Predicting Species Distributions Research Articles

Diversifying modelling techniques to disentangle the complex patterns of species richness and diversity in the protected afromontane grasslands

Ecological research has focused on the importance of environmental factors on spatial biodiversity variations and organisation. This is important because of scant conservation resources. We used stepwise backward selection and random feature selection (RFE) to identify a parsimonious model that can predict species richness and diversity metrics in response to three models; biotic, abiotic, and topo-edaphic. Our results show that both metrics are good predictors of one another, mainly because species diversity is a combination of species richness and abundance, and further highlights the importance of biotic variables in predicting species distribution. The two modelling techniques selected soil texture and its interactions with topographic variables as the most important variables. However, random forest performed worse than multiple linear regression in the prediction of diversity metrics. This research highlights the importance of topographically controlled edaphic factors as drivers of species richness and diversity in mountainous grasslands where topography inherently controls the geomorphic, hydrological, and, as a result, ecological processes.

MaxEnt Modeling for Predicting Suitable Habitat for Endangered Tree Keteleeria davidiana (Pinaceae) in China

Understanding species response to climate change is essential for the conservation and utilization of species resources under rapid climate change in the future. In this study, the present and future suitable distribution range of Keteleeria davidiana, a tertiary relict gymnosperm, was predicted based on the maximum entropy model (MaxEnt). A total of 158 occurrence records were collected after removing the duplicated records. Six low-correlation climate variables were used to predict species distributions. The three key climate factors that affect the distribution of K. davidiana were temperature seasonality (34.96%), mean temperature of the coldest quarter (28.30%) and precipitation seasonality (13.58%). The most suitable zone of the temperature seasonality for K. davidiana was between 377.4 and 843.4. The highly suitable area was located in the mountainous regions of central and southeast China, which accounted for 13.39% of the whole study area. With climate warming in the future, the highly suitable distribution area of K. davidiana was estimated to decrease by 35% (SSP1-2.6 scenario) or 85% (SSP5-8.5 scenario). This study has provided a sufficient scientific basis for the future in situ and ex situ conservation of K. davidiana.

Modeling climate‐driven range shifts in populations of two bird species limited by habitat independent of climate

AbstractRanges of species around the world are expected to contract in response to climate change. Species distribution models (SDMs) are a powerful tool for predicting changes in habitat availability, but the variables selected to create SDMs influence their performance. In addition to climate, habitat characteristics and species traits can play a role in predicting species distribution. In this paper, we consider how variable selection influences the accuracy of SDMs when applied to isolated subpopulations of two widely distributed bird species: great gray owl (Strix nebulosa) and willow flycatcher (Empidonax traillii). In the Sierra Nevada of California, these species are restricted largely to discrete patches of meadow habitat within a forest matrix, providing the potential to identify specific locations to target conservation efforts. We contrast predictions made by SDMs that consider climatic variables alone with those that incorporate both climate and geophysical variables. Adding geophysical variables resulted in differing model predictions. For willow flycatchers, adding geophysical variables improved predictive performance. In the case of great gray owls, models with and without geophysical variables had nearly identical performance under historical conditions but differed starkly in their predictions. The full model (climatic and geophysical variables) predicted habitat availability to decrease moderately, whereas the climate‐only model predicted nearly complete loss of favorable habitat by 2099. The climate‐only model is consistent with expectations based on previous SDMs of birds across North America, but previous studies also assumed homogeneity in species traits and range‐wide habitat requirements. The full model appears more consistent with recent trends in great gray owl numbers in the Sierra Nevada specifically, where the population has remained relatively stable over recent decades. Given contradictions in our model predictions, care should be taken when trying to apply similar SDMs to other systems.

Dealing with disjunct populations of vascular plants: implications for assessing the effect of climate change

Species distribution models are the most widely used tool to predict species distributions for species conservation and assessment of climate change impact. However, they usually do not consider intraspecific ecological variation exhibited by many species. Overlooking the potential differentiation among groups of populations may lead to misplacing any conservation actions. This issue may be particularly relevant in species in which few populations with potential local adaptation occur, as in species with disjunct populations. Here, we used ecological niche modeling to analyze how the projections of current and future climatically suitable areas of 12 plant species can be affected using the whole taxa occurrences compared to occurrences from geographically disjunct populations. Niche analyses suggest that usually the disjunct group of populations selects the climatic conditions as similar as possible to the other according to climate availability. Integrating intraspecific variability only slightly increases models’ ability to predict species occurrences. However, it results in different predictions of the magnitude of range change. In some species, integrating or not integrating intraspecific variability may lead to opposite trend in projected range change. Our results suggest that integrating intraspecific variability does not strongly improve overall models’ accuracy, but it can result in considerably different conclusions about future range change. Consequently, accounting for intraspecific differentiation may enable the detection of potential local adaptations to new climate and so to design targeted conservation strategies.

Tree ecophysiology in the context of climate change

Forest structure and function strongly depend on and concurrently influence environmental conditions. Tree performance is generally governed by its genetics and environment; thus, recent hotspots in this field include tree genotype × environment, phenotype × environment, and functional trait × environment interactions. The editorial, review, and 22 original research articles in this Special Issue, “Tree ecophysiology in the context of climate change”, highlight ecophysiological phenomena (e.g., climate hormesis, seed germination, tree mortality), processes (e.g., tree metabolism, photosynthate allocation, nutrient uptake and transport), indicators (e.g., carbon sequestration, pollutants), measurements (e.g., thermal time methods, soil quality indices, vegetation spectral index, and near-infrared leaf reflectance), and modeling (e.g., climate correlations with tree growth, photosynthetic phenology, hydraulic strategies, OliveCan model) in the context of global climate change. Understanding forest–environment interactions from an ecophysiological perspective as climate changes provides insights into species fitness in suboptimal environments, species competition for limited resources, and phylogenetic divergence or convergence of species, and predicting species distributions.

Expanded distribution and predicted suitable habitat for the critically endangered yellow-tailed woolly monkey (Lagothrix flavicauda) in Perú.

The Tropical Andes Biodiversity Hotspot holds a remarkable number of species at risk of extinction due to anthropogenic habitat loss, hunting, and climate change. One of these species, the critically endangered yellow-tailed woolly monkey (Lagothrix flavicauda), was recently observed in the region Junín, 206 kmsouth of its previously known distribution. This range extension, combined with continued habitat loss, calls for a reevaluation of the species distribution, and available suitable habitat. Here, we present novel data from surveys at 53 sites in the regions of Junín, Cerro de Pasco, Ayacucho, and Cusco. We encountered L. flavicauda at 9 sites, all in Junín, and the congeneric Lagothrix lagotricha tschudii at 20 sites, but never in sympatry. Using these new localities along with all previous geographic localities for the species, we made predictive species distribution models based on ecological niche modeling using a generalized linear model and maximum entropy. Each model incorporated bioclimatic variables, forest cover, vegetation measurements, and elevation as predictor variables. The model evaluation showed >80% accuracy for all measures. Precipitation was the strongest predictor of species presence. Habitat suitability maps illustrate potential corridors for gene flow between the southern and northern populations, although much of this area is inhabited by L. l. tschudii whereas L. flavicauda has yet to be officially confirmed in these areas, by these or any other scientific surveys. An analysis of the current protected area (PA) network showed that ~75% of remaining suitable habitat is unprotected. With this, we suggest priority areas for new PAs or expansions to existing reserves that would conserve potential corridors between L. flavicauda populations. Further surveys and characterization of the distribution in intermediate areas, combined with studies on gene flow through these areas, are still needed to protect this species.

Empowering ecological modellers with a PERFICT workflow: Seamlessly linking data, parameterisation, prediction, validation and visualisation

Abstract Modelling is widely used in ecology and its utility continues to increase as scientists, managers and policy‐makers face pressure to effectively manage ecosystems and meet conservation goals with limited resources. As the urgency to forecast ecosystem responses to global change grows, so do the number and complexity of predictive ecological models and the value of iterative prediction, both of which demand validation and cross‐model comparisons. This challenges ecologists to provide predictive models that are reusable, interoperable, transparent and able to accommodate updates to both data and algorithms. We propose a practical solution to this challenge based on the PERFICT principles (frequent Predictions and Evaluations of Reusable, Freely accessible, Interoperable models, built within Continuous workflows that are routinely Tested), using a modular and integrated framework. We present its general implementation across seven common components of ecological model applications—(i) the modelling toolkit; (ii) data acquisition and treatment; (iii) model parameterisation and calibration; (iv) obtaining predictions; (v) model validation; (vi) analysing and presenting model outputs; and (vii) testing model code—and apply it to two approaches used to predict species distributions: (1) a static statistical model, and (2) a complex spatiotemporally dynamic model. Adopting a continuous workflow enabled us to reuse our models in new study areas, update predictions with new data, and re‐parameterise with different interoperable modules using freely accessible data sources, all with minimal user input. This allowed repeating predictions and automatically evaluating their quality, while centralised inputs, parameters and outputs, facilitated ensemble forecasting and tracking uncertainty. Importantly, the integrated model validation promotes a continuous evaluation of the quality of more‐ or less‐parsimonious models, which is valuable in predictive ecological modelling. By linking all stages of an ecological modelling exercise, it is possible to overcome common challenges faced by ecological modellers, such as changing study areas, choosing between different modelling approaches, and evaluating the appropriateness of the model. This ultimately creates a more equitable and robust playing field for both modellers and end users (e.g. managers), and contributes to position predictive ecology as a central contributor to global change forecasting.

Combining tracking with at-sea surveys to improve occurrence and distribution estimates of two threatened seabirds in Peru

SummarySeabirds are highly threatened, including by fisheries bycatch. Accurate understanding of offshore distribution of seabirds is crucial to address this threat. Tracking technologies revolutionised insights into seabird distributions but tracking data may contain a variety of biases. We tracked two threatened seabirds (Salvin’s AlbatrossThalassarche salvini n= 60 and Black PetrelProcellaria parkinsoni n= 46) from their breeding colonies in Aotearoa (New Zealand) to their non-breeding grounds in South America, including Peru, while simultaneously completing seven surveys in Peruvian waters. We then used species distribution models to predict occurrence and distribution using either data source alone, and both data sources combined. Results showed seasonal differences between estimates of occurrence and distribution when using data sources independently. Combining data resulted in more balanced insights into occurrence and distributions, and reduced uncertainty. Most notably, both species were predicted to occur in Peruvian waters during all four annual quarters: the northern Humboldt upwelling system for Salvin’s Albatross and northern continental shelf waters for Black Petrels. Our results highlighted that relying on a single data source may introduce biases into distribution estimates. Our tracking data might have contained ontological and/or colony-related biases (e.g. only breeding adults from one colony were tracked), while our survey data might have contained spatiotemporal biases (e.g. surveys were limited to waters <200 nm from the coast). We recommend combining data sources wherever possible to refine predictions of species distributions, which ultimately will improve fisheries bycatch management through better spatiotemporal understanding of risks.

Thermal Limits Determination for Zooplankton Using a Heat Block.

Thermal limits and breadth have been widely used to predict species distribution. As the global temperature continues to rise, understanding how thermal limit changes with acclimation and how it varies between life stages and populations are vital for determining the vulnerability of species to future warming. Most marine organisms have complex life cycles that include early planktonic stages. While quantifying the thermal limit of these small early developmental stages (tens to hundreds of microns) helps identify developmental bottlenecks, this process can be challenging due to the small size of target organisms, large bench space requirement, and high initial fabrication cost. Here, a setup that is geared toward small volumes (mL to tens of mL) is presented. This setup combines commercially available components to generate a stable and linear thermal gradient. Production specifications of the setup, as well as procedures to introduce and enumerate live versus dead individuals and compute lethal temperature, are also presented.

Integrating Different Scales into Species Distribution Models: A Case for Evaluating the Risk of Plant Invasion in Chinese Protected Areas under Climate Change

Species distribution models (SDMs) based on fine-scale environmental data may reduce the uncertainty in predicting species distributions. However, many scientists have also projected the robust potential distributions of species using environmental data of different scales and found that the potential distributions modeled using SDMs are scale dependent. This may be due to the impact of the scale effect on species richness (as well as on multi-species distributions). To eliminate the impact of the scale effect, we aim to develop an improved method to integrate different scales into species distribution models. We use protected areas as the study regions and propose the hypothesis that there is a spatial element to the threat of invasive species for protected areas under climate change. We use Maxent to compute the current and future invasion ability and invasion inequality of invasive species for protected areas based on the potential distributions of species across different scales to evaluate the risk of invasive species. We find that an increase in the number of present records could reduce the accuracy of SDMs. There is a significant linear relationship between the fine-scale and coarse-scale risk of invasive species of alien plants in protected areas, and an appropriate scale should thus be selected to assess species risk based on this linear relationship of invasive risk. There is a significant relationship between the potential of IAPS to invade protected areas and the invasion inequality of IAPS in protected areas across all scales, and 5.0 arcminutes is the most appreciate scale to evaluate the risk of IAPS for protected areas under climate change based on principal component analysis. We provide new insights into the use of species distribution models coupled with different spatial scales to analyze the regional risks associated with species and to assess regional biodiversity.

A Novel Multimodal Species Distribution Model Fusing Remote Sensing Images and Environmental Features

Species distribution models (SDMs) are critical in conservation decision-making and ecological or biogeographical inference. Accurately predicting species distribution can facilitate resource monitoring and management for sustainable regional development. Currently, species distribution models usually use a single source of information as input for the model. To determine a solution to the lack of accuracy of the species distribution model with a single information source, we propose a multimodal species distribution model that can input multiple information sources simultaneously. We used ResNet50 and Transformer network structures as the backbone for multimodal data modeling. The model’s accuracy was tested using the GEOLIFE2020 dataset, and our model’s accuracy is state-of-the-art (SOTA). We found that the prediction accuracy of the multimodal species distribution model with multiple data sources of remote sensing images, environmental variables, and latitude and longitude information as inputs (29.56%) was higher than that of the model with only remote sensing images or environmental variables as inputs (25.72% and 21.68%, respectively). We also found that using a Transformer network structure to fuse data from multiple sources can significantly improve the accuracy of multimodal models. We present a novel multimodal model that fuses multiple sources of information as input for species distribution prediction to advance the research progress of multimodal models in the field of ecology.

Plant species biogeographic origin shapes their current and future distribution on the world's highest island mountain

Abstract In New Guinea, low elevations are thought to be dominated by plant species with biogeographic origins in Sunda (Borneo, Sumatra, Java, Mainland Asia) and high elevations by species from Sahul (Australia and New Guinea). In regions where frequent surveys of the flora are unfeasible, herbarium specimens collected over time offer an opportunity to study the co‐occurrence of species. Over 7000 collections have been made on New Guinea's Mount Jaya allowing us to examine species presence under current and future temperatures. Applying a Bayesian phylogenetic mixed effects model to the Mount Jaya elevation gradient, we show that species of Sunda origin are more likely to occupy warmer lowlands and Sahul species cooler higher elevations. Using the model to predict species distributions under a 2°C increase, we find that Sahul species are less likely to tolerate future temperatures, especially below 1000 m. Predictions also show a potential increase in species richness across the gradient. It is unclear if these increases are possible due to limits imposed by the species area relationship. Synthesis. These results emphasise how biogeographic origin influences current and future species distributions on tropical mountains. It remains to be seen how our results link to traits that dictate species' ability to tolerate increasing temperatures.

Assessing the effect of sample bias correction in species distribution models

1. Open-source biodiversity databases contain a large number of species occurrence records but are often spatially biased; which affects the reliability of species distribution models based on these records. Sample bias correction techniques require data filtering which comes at the cost of record numbers, or require considerable additional sampling effort. Since independent data is rarely available, assessment of the correction technique often relies solely on performance metrics computed using subsets of the available – biased – data, which may prove misleading.2. Here, we assess the extent to which an acknowledged sample bias correction technique is likely to improve models’ ability to predict species distributions in the absence of independent data. We assessed variation in model predictions induced by the aforementioned correction and model stochasticity; the variability between model replicates related to a random component (pseudo-absences sets and cross-validation subsets). We present, then, an index of the effect of correction relative to model stochasticity; the Relative Overlap Index (ROI). We investigated whether the ROI better represented the effect of correction than classic performance metrics (Boyce index, cAUC, AUC and TSS) and absolute overlap metrics (Schoener’s D, Pearson’s and Spearman’s correlation coefficients) when considering data related to 64 vertebrate species and 21 virtual species with a generated sample bias.3. When based on absolute overlaps and cross-validation performance metrics, we found that correction produced no significant effects. When considering its effect relative to model stochasticity, the effect of correction was strong for most species at one of the three sites. The use of virtual species enabled us to verify that the correction technique improved both distribution predictions and the biological relevance of the selected variables at the specific site, when these were not correlated with sample bias patterns.4. In the absence of additional independent data, the assessment of sample bias correction based on subsample data may be misleading. We propose to investigate both the biological relevance of environmental variables selected, and, the effect of sample bias correction based on its effect relative to model stochasticity.

Predicting the changes in suitable habitats for six common woody species in Central Asia.

Climate change over the past decades has significantly altered global hydrothermal conditions and caused an evident shift in species distribution. Predicting species distribution patterns and identifying their influencing factors will be essential in developing coping strategies to prevent species extirpation and extinction. Yet, environmental factors affecting the distribution of woody species in Central Asia remain largely unknown. Here, I used the MaxEnt model to predict the current distributions and future distribution under three SSP-RCP scenarios of six common woody species in Central Asia. The results indicated a good performance of the MaxEnt model. Precipitation of driest month and annual mean temperature were the dominant factors affecting species distribution. For the species with wide ecological niches, i.e., Acer negundo and Rosa chinensis, the suitable areas showed an evident expansion trend under future scenarios. In addition, a trend toward higher elevation was found for the species that grew at high altitudes (1600-3200m). However, the average elevation of suitable area for A. negundo and R. chinensis firstly increased but then decreased under future scenarios. Even though the areas with high species diversity increased from 0.59% under the current situation to 0.82% and 0.81% under ssp245 in 2021-2040 and 2041-2060, respectively, species diversity showed an apparent loss in parts of the northwest and southeast areas under ssp370 and ssp585. This study can guide susceptible habitat protections under climate change.

Modeled distribution shifts of North American birds over four decades based on suitable climate alone do not predict observed shifts

As climate change alters the global environment, it is critical to understand the relationship between shifting climate suitability and species distributions. Key questions include whether observed changes in population abundance are aligned with the velocity and direction of shifts predicted by climate suitability models and if the responses are consistent among species with similar ecological traits. We examined the direction and velocity of the observed abundance-based distribution centroids compared with the model-predicted bioclimatic distribution centroids of 250 bird species across the United States from 1969 to 2011. We hypothesized that there is a significant positive correlation in both direction and velocity between the observed and the modeled shifts. We then tested five additional hypotheses that predicted differential shifting velocity based on ecological adaptability and climate change exposure. Contrary to our hypotheses, we found large differences between the observed and modeled shifts among all studied bird species and within specific ecological guilds. However, temperate migrants and habitat generalist species tended to have higher velocity of observed shifts than other species. Neotropical migratory and wetland birds also had significantly different observed velocities than their counterparts, which may be due to their climate change exposure. The velocity based on modeled bioclimatic suitability did not exhibit significant differences among most guilds. Boreal forest birds were the only guild with significantly faster modeled-shifts than the other groups, suggesting an elevated conservation risk for high latitude and altitude species. The highly idiosyncratic species responses to climate and the mismatch between shifts in modeled and observed distribution centroids highlight the challenge of predicting species distribution change based solely on climate suitability and the importance of non-climatic factors traits in shaping species distributions.

Projecting species distributions using fishery‐dependent data

AbstractMany marine species are shifting their distributions in response to changing ocean conditions, posing significant challenges and risks for fisheries management. Species distribution models (SDMs) are used to project future species distributions in the face of a changing climate. Information to fit SDMs generally comes from two main sources: fishery‐independent (scientific surveys) and fishery‐dependent (commercial catch) data. A concern with fishery‐dependent data is that fishing locations are not independent of the underlying species abundance, potentially biasing predictions of species distributions. However, resources for fishery‐independent surveys are increasingly limited; therefore, it is critical we understand the strengths and limitations of SDMs developed from fishery‐dependent data. We used a simulation approach to evaluate the potential for fishery‐dependent data to inform SDMs and abundance estimates and quantify the bias resulting from different fishery‐dependent sampling scenarios in the California Current System (CCS). We then evaluated the ability of the SDMs to project changes in the spatial distribution of species over time and compare the time scale over which model performance degrades between the different sampling scenarios and as a function of climate bias and novelty. Our results show that data generated from fishery‐dependent sampling can still result in SDMs with high predictive skill several decades into the future, given specific forms of preferential sampling which result in low climate bias and novelty. Therefore, fishery‐dependent data may be able to supplement information from surveys that are reduced or eliminated for budgetary reasons to project species distributions into the future.

Realizing certainty in an uncertain future climate: modeling suitable areas for conserving wild Citrus species under different change scenarios in India.

Citrus is an important horticultural crop of India and is often prone to diseases, particularly under increased temperature scenarios. For developing disease-resistant Citrus varieties, conservation of wild relatives is extremely important. However, our knowledge on temperature tolerance of these wild relatives of Citrus to varied climate change scenarios is extremely limited. Therefore, we determined the climatic niche of six wild relatives of cultivated Citrus species (C. indica Tanaka, C. karna Rafin., C. latipes (Swingle) Tanaka, C. macroptera Montrouz., C. medica L., and C. sinensis (L.) Osbeck.) and identified the geographical areas in India that would remain climatically stable in future through ecological niche modeling (ENM). Raster data on 19 bioclimatic variables with a resolution of 0.04° were used to generate niche models for each Citrus species that delineated their potential distribution areas. Future species distribution predictions for the year 2050 were made using the climate change scenarios from the most appropriate climate models, i.e., IPSL-CM5A-LR and NIMR-HADGEM2-AO with four Representative Concentration Pathways (RCPs). Ensemble of current and future projections was used to identify climatically stable areas for each species. Precipitation-related bioclimatic variables were the key climatic determinants for the modeled distribution pattern. The consensus of current and future projections suggests that most areas with stable climates for the species in the future would be available in the northeastern states of Arunachal Pradesh, Meghalaya, Mizoram, and Tripura. Efforts for in situ conservation and establishment of germplasm banks and citrus orchards may be encouraged in these identified areas.

Co-occurrences enhance our understanding of aquatic fungal metacommunity assembly and reveal potential host-parasite interactions.

Our knowledge of aquatic fungal communities, their assembly, distributions and ecological roles in marine ecosystems is scarce. Hence, we aimed to investigate fungal metacommunities of coastal habitats in a subarctic zone (northern Baltic Sea, Sweden). Using a novel joint species distribution model and network approach, we quantified the importance of biotic associations contributing to the assembly of mycoplankton, further, detected potential biotic interactions between fungi-algae pairs, respectively. Our long-read metabarcoding approach identified 493 fungal taxa, of which a dominant fraction (44.4%) was assigned as early-diverging fungi (i.e. Cryptomycota and Chytridiomycota). Alpha diversity of mycoplankton declined and community compositions changed along inlet-bay-offshore transects. The distributions of most fungi were rather influenced by environmental factors than by spatial drivers, and the influence of biotic associations was pronounced when environmental filtering was weak. We found great number of co-occurrences (120) among the dominant fungal groups, and the 25 associations between fungal and algal OTUs suggested potential host-parasite and/or saprotroph links, supporting a Cryptomycota-based mycoloop pathway. We emphasize that the contribution of biotic associations to mycoplankton assembly are important to consider in future studies as it helps to improve predictions of species distributions in aquatic ecosystems.

Tree species diversity analysis using species distribution models: A Faidherbia albida parkland case study in Senegal

Around the world, SDMs have been widely used to support forest management planning and biodiversity conservation. Beyond the prediction of species distribution provided by the SDMs, this study aimed to analyze the spatial distribution of tree species diversity using SDMs. The study area is a Faidherbia albida parkland in Central Senegal. It is characterized by a tree-based farming system dominated by Faidherbia albida.Using a robust and representative dataset of 9258 tree species occurrence, we first determined by an SDM the current potential spatial distribution of the 16 main tree species forming the parkland. Specifically, using 6 SDM algorithms and applying several modeling techniques with different categories of predictor variables (e.g., climate, topography, soil properties and human impact) we benchmarked 576 SDMs to achieve best model predictions for tree species. Then, tree species diversity maps were created on the basis of the resulting SDM predictions. Finally, the spatial dynamics of tree species diversity were discussed in relation to landscape characteristics, including heterogeneity, composition and human impact.The results showed that there is no single ‘best’ SDM algorithm (among the 6 algorithms tested) or modeling approach for all species. Benchmarking several modeling techniques allowed strengthening SDM performance, achieving AUC values that ranged from 0.64 (intermediate accuracy) to 0.87 (very good accuracy). The spatial dynamics of tree species diversity is related to the landscape heterogeneity and composition. In the Sahelian agroforestry systems (AFS), tree diversity is sustained by anthropization. A significant negative correlation with the distance to the village was found, i.e. the closer you get to the village, the greater the diversity of trees.This study could be crucial for analyzing tree species diversity when abundance information is not available.

Transferable species distribution modelling: Comparative performance of Generalised Functional Response models

Predictive species distribution models (SDMs) are becoming increasingly important in ecology, in the light of rapid environmental change. However, the predictions of most current SDMs are specific to the habitat composition of the environments in which they were fitted. This may limit SDM predictive power because species may respond differently to a given habitat depending on the availability of all habitats in their environment, a phenomenon known as a functional response in resource selection. The Generalised Functional Response (GFR) framework captures this dependence by formulating the SDM coefficients as functions of habitat availability. The original GFR implementation used global polynomial functions of habitat availability to describe the functional responses. In this study, we develop several refinements of this approach and compare their predictive performance using two simulated and two real datasets. We first use local radial basis functions (RBF), a more flexible approach than global polynomials, to represent the habitat selection coefficients, and balance bias with precision via regularization to prevent overfitting. Second, we use the RBF-GFR and GFR models in combination with the classification and regression tree CART, which has more flexibility and better predictive powers for non-linear modelling. As further extensions, we use random forests (RFs) and extreme gradient boosting (XGBoost), ensemble approaches that consistently lead to variance reduction in generalization error. We find that the different methods are ranked consistently across the datasets for out-of-data prediction. The traditional stationary approach to SDMs and the GFR model consistently perform at the bottom of the ranking (simple SDMs underfit, and polynomial GFRs overfit the data). The best methods in our list provide non-negligible improvements in predictive performance, in some cases taking the out-of-sample R2 from 0.3 up to 0.7 across datasets. At times of rapid environmental change and spatial non-stationarity ignoring the effects of functional responses on SDMs, results in two different types of prediction bias (under-prediction or mis-positioning of distribution hotspots). However, not all functional response models perform equally well. The more volatile polynomial GFR models can generate biases through over-prediction. Our results indicate that there are consistently robust GFR approaches that achieve impressive gains in transferability across very different datasets.