frequency with which exceptions to the ENH are detected has important implications for distributional ecology and methods for estimating distributions of species.     

The mid‐domain effect matters: simulation analyses of range‐size distribution data from Mount Kinabalu,Borneo

Aim In simulation exercises, mid‐domain peaks in species richness arise as a result of the random placement of modelled species ranges within simulated geometric constraints. This has been called the mid‐domain effect (MDE). Where close correspondence is found between such simulations and empirical data, it is not possible to reject the hypothesis that empirical species richness patterns result from the MDE rather than being the outcome (wholly or largely) of other factors. To separate the influence of the MDE from other factors we therefore need to evaluate variables other than species richness. The distribution of range sizes gives different predictions between models including the MDE or not. Here, we produce predictions for species richness and distribution of range sizes from one model without the MDE and from two MDE models: a classical MDE model encompassing only species with their entire range within the domain (range‐restricted MDE), and a model encompassing all species with the theoretical midpoint within the domain (midpoint‐restricted MDE). These predictions are compared with observations from the elevational pattern of range‐size distributions and species richness of vascular plants. Location Mount Kinabalu, Borneo. Methods The data set analysed comprises more than 28,000 plant specimens with information on elevation. Species ranges are simulated with various assumptions for the three models, and the species simulated are subsequently subjected to a sampling that simulates the actual collection of species on Mount Kinabalu. The resulting pattern of species richness and species range‐size distributions are compared with the observed pattern. Results The comparison of simulated and observed patterns indicates that an underlying monotonically decreasing trend in species richness with elevation is essential to explain fully the observed pattern of richness and range size. When the underlying trend is accounted for, the MDE model that restricts the distributions of theoretical midpoints performs better than both the classical MDE model and the model that does not incorporate geometric constraints. Main conclusions Of the three models evaluated here, the midpoint‐restricted MDE model is found to be the best for explaining species richness and species range‐size distributions on Mount Kinabalu.     

Climatic associations of British species distributions show good transferability in time but low predictive accuracy for range change

Conservation planners often wish to predict how species distributions will change in response to environmental changes. Species distribution models (SDMs) are the primary tool for making such predictions. Many methods are widely used; however, they all make simplifying assumptions, and predictions can therefore be subject to high uncertainty. With global change well underway, field records of observed range shifts are increasingly being used for testing SDM transferability. We used an unprecedented distribution dataset documenting recent range changes of British vascular plants, birds, and butterflies to test whether correlative SDMs based on climate change provide useful approximations of potential distribution shifts. We modelled past species distributions from climate using nine single techniques and a consensus approach, and projected the geographical extent of these models to a more recent time period based on climate change; we then compared model predictions with recent observed distributions in order to estimate the temporal transferability and prediction accuracy of our models. We also evaluated the relative effect of methodological and taxonomic variation on the performance of SDMs. Models showed good transferability in time when assessed using widespread metrics of accuracy. However, models had low accuracy to predict where occupancy status changed between time periods, especially for declining species. Model performance varied greatly among species within major taxa, but there was also considerable variation among modelling frameworks. Past climatic associations of British species distributions retain a high explanatory power when transferred to recent time--due to their accuracy to predict large areas retained by species--but fail to capture relevant predictors of change. We strongly emphasize the need for caution when using SDMs to predict shifts in species distributions: high explanatory power on temporally-independent records--as assessed using widespread metrics--need not indicate a model's ability to predict the future.     

Mid-domain models as predictors of species diversity patterns: bathymetric diversity gradients in the deep sea

Geometric constraints represent a class of null models that describe how species diversity may vary between hard boundaries that limit geographic distributions. Recent studies have suggested that a number of large scale biogeographic patterns of diversity (e.g. latitude, altitude, depth) may reflect boundary constraints. However, few studies have rigorously tested the degree to which mid-domain null predictions match empirical patterns or how sensitive the null models are to various assumptions. We explore how variation in the assumptions of these models alter null depth ranges and consequently bathymetric variation in diversity, and test the extent to which bathymetric patterns of species diversity in deep sea gastropods, bivalves, and polychaetes match null predictions based on geometric constraints.
Range–size distributions and geographic patterns of diversity produced by these null models are sensitive to the relative position of the hard boundaries, the specific algorithms used to generate range sizes, and whether species are continuously or patchily distributed between range end points. How well empirical patterns support null expectations is highly dependent on these assumptions. Bathymetric patterns of species diversity for gastropods, bivalves and polychaetes differ substantially from null expectations suggesting that geometric constraints do not account for diversity–depth patterns in the deep sea benthos.     

Palaeodistribution modelling does not support disjunct Pleistocene refugia in several Central American plant taxa

Aim We combine evidence from palaeoniche modelling studies of several tree species to estimate the extent of Central American forest during the Last Glacial Maximum (LGM). In particular, we ask whether the distributions of these species are likely to have changed since the LGM, and whether LGM distributions coincide with previously proposed Pleistocene refugia in this area. Location Central American wet and seasonally dry forests. Methods We developed ecological niche models using two simulations of Pleistocene climate and occurrence data for 15 Neotropical plant species. We focused on palaeodistribution models of three ‘focal’ tree species that occur in wet and seasonally dry Central American forests, where recent phylogeographic data suggest Pleistocene differentiation coincident with previously proposed refugia. We added predictions from six wet‐forest and six seasonally dry‐forest obligate plant species to gauge whether Pleistocene range shifts were specific to habitat type. Correlation analyses were performed between projected LGM and present distributions, LGM distributions and previously proposed refugia. We also asked whether modelled palaeodistributions were smaller than their current extents. Results According to our models, the ranges of the study species were not reduced during the LGM, and did not correlate with refugial models, regardless of habitat type. Relative range sizes between present and LGM distributions did not indicate significant range changes since the LGM. However, relative range sizes differed overall between the two palaeoclimate models. Main conclusions Many of the modelled palaeodistributions of study species were not restricted to refugia during the LGM, regardless of forest type. While constrained from higher elevations, most species found suitable habitat at coastal margins and on newly exposed land due to lowered sea levels during the LGM. These results offer no corroboration for Pleistocene climate change as a driver of genetic differentiation in the ‘focal’ species. We offer alternative explanations for genetic differentiation found in plant species in this area.     

Is the Rapoport effect widespread? Null models revisited

Aim  To test the Rapoport effect using null models and data sets taken from the literature. We propose an improvement on an existing method, testing the Rapoport effect in elevational and latitudinal distributions when distributions are restricted by sampling.
Location  Global.
Methods  First, we hypothesized that real range size distributions are similar to those expected by null assumptions (expected by only imposing boundaries to species distributions). When these distributions were different from those expected under the null assumptions, we tested the hypothesis that these distributions correspond to those expected when a Rapoport effect occurs. We used two simulation methods, random and pseudo-random, which differed only in that the latter one assumes fixed species mid-points, coinciding with real mid-points. Observed correlations between range size and mid-point were compared with the frequency distribution of 1000 simulations, using both simulation methods. We compared the correlation curves generated by 1000 simulations with those of the observed distributions, testing whether correlations indicated a Rapoport effect.
Results  Several significant patterns of correlations between range size and mid-point were observed in the data sets when compared with random and pseudo-random simulations. However, few of these correlations were consistent with a Rapoport effect.
Main conclusions  Although some recent studies are consistent with a Rapoport effect, our results suggest that the Rapoport effect is not a widespread pattern in global ecology.     

Genetic trials improve the transfer of Douglas‐fir distribution models across continents

Climate change is likely to result in novel conditions with no analogy to current climate. Therefore, the application of species distribution models (SDMs) based on the correlation between observed species’ occurrence and their environment is questionable and calls for a better understanding of the traits that determine species occurrence. Here, we compared two intraspecific, trait‐based SDMs with occurrence‐based SDMs, all developed from European data, and analyzed their transferability to the native range of Douglas‐fir in North America. With data from 50 provenance trials of Douglas‐fir in central Europe multivariate universal response functions (URFs) were developed for two functional traits (dominant tree height and basal area) which are good indicators of growth and vitality under given environmental conditions. These trials included 290 North American provenances of Douglas‐fir. The URFs combine genetic effects i.e. the climate of provenance origin and environmental effects, i.e. the climate of planting locations into an integrated model to predict the respective functional trait from climate data. The URFs were applied as SDMs (URF‐SDMs) by converting growth performances into occurrence. For comparison, an ensemble occurrence‐based SDM was developed and block cross validated with the BIOMOD2 modeling platform utilizing the observed occurrence of Douglas‐fir in Europe. The two trait based SDMs and the occurrence‐based SDM, all calibrated with data from Europe, were applied to predict the known distribution of Douglas‐fir in its introduced and native range in Europe and North America. Both models performed well within their calibration range in Europe, but model transfer to its native range in North America was superior in case of the URF‐SDMs showing similar predictive power as SDMs developed in North America itself. The high transferability of the URF‐SDMs is a testimony of their applicability under novel climatic conditions highlighting the role of intraspecific trait variation for adaptation planning in climate change.     

Habitat disturbance selects against both small and large species across varying climates

Global extinction drivers, including habitat disturbance and climate change, are thought to affect larger species more than smaller species. However, it is unclear if such drivers interact to affect assemblage body size distributions. We asked how these two key global change drivers differentially affect the interspecific size distributions of ants, one of the most abundant and ubiquitous animal groups on earth. We also asked whether there is evidence of synergistic interactions and whether effects are related to species’ trophic roles. We generated a global dataset on ant body size from 333 local ant assemblages collected by the authors across a broad range of climates and in disturbed and undisturbed habitats. We used head length (range: 0.22–4.55 mm) as a surrogate of body size and classified species to trophic groups. We used generalized linear models to test whether body size distributions changed with climate and disturbance, independent of species richness. Our analysis yielded three key results: 1) climate and disturbance showed independent associations with body size; 2) assemblages included more small species in warmer climates and fewer large species in wet climates; and 3) both the largest and smallest species were absent from disturbed ecosystems, with predators most affected in both cases. Our results indicate that temperature, precipitation and disturbance have differing effects on the body size distributions of local communities, with no evidence of synergistic interactions. Further, both large and small predators may be vulnerable to global change, particularly through habitat disturbance.     

Quantifying species' range shifts in relation to climate change: a case study of Abies spp. in China

Predicting species range shifts in response to climatic change is a central aspect of global change studies. An ever growing number of species have been modeled using a variety of species distribution models (SDMs). However, quantitative studies of the characteristics of range shifts are rare, predictions of range changes are hard to interpret, analyze and summarize, and comparisons between the various models are difficult to make when the number of species modeled is large. Maxent was used to model the distribution of 12 Abies spp. in China under current and possible future climate conditions. Two fuzzy set defined indices, range increment index (I) and range overlapping index (O), were used to quantify range shifts of the chosen species. Correlation analyses were used to test the relationships between these indices and species distribution characteristics. Our results show that Abies spp. range increments (I) were highly correlated with longitude, latitude, and mean roughness of their current distributions. Species overlapping (O) was moderately, or not, correlated with these parameters. Neither range increments nor overlapping showed any correlation with species prevalence. These fuzzy sets defined indices provide ideal measures of species range shifts because they are stable and threshold-free. They are reliable indices that allow large numbers of species to be described, modeled, and compared on a variety of taxonomic levels.